Training Results Master Summary

Executive Snapshot

• Generated At: 2026-06-22T17:01:10

• Program State: active.

• Current Completed Wave: polished_dataset Stage 1 smoke campaign closeout

• Current Focus: prepare the next polished-dataset retraining stage; keep TE Curve Verification Pipeline refresh separate

• Active Campaign Status: none

• Active Campaign Name: N/A

• Current Program Winner: te_periodic_gru_sequence_remote_global | Family periodic_gru_sequence | Test MAE 0.001279

Main Takeaways

• Strongest current neural family: periodic_gru_sequence

• Current plain MLP anchor: te_feedforward_trial

• Active family-improvement branch count: 0

• Implemented and benchmarked family count: 107

Current Project Status

Implemented And Benchmarked Families

• Multi-scope waves must keep global, Fw, and Bw reporting surfaces separated in this canonical summary.

Global Models

Family	Current Role	Best Run	Model Type	Test MAE [deg]	Params	Last Update
periodic_gru_sequence	Current Program Winner	te_periodic_gru_sequence_remote_global	periodic_gru_sequence	0.001279	157,569	2026-06-22 14:06:48
tree	Implemented Benchmark	te_hist_gbr_tabular_global	hist_gradient_boosting	0.001753	4	2026-06-22 16:58:47
track2g_curve_aware_harmonic_residual_offset_full_curve_composite_global	Implemented Benchmark	te_track2g_curve_aware_full_curve_composite_global	curve_aware_harmonic_residual_offset_probe	0.002008	85,440	2026-06-22 15:16:05
residual_harmonic_gru_sequence_sparse_rcim	Implemented Benchmark	te_residual_harmonic_gru_sequence_remote_global_sparse_rcim	residual_harmonic_gru_sequence	0.002112	150,676	2026-06-22 14:31:08
wave3_harmonic_prior_residual_smooth_l1_structured_global	Implemented Benchmark	te_wave3_harmonic_prior_residual_smooth_l1_structured_global	wave3_harmonic_prior_residual	0.002168	7,168	2026-06-22 15:37:36
track2h_latent_state_hysteresis_gru_offset_residual_global	Implemented Benchmark	te_track2h_l_gru_offset_residual_global	latent_state_hysteresis_probe	0.002339	124,899	2026-06-22 16:00:39
periodic_lstm_sequence	Implemented Benchmark	te_periodic_lstm_sequence_remote_global	periodic_lstm_sequence	0.002682	210,561	2026-05-25 19:20:56
feedforward	Current Plain MLP Anchor	te_feedforward_trial	feedforward	0.002877	26,113	2026-06-22 13:13:25
track2h_quantile_probabilistic_gaussian_nll_global	Implemented Benchmark	te_track2h_gaussian_nll_global	curve_aware_harmonic_residual_offset_probe	0.003013	85,958	2026-06-12 13:12:35
residual_harmonic_mlp	Implemented Benchmark	te_residual_h12_deep_joint_wave1_global_optuna_t0006	residual_harmonic_mlp	0.003034	26,266	2026-05-20 11:41:03
periodic_mlp	Implemented Benchmark	te_periodic_mlp_h04_standard_global_optuna_t0010	periodic_mlp	0.003186	27,265	2026-05-21 08:12:57
track2g_curve_aware_harmonic_residual_offset_raw_centered_shape_global	Implemented Benchmark	te_track2g_curve_aware_raw_centered_shape_global	curve_aware_harmonic_residual_offset_probe	0.003350	85,747	2026-06-08 19:45:16
track2h_latent_state_hysteresis_causal_tcn_offset_residual_global	Implemented Benchmark	te_track2h_l_causal_tcn_offset_residual_global	latent_state_hysteresis_probe	0.003368	97,923	2026-06-16 19:16:49
residual_harmonic_lstm_sequence_sparse_rcim	Implemented Benchmark	te_residual_harmonic_lstm_sequence_remote_global_sparse_rcim	residual_harmonic_lstm_sequence	0.003368	201,364	2026-05-27 20:55:58
track2h_quantile_probabilistic_quantile_p10_p50_p90_global	Implemented Benchmark	te_track2h_quantile_p10_p50_p90_global	curve_aware_harmonic_residual_offset_probe	0.003383	86,169	2026-06-12 11:34:56
track2h_dispersion_aware_mae_robust_global	Implemented Benchmark	te_track2h_mae_robust_global	curve_aware_harmonic_residual_offset_probe	0.003406	85,747	2026-06-11 12:07:43
track2h_dispersion_aware_smooth_l1_robust_global	Implemented Benchmark	te_track2h_smooth_l1_robust_global	curve_aware_harmonic_residual_offset_probe	0.003422	85,747	2026-06-11 12:48:44
wave3_harmonic_prior_residual_pointwise_control_global	Implemented Benchmark	te_wave3_harmonic_prior_residual_pointwise_control_global	wave3_harmonic_prior_residual	0.003451	7,283	2026-06-15 14:27:23
track2g_curve_aware_harmonic_residual_offset_raw_offset_global	Implemented Benchmark	te_track2g_curve_aware_raw_offset_global	curve_aware_harmonic_residual_offset_probe	0.003465	85,747	2026-06-08 20:43:53
residual_harmonic_lstm_sequence_dense240	Implemented Benchmark	te_residual_harmonic_lstm_sequence_remote_global_dense240	residual_harmonic_lstm_sequence	0.003473	201,826	2026-05-27 21:22:30
residual_harmonic_lstm_sequence_dense360	Implemented Benchmark	te_residual_harmonic_lstm_sequence_remote_global_dense360	residual_harmonic_lstm_sequence	0.003477	202,066	2026-05-27 22:09:01
lstm_sequence	Implemented Benchmark	te_lstm_sequence_remote_global	lstm_sequence	0.003482	201,345	2026-05-24 12:16:30
track2h_mixture_density_heads_mdn_k2_global	Implemented Benchmark	te_track2h_mdn_k2_global	curve_aware_harmonic_residual_offset_probe	0.003503	86,802	2026-06-13 11:32:06
track2h_dispersion_aware_log_cosh_robust_global	Implemented Benchmark	te_track2h_log_cosh_robust_global	curve_aware_harmonic_residual_offset_probe	0.003505	85,747	2026-06-11 13:43:04
periodic_temporal_convolution	Implemented Benchmark	te_periodic_temporal_convolution_sequence_remote_global	periodic_temporal_convolution	0.003508	158,529	2026-05-25 16:10:13
residual_harmonic_gru_sequence_dense240	Implemented Benchmark	te_residual_harmonic_gru_sequence_remote_global_dense240	residual_harmonic_gru_sequence	0.003511	151,522	2026-05-27 19:32:17
track2f_bis_clean_sequential_residual_offset_global	Implemented Benchmark	te_track2f_bis_clean_residual_offset_global	sequential_residual_offset_probe	0.003528	92,802	2026-06-04 23:43:38
residual_harmonic_gru_sequence_dense360	Implemented Benchmark	te_residual_harmonic_gru_sequence_remote_global_dense360	residual_harmonic_gru_sequence	0.003535	151,762	2026-05-27 20:21:50
sequential_residual_offset_probe	Implemented Benchmark	te_sequential_residual_offset_probe_remote_global	sequential_residual_offset_probe	0.003537	92,802	2026-06-04 11:45:31
track2f_bis_harmonic_residual_offset_global	Implemented Benchmark	te_track2f_bis_harmonic_residual_offset_global	harmonic_residual_offset_probe	0.003538	85,747	2026-06-05 16:19:21
track2h_mixture_density_heads_mdn_k3_global	Implemented Benchmark	te_track2h_mdn_k3_global	curve_aware_harmonic_residual_offset_probe	0.003564	87,435	2026-06-13 12:34:37
track2g_curve_aware_harmonic_residual_offset_pointwise_control_global	Implemented Benchmark	te_track2g_curve_aware_pointwise_control_global	curve_aware_harmonic_residual_offset_probe	0.003587	85,747	2026-06-08 18:56:59
gru_sequence	Implemented Benchmark	te_gru_sequence_remote_global	gru_sequence	0.003591	151,041	2026-05-24 11:54:03
temporal_convolution	Implemented Benchmark	te_temporal_convolution_sequence_remote_global	temporal_convolution	0.003754	147,009	2026-05-24 11:30:23
harmonic_regression	Implemented Benchmark	te_harmonic_order12_linear_conditioned_recovery_global	harmonic_regression	0.003839	125	2026-06-22 13:26:44
feedforward_recovery_micro	Implemented Benchmark	te_feedforward_optuna_recovery_micro_global_optuna_t0000	feedforward	0.004164	109,953	2026-05-12 11:12:51
feedforward_recovery_probe_dense	Implemented Benchmark	te_feedforward_optuna_recovery_probe_dense_global_optuna_t0000	feedforward	0.004602	109,953	2026-05-12 17:16:41

Forward Models

Family	Current Role	Best Run	Model Type	Test MAE [deg]	Params	Last Update
tree_fw	Implemented Benchmark	te_hist_gbr_tabular_Fw_grid_depth6_lr008_leaf10	hist_gradient_boosting	0.002743	5	2026-05-11 20:58:32
track2f_bis_harmonic_residual_offset_fw	Implemented Benchmark	te_track2f_bis_harmonic_residual_offset_fw	harmonic_residual_offset_probe	0.002862	85,747	2026-06-05 16:32:38
harmonic_regression_fw	Implemented Benchmark	te_harmonic_dense360_tracking_Fw	harmonic_regression	0.002916	4,326	2026-05-20 10:50:22
periodic_mlp_fw	Implemented Benchmark	te_periodic_mlp_dense240_tracking_Fw	periodic_mlp	0.003055	87,681	2026-05-21 08:48:01
residual_harmonic_mlp_fw	Implemented Benchmark	te_residual_harmonic_rcim_sparse_tracking_Fw	residual_harmonic_mlp	0.003089	26,260	2026-05-20 11:57:15
track2h_dispersion_aware_mae_robust_fw	Implemented Benchmark	te_track2h_mae_robust_fw	curve_aware_harmonic_residual_offset_probe	0.003146	85,747	2026-06-11 12:14:52
track2h_quantile_probabilistic_gaussian_nll_fw	Implemented Benchmark	te_track2h_gaussian_nll_fw	curve_aware_harmonic_residual_offset_probe	0.003165	85,958	2026-06-12 13:25:53
track2g_curve_aware_harmonic_residual_offset_raw_centered_shape_fw	Implemented Benchmark	te_track2g_curve_aware_raw_centered_shape_fw	curve_aware_harmonic_residual_offset_probe	0.003181	85,747	2026-06-08 19:56:04
periodic_gru_sequence_fw	Implemented Benchmark	te_periodic_gru_sequence_remote_Fw	periodic_gru_sequence	0.003193	157,953	2026-05-25 17:38:18
residual_harmonic_gru_sequence_fw_sparse_rcim	Implemented Benchmark	te_residual_harmonic_gru_sequence_remote_Fw_sparse_rcim	residual_harmonic_gru_sequence	0.003200	151,060	2026-05-27 19:12:38
feedforward_fw	Implemented Benchmark	te_feedforward_stride1_high_compute_long_remote_Fw_optuna_t0008	feedforward	0.003203	109,953	2026-05-14 22:03:06
residual_harmonic_gru_sequence_fw_dense240	Implemented Benchmark	te_residual_harmonic_gru_sequence_remote_Fw_dense240	residual_harmonic_gru_sequence	0.003219	151,522	2026-05-27 19:40:30
residual_harmonic_lstm_sequence_fw_sparse_rcim	Implemented Benchmark	te_residual_harmonic_lstm_sequence_remote_Fw_sparse_rcim	residual_harmonic_lstm_sequence	0.003234	201,364	2026-05-27 21:00:48
track2h_mixture_density_heads_mdn_k3_fw	Implemented Benchmark	te_track2h_mdn_k3_fw	curve_aware_harmonic_residual_offset_probe	0.003235	87,435	2026-06-13 12:43:18
residual_harmonic_gru_sequence_fw_dense360	Implemented Benchmark	te_residual_harmonic_gru_sequence_remote_Fw_dense360	residual_harmonic_gru_sequence	0.003241	151,762	2026-05-27 20:33:03
track2g_curve_aware_harmonic_residual_offset_full_curve_composite_fw	Implemented Benchmark	te_track2g_curve_aware_full_curve_composite_fw	curve_aware_harmonic_residual_offset_probe	0.003260	85,747	2026-06-08 21:49:46
residual_harmonic_lstm_sequence_fw_dense240	Implemented Benchmark	te_residual_harmonic_lstm_sequence_remote_Fw_dense240	residual_harmonic_lstm_sequence	0.003262	201,826	2026-05-27 21:29:55
periodic_lstm_sequence_fw	Implemented Benchmark	te_periodic_lstm_sequence_remote_Fw	periodic_lstm_sequence	0.003274	210,561	2026-05-25 19:30:17
track2g_curve_aware_harmonic_residual_offset_raw_offset_fw	Implemented Benchmark	te_track2g_curve_aware_raw_offset_fw	curve_aware_harmonic_residual_offset_probe	0.003279	85,747	2026-06-08 20:51:34
track2h_quantile_probabilistic_quantile_p10_p50_p90_fw	Implemented Benchmark	te_track2h_quantile_p10_p50_p90_fw	curve_aware_harmonic_residual_offset_probe	0.003285	86,169	2026-06-12 11:43:50
track2h_dispersion_aware_smooth_l1_robust_fw	Implemented Benchmark	te_track2h_smooth_l1_robust_fw	curve_aware_harmonic_residual_offset_probe	0.003314	85,747	2026-06-11 12:56:26
gru_sequence_fw	Implemented Benchmark	te_gru_sequence_remote_Fw	gru_sequence	0.003333	151,041	2026-05-24 12:00:05
periodic_temporal_convolution_fw	Implemented Benchmark	te_periodic_temporal_convolution_sequence_remote_Fw	periodic_temporal_convolution	0.003337	158,529	2026-05-25 16:18:28
track2h_mixture_density_heads_mdn_k2_fw	Implemented Benchmark	te_track2h_mdn_k2_fw	curve_aware_harmonic_residual_offset_probe	0.003339	86,802	2026-06-13 11:41:07
residual_harmonic_lstm_sequence_fw_dense360	Implemented Benchmark	te_residual_harmonic_lstm_sequence_remote_Fw_dense360	residual_harmonic_lstm_sequence	0.003351	202,066	2026-05-27 22:19:22
track2h_dispersion_aware_log_cosh_robust_fw	Implemented Benchmark	te_track2h_log_cosh_robust_fw	curve_aware_harmonic_residual_offset_probe	0.003355	85,747	2026-06-11 13:51:00
lstm_sequence_fw	Implemented Benchmark	te_lstm_sequence_remote_Fw	lstm_sequence	0.003370	201,345	2026-05-24 12:21:01
track2g_curve_aware_harmonic_residual_offset_pointwise_control_fw	Implemented Benchmark	te_track2g_curve_aware_pointwise_control_fw	curve_aware_harmonic_residual_offset_probe	0.003371	85,747	2026-06-08 19:08:39
wave3_harmonic_prior_residual_pointwise_control_fw	Implemented Benchmark	te_wave3_harmonic_prior_residual_pointwise_control_fw	wave3_harmonic_prior_residual	0.003382	7,283	2026-06-15 14:34:34
sequential_residual_offset_probe_fw	Implemented Benchmark	te_sequential_residual_offset_probe_remote_fw	sequential_residual_offset_probe	0.003385	92,802	2026-06-04 11:57:40
track2f_bis_clean_sequential_residual_offset_fw	Implemented Benchmark	te_track2f_bis_clean_residual_offset_fw	sequential_residual_offset_probe	0.003446	92,802	2026-06-04 23:48:53
track2h_latent_state_hysteresis_causal_tcn_offset_residual_fw	Implemented Benchmark	te_track2h_l_causal_tcn_offset_residual_fw	latent_state_hysteresis_probe	0.003470	97,923	2026-06-16 19:22:12
wave3_harmonic_prior_residual_smooth_l1_structured_fw	Implemented Benchmark	te_wave3_harmonic_prior_residual_smooth_l1_structured_fw	wave3_harmonic_prior_residual	0.003527	7,283	2026-06-15 15:16:24
track2h_latent_state_hysteresis_gru_offset_residual_fw	Implemented Benchmark	te_track2h_l_gru_offset_residual_fw	latent_state_hysteresis_probe	0.003537	125,475	2026-06-16 18:34:12
temporal_convolution_fw	Implemented Benchmark	te_temporal_convolution_sequence_remote_Fw	temporal_convolution	0.003611	147,009	2026-05-24 11:37:07

Backward Models

Family	Current Role	Best Run	Model Type	Test MAE [deg]	Params	Last Update
periodic_gru_sequence_bw	Implemented Benchmark	te_periodic_gru_sequence_remote_Bw	periodic_gru_sequence	0.002344	157,953	2026-05-25 18:09:44
periodic_lstm_sequence_bw	Implemented Benchmark	te_periodic_lstm_sequence_remote_Bw	periodic_lstm_sequence	0.002556	210,561	2026-05-25 20:05:38
track2h_mixture_density_heads_mdn_k2_bw	Implemented Benchmark	te_track2h_mdn_k2_bw	curve_aware_harmonic_residual_offset_probe	0.002658	86,802	2026-06-13 12:14:13
track2h_mixture_density_heads_mdn_k3_bw	Implemented Benchmark	te_track2h_mdn_k3_bw	curve_aware_harmonic_residual_offset_probe	0.002721	87,435	2026-06-13 13:10:07
track2h_quantile_probabilistic_quantile_p10_p50_p90_bw	Implemented Benchmark	te_track2h_quantile_p10_p50_p90_bw	curve_aware_harmonic_residual_offset_probe	0.002927	86,169	2026-06-12 12:10:23
tree_bw	Implemented Benchmark	te_hist_gbr_tabular_Bw_grid_depth6_lr008_leaf10	hist_gradient_boosting	0.002954	5	2026-05-11 21:18:29
track2h_quantile_probabilistic_gaussian_nll_bw	Implemented Benchmark	te_track2h_gaussian_nll_bw	curve_aware_harmonic_residual_offset_probe	0.002998	85,958	2026-06-12 13:53:26
residual_harmonic_mlp_bw	Implemented Benchmark	te_residual_harmonic_rcim_sparse_tracking_Bw	residual_harmonic_mlp	0.003042	26,260	2026-05-20 12:25:49
track2h_dispersion_aware_smooth_l1_robust_bw	Implemented Benchmark	te_track2h_smooth_l1_robust_bw	curve_aware_harmonic_residual_offset_probe	0.003074	85,747	2026-06-11 13:24:47
feedforward_bw	Implemented Benchmark	te_feedforward_stride1_high_compute_long_remote_Bw_optuna_t0005	feedforward	0.003099	167,937	2026-05-14 13:49:53
periodic_mlp_bw	Implemented Benchmark	te_periodic_mlp_h04_standard_Bw_optuna_t0006	periodic_mlp	0.003233	27,777	2026-05-21 09:38:37
track2f_bis_harmonic_residual_offset_bw	Implemented Benchmark	te_track2f_bis_harmonic_residual_offset_bw	harmonic_residual_offset_probe	0.003336	85,747	2026-06-05 16:44:49
wave3_harmonic_prior_residual_pointwise_control_bw	Implemented Benchmark	te_wave3_harmonic_prior_residual_pointwise_control_bw	wave3_harmonic_prior_residual	0.003363	7,283	2026-06-15 14:49:19
harmonic_regression_bw	Implemented Benchmark	te_harmonic_dense240_tracking_Bw	harmonic_regression	0.003400	2,886	2026-05-20 11:08:01
track2h_dispersion_aware_mae_robust_bw	Implemented Benchmark	te_track2h_mae_robust_bw	curve_aware_harmonic_residual_offset_probe	0.003430	85,747	2026-06-11 12:33:14
track2g_curve_aware_harmonic_residual_offset_pointwise_control_bw	Implemented Benchmark	te_track2g_curve_aware_pointwise_control_bw	curve_aware_harmonic_residual_offset_probe	0.003430	85,747	2026-06-08 19:23:08
wave3_harmonic_prior_residual_smooth_l1_structured_bw	Implemented Benchmark	te_wave3_harmonic_prior_residual_smooth_l1_structured_bw	wave3_harmonic_prior_residual	0.003431	7,283	2026-06-15 15:30:20
residual_harmonic_lstm_sequence_bw_sparse_rcim	Implemented Benchmark	te_residual_harmonic_lstm_sequence_remote_Bw_sparse_rcim	residual_harmonic_lstm_sequence	0.003440	201,364	2026-05-27 21:08:36
track2g_curve_aware_harmonic_residual_offset_raw_centered_shape_bw	Implemented Benchmark	te_track2g_curve_aware_raw_centered_shape_bw	curve_aware_harmonic_residual_offset_probe	0.003465	85,747	2026-06-08 20:11:41
residual_harmonic_gru_sequence_bw_dense360	Implemented Benchmark	te_residual_harmonic_gru_sequence_remote_Bw_dense360	residual_harmonic_gru_sequence	0.003468	151,762	2026-05-27 20:46:25
track2g_curve_aware_harmonic_residual_offset_raw_offset_bw	Implemented Benchmark	te_track2g_curve_aware_raw_offset_bw	curve_aware_harmonic_residual_offset_probe	0.003471	85,747	2026-06-08 21:06:56
track2h_dispersion_aware_log_cosh_robust_bw	Implemented Benchmark	te_track2h_log_cosh_robust_bw	curve_aware_harmonic_residual_offset_probe	0.003481	85,747	2026-06-11 14:01:57
residual_harmonic_gru_sequence_bw_dense240	Implemented Benchmark	te_residual_harmonic_gru_sequence_remote_Bw_dense240	residual_harmonic_gru_sequence	0.003492	151,522	2026-05-27 20:00:10
residual_harmonic_gru_sequence_bw_sparse_rcim	Implemented Benchmark	te_residual_harmonic_gru_sequence_remote_Bw_sparse_rcim	residual_harmonic_gru_sequence	0.003502	151,060	2026-05-27 19:18:56
track2g_curve_aware_harmonic_residual_offset_full_curve_composite_bw	Implemented Benchmark	te_track2g_curve_aware_full_curve_composite_bw	curve_aware_harmonic_residual_offset_probe	0.003511	85,747	2026-06-08 22:05:10
track2f_bis_clean_sequential_residual_offset_bw	Implemented Benchmark	te_track2f_bis_clean_residual_offset_bw	sequential_residual_offset_probe	0.003540	92,802	2026-06-04 23:58:31
track2h_latent_state_hysteresis_gru_offset_residual_bw	Implemented Benchmark	te_track2h_l_gru_offset_residual_bw	latent_state_hysteresis_probe	0.003545	125,475	2026-06-16 18:48:13
residual_harmonic_lstm_sequence_bw_dense360	Implemented Benchmark	te_residual_harmonic_lstm_sequence_remote_Bw_dense360	residual_harmonic_lstm_sequence	0.003556	202,066	2026-05-27 22:35:20
lstm_sequence_bw	Implemented Benchmark	te_lstm_sequence_remote_Bw	lstm_sequence	0.003557	201,345	2026-05-24 12:27:31
residual_harmonic_lstm_sequence_bw_dense240	Implemented Benchmark	te_residual_harmonic_lstm_sequence_remote_Bw_dense240	residual_harmonic_lstm_sequence	0.003605	201,826	2026-05-27 21:40:13
periodic_temporal_convolution_bw	Implemented Benchmark	te_periodic_temporal_convolution_sequence_remote_Bw	periodic_temporal_convolution	0.003614	158,529	2026-05-25 16:26:53
track2h_latent_state_hysteresis_causal_tcn_offset_residual_bw	Implemented Benchmark	te_track2h_l_causal_tcn_offset_residual_bw	latent_state_hysteresis_probe	0.003630	97,923	2026-06-16 19:34:05
gru_sequence_bw	Implemented Benchmark	te_gru_sequence_remote_Bw	gru_sequence	0.003631	151,041	2026-05-24 12:06:34
sequential_residual_offset_probe_bw	Implemented Benchmark	te_sequential_residual_offset_probe_remote_bw	sequential_residual_offset_probe	0.003638	92,802	2026-06-04 12:04:47
temporal_convolution_bw	Implemented Benchmark	te_temporal_convolution_sequence_remote_Bw	temporal_convolution	0.003739	147,009	2026-05-24 11:45:19

Active Training Or Improvement Branches

• No campaign is currently in prepared or running state.

• The next active implementation branch should therefore be read from the live backlog focus and the next approved campaign plan.

Roadmap And Planned Work

Wave Or Track	Status
Wave 0. Shared Infrastructure	completed.

Low-priority exploratory families currently listed in the backlog:

• low priority.

• Lightweight Transformer

• State-Space Sequence Model

• Neural ODE

• Hamiltonian-Inspired Model

• optional Kernel Ridge / Gaussian Process benchmark | Wave 1. Structured Static Baselines | planning report: completed;; implementation: completed;; smoke tests: completed;; validation checks: completed;; campaign execution: completed;; directional HPO closeout: completed;; exported global, forward, and backward surfaces: completed;; results report: completed;; status: closed. |

Low-priority exploratory families currently listed in the backlog:

• low priority.

• Lightweight Transformer

• State-Space Sequence Model

• Neural ODE

• Hamiltonian-Inspired Model

• optional Kernel Ridge / Gaussian Process benchmark | RCIM Model-Bank Reproduction. RCIM Paper-Faithful Model Bank | recovered original workflow: preserved;; original-dataset reimplementation: completed;; retuned reference archive: completed;; forward campaign: completed;; backward campaign: completed;; paper-reference archives: refreshed;; Tables 2-5: repopulated;; status: closed as faithful full-bank reproduction, not all-green |

Low-priority exploratory families currently listed in the backlog:

• low priority.

• Lightweight Transformer

• State-Space Sequence Model

• Neural ODE

• Hamiltonian-Inspired Model

• optional Kernel Ridge / Gaussian Process benchmark | TE Curve Verification Pipeline. Directional Offline Comparison | direction-aware loader and candidate matrix: completed;; recovered original forward candidates: included;; retuned forward and backward candidates: included;; RCIM Model-Bank Reproduction forward and backward candidates: included;; Wave 1 global, forward, and backward exports: included;; Wave 2.1 temporal global, forward, and backward registry candidates:; grouped source tables: completed;; composite best-reference visibility: completed;; direction/truth and preview audit: completed;; official model-verification report: completed;; multi-index curve-first selection policy: adopted;; complete multi-index reranking over all current official candidates:; status: closed. |

Low-priority exploratory families currently listed in the backlog:

• low priority.

• Lightweight Transformer

• State-Space Sequence Model

• Neural ODE

• Hamiltonian-Inspired Model

• optional Kernel Ridge / Gaussian Process benchmark | Wave 2.1. Temporal Models | status: entry campaign completed; closeout report prepared; official; initial families: temporal_convolution, gru_sequence, lstm_sequence;; configuration root: config/training/hydra/wave2/;; preliminary campaign plan:; closeout report:; campaign winner: te_gru_sequence_remote_Fw from family; refresh plan:; official verification report:; curve-verification decision: verified exploratory baselines, not promoted over tree;; mandatory rule: prepare or justify global, forward, and backward; baseline comparison: TE Curve Verification Pipeline plus closed Wave 1. |

Low-priority exploratory families currently listed in the backlog:

• low priority.

• Lightweight Transformer

• State-Space Sequence Model

• Neural ODE

• Hamiltonian-Inspired Model

• optional Kernel Ridge / Gaussian Process benchmark | Wave 2.2. Harmonic Temporal Hybrid Models | status: harmonic-temporal hybrid campaign completed; normal closeout report; families: periodic_temporal_convolution, periodic_gru_sequence,; configuration root:; preliminary campaign plan:; closeout report:; campaign winner: te_periodic_gru_sequence_remote_Bw from family; strongest bidirectional candidate: te_periodic_gru_sequence_remote_global; curve-verification decision: strongest repository-owned neural branch after official; mandatory rule: prepare or justify global, forward, and backward; baseline comparison: official curve-verification matrix plus visual collage and overlay |

Low-priority exploratory families currently listed in the backlog:

• low priority.

• Lightweight Transformer

• State-Space Sequence Model

• Neural ODE

• Hamiltonian-Inspired Model

• optional Kernel Ridge / Gaussian Process benchmark | Wave 2.3. Residual Harmonic Temporal Hybrid Models | status: residual harmonic temporal hybrid campaign completed; official; families: residual_harmonic_gru_sequence,; harmonic banks: sparse RCIM, dense 240, dense 360;; closeout report:; official verification report:; strongest Wave 2.3 forward candidate:; strongest Wave 2.3 backward candidate:; strongest Wave 2.3 global candidate:; curve-verification decision: verified exploratory baseline, not promoted over the; design conclusion: sparse RCIM harmonics remain useful, while dense 240 |

Low-priority exploratory families currently listed in the backlog:

• low priority.

• Lightweight Transformer

• State-Space Sequence Model

• Neural ODE

• Hamiltonian-Inspired Model

• optional Kernel Ridge / Gaussian Process benchmark | Wave 3.1. Offset-Aware Sequential Residual Probe | status: offset-aware probe campaign completed; official TE Curve Verification Pipeline matrix; family: sequential_residual_offset_probe;; official verification report:; strongest Wave 3.1 forward candidate:; strongest Wave 3.1 backward candidate:; strongest Wave 3.1 global candidate:; curve-verification decision: verified exploratory baseline, not promoted over the; design conclusion: a sequential residual offset head alone does not solve |

Low-priority exploratory families currently listed in the backlog:

• low priority.

• Lightweight Transformer

• State-Space Sequence Model

• Neural ODE

• Hamiltonian-Inspired Model

• optional Kernel Ridge / Gaussian Process benchmark | Wave 3.2. Harmonic-Offset Probe | status: campaign completed after runner registration repair; official; families:; track2f_bis_clean_sequential_residual_offset_global;; track2f_bis_clean_sequential_residual_offset_fw;; track2f_bis_clean_sequential_residual_offset_bw;; track2f_bis_harmonic_residual_offset_global;; track2f_bis_harmonic_residual_offset_fw;; track2f_bis_harmonic_residual_offset_bw;; closeout report:; official verification report:; clean global candidate:; harmonic global candidate:; clean forward candidate:; harmonic forward candidate:; clean backward candidate:; harmonic backward candidate:; strongest Wave 3.2 forward candidate:; strongest Wave 3.2 backward candidate:; strongest Wave 3.2 global candidate:; curve-verification decision: verified exploratory baseline, not promoted over the; design conclusion: harmonic forcing helps the direction-specific Fw and |

Low-priority exploratory families currently listed in the backlog:

• low priority.

• Lightweight Transformer

• State-Space Sequence Model

• Neural ODE

• Hamiltonian-Inspired Model

• optional Kernel Ridge / Gaussian Process benchmark | Wave 4.1. Dispersion-Aware Robust-Loss Probe | status: robust-loss campaign completed; official TE Curve Verification Pipeline matrix refresh; families:; track2h_dispersion_aware_mae_robust_global;; track2h_dispersion_aware_mae_robust_fw;; track2h_dispersion_aware_mae_robust_bw;; track2h_dispersion_aware_smooth_l1_robust_global;; track2h_dispersion_aware_smooth_l1_robust_fw;; track2h_dispersion_aware_smooth_l1_robust_bw;; track2h_dispersion_aware_log_cosh_robust_global;; track2h_dispersion_aware_log_cosh_robust_fw;; track2h_dispersion_aware_log_cosh_robust_bw;; closeout report:; official verification report:; robust global candidate:; robust forward candidate:; robust backward candidate:; campaign scalar winner:; TE Curve Verification Pipeline strongest forward candidate:; TE Curve Verification Pipeline strongest backward candidate:; TE Curve Verification Pipeline strongest global candidate:; curve-verification decision: verified exploratory baseline, not promoted over the; design conclusion: robust losses are useful enough to keep in the |

Low-priority exploratory families currently listed in the backlog:

• low priority.

• Lightweight Transformer

• State-Space Sequence Model

• Neural ODE

• Hamiltonian-Inspired Model

• optional Kernel Ridge / Gaussian Process benchmark | Wave 4.2. Quantile Probabilistic Probe | status: quantile/probabilistic campaign completed; official TE Curve Verification Pipeline matrix; families:; track2h_quantile_probabilistic_quantile_p10_p50_p90_global;; track2h_quantile_probabilistic_quantile_p10_p50_p90_fw;; track2h_quantile_probabilistic_quantile_p10_p50_p90_bw;; track2h_quantile_probabilistic_gaussian_nll_global;; track2h_quantile_probabilistic_gaussian_nll_fw;; track2h_quantile_probabilistic_gaussian_nll_bw;; closeout report:; official verification report:; strongest probabilistic global candidate:; strongest probabilistic forward-only candidate:; strongest probabilistic forward-evaluated candidate:; strongest probabilistic backward candidate:; curve-verification decision: verified exploratory baseline, not promoted over the; design conclusion: probabilistic losses improve over robust losses on the |

Low-priority exploratory families currently listed in the backlog:

• low priority.

• Lightweight Transformer

• State-Space Sequence Model

• Neural ODE

• Hamiltonian-Inspired Model

• optional Kernel Ridge / Gaussian Process benchmark | Wave 4.3. Mixture Density Heads Probe | status: mixture-density heads campaign completed; official TE Curve Verification Pipeline matrix; families:; track2h_mixture_density_heads_mdn_k2_global;; track2h_mixture_density_heads_mdn_k2_fw;; track2h_mixture_density_heads_mdn_k2_bw;; track2h_mixture_density_heads_mdn_k3_global;; track2h_mixture_density_heads_mdn_k3_fw;; track2h_mixture_density_heads_mdn_k3_bw;; closeout report:; official verification report:; matrix output:; strongest MDN global candidate:; strongest MDN forward candidate:; strongest MDN backward candidate:; strongest TE Curve Verification Pipeline forward MDN candidate:; strongest TE Curve Verification Pipeline backward MDN candidate:; strongest TE Curve Verification Pipeline global MDN candidate:; campaign scalar winner:; program scalar winner changed: no, te_periodic_gru_sequence_remote_Bw; curve-verification decision: verified exploratory baseline, not promoted over the; design conclusion: MDN improves the scalar Bw dispersion-aware branch by |

Low-priority exploratory families currently listed in the backlog:

• low priority.

• Lightweight Transformer

• State-Space Sequence Model

• Neural ODE

• Hamiltonian-Inspired Model

• optional Kernel Ridge / Gaussian Process benchmark | Wave 4.4. Latent-State Hysteresis Probe | status: latent-state / hysteresis-aware campaign completed; official; families:; track2h_latent_state_hysteresis_gru_offset_residual_global;; track2h_latent_state_hysteresis_gru_offset_residual_fw;; track2h_latent_state_hysteresis_gru_offset_residual_bw;; track2h_latent_state_hysteresis_causal_tcn_offset_residual_global;; track2h_latent_state_hysteresis_causal_tcn_offset_residual_fw;; track2h_latent_state_hysteresis_causal_tcn_offset_residual_bw;; closeout report:; official TE curve-verification report:; official curve-verification matrix:; strongest Wave 4.4 global candidate:; strongest Wave 4.4 forward candidate:; strongest Wave 4.4 backward candidate:; campaign scalar winner:; program scalar winner changed: no, te_periodic_gru_sequence_remote_Bw; scalar comparison: Wave 4.4 improves the global scalar surface versus MDN; official TE Curve Verification Pipeline strongest refreshed candidates:; curve-verification decision: verified exploratory baseline, not promoted over; design conclusion: causal history is useful as a diagnostic signal, but this |

Low-priority exploratory families currently listed in the backlog:

• low priority.

• Lightweight Transformer

• State-Space Sequence Model

• Neural ODE

• Hamiltonian-Inspired Model

• optional Kernel Ridge / Gaussian Process benchmark | Wave 5.1. Hybrid Structured Models | status: first real campaign closed successfully as a scalar training; current scaffold:; model type: wave3_harmonic_prior_residual;; model class:; dry-run skeleton checker:; training-smoke-ready checker:; final one-batch validation artifact:; prepared package:; prepared launcher:; campaign closeout report:; scalar training winner:; scalar decision: no program-best promotion; the current program winner; official TE Curve Verification Pipeline launcher:; official TE curve-verification report:; strongest Wave 5.1 TE Curve Verification Pipeline candidate:; updated priority: use the completed Wave 5.1 curve, offset, collage, overlay,; mandatory rule: prepare or justify global, forward, and backward; paper-reproduction scope:; compare hybrid structured predictors against the paper-style harmonic stack;; test condition-conditioned residual structure and separate treatment of; prepare the repository-owned deployable predictor package after the; next implementation steps:; proceed to Wave 5.2 / integrated multi-head planning with hidden-state |

Low-priority exploratory families currently listed in the backlog:

• low priority.

• Lightweight Transformer

• State-Space Sequence Model

• Neural ODE

• Hamiltonian-Inspired Model

• optional Kernel Ridge / Gaussian Process benchmark | Wave 5.2. PINN Formulation And First PINN | status: pre-implemented at Wave 5.2A diagnostic level, not campaign-ready;; current scaffold:; diagnostic adapter:; diagnostic report builder:; parameter-inventory report builder:; generated diagnostic report:; generated parameter-inventory report:; companion artifacts:; parameter-inventory artifacts:; updated priority: execute dataset-aligned diagnostic calibration after the; mandatory rule: prepare or justify global, forward, and backward; paper-reproduction scope:; prepare PINN-side model and loss formulations for later offline and; test whether soft physics, periodicity, smoothness, harmonic-consistency,; keep online compensation execution out of Wave 5.2 unless Track 3 is; completed inventory conclusions:; known geometry constants are safe for diagnostics and feature generation;; operating metadata can be used for stratification and causal conditioning;; five equivalent-error groups are train-only calibratable;; contact geometry remains unavailable or ambiguous and blocks calibrated; measured TE remains target-only and must not become an inference input.; next implementation steps:; compare MMT diagnostic signatures against dataset-aligned curve summaries; design a train-only equivalent-error calibration policy for candidate; decide whether the MMT path remains diagnostic-only, becomes a feature; do not treat the current demonstration harmonic summary as dataset |

Low-priority exploratory families currently listed in the backlog:

• low priority.

• Lightweight Transformer

• State-Space Sequence Model

• Neural ODE

• Hamiltonian-Inspired Model

• optional Kernel Ridge / Gaussian Process benchmark | Wave 5. Cross-Wave Comparison And Best Solution | status: pending;; mandatory rule: preserve direction-separated reporting;; paper-reproduction scope:; compare closed offline waves and Track 3 results when available;; finalize the real paper vs repository comparison only after Track 3 |

Low-priority exploratory families currently listed in the backlog:

• low priority.

• Lightweight Transformer

• State-Space Sequence Model

• Neural ODE

• Hamiltonian-Inspired Model

• optional Kernel Ridge / Gaussian Process benchmark | Track 3. Online Compensation And Deployment Evaluation | status: future implementation branch;; canonical objective: close Target B;; scope:; online compensation loop in the TestRig / TwinCAT path;; old future Pipelines 8-10;; Robot and Cycloidal motion-profile validation;; uncompensated versus compensated TE RMS and TE max;; final paper-style Table 9 report;; deployment-readiness interpretation for the selected repository model path. |

Low-priority exploratory families currently listed in the backlog:

• low priority.

• Lightweight Transformer

• State-Space Sequence Model

• Neural ODE

• Hamiltonian-Inspired Model

• optional Kernel Ridge / Gaussian Process benchmark

Recent Campaign Changes

Campaign	Generated At	Completed	Failed	Winner	Impact
polished_dataset_stage1_smoke_2026_06_21	2026-06-22 16:00:39	8	0	te_periodic_gru_sequence_remote_global	Updated global best
polished_dataset_stage1_smoke_2026_06_21	2026-06-22 12:04:16	0	1	N/A	No winner artifact
track2h_latent_state_hysteresis_campaign_2026_06_16	2026-06-16 19:34:05	6	0	te_track2h_l_causal_tcn_offset_residual_global	Updated track2h_latent_state_hysteresis_causal_tcn_offset_residual_global family best
wave3_harmonic_prior_residual_campaign_2026_06_14	2026-06-15 15:30:20	6	0	te_wave3_harmonic_prior_residual_pointwise_control_bw	Updated wave3_harmonic_prior_residual_pointwise_control_bw family best
track2h_mixture_density_heads_campaign_2026_06_13	2026-06-13 13:10:07	6	0	te_track2h_mdn_k2_bw	Updated track2h_mixture_density_heads_mdn_k2_bw family best

Ranking Policy

• Primary metric: test_mae

• First tie-breaker: test_rmse

• Second tie-breaker: val_mae

• Third tie-breaker: trainable_parameter_count

• Direction: minimize

Best Result Per Family

• Scope-separated family ranking is mandatory for every future wave that introduces more than one canonical training surface.

Global Models

Family	Best Run	Model Type	Val MAE [deg]	Test MAE [deg]	Test RMSE [deg]	Params	Artifact Size	Training Cost	Current Role
periodic_gru_sequence	te_periodic_gru_sequence_remote_global	periodic_gru_sequence	0.001274	0.001279	0.001638	157,569	1.82 MB	High	Current Program Winner
tree	te_hist_gbr_tabular_global	hist_gradient_boosting	0.001591	0.001753	0.002892	4	0.44 MB	Very Low	Implemented Benchmark
track2g_curve_aware_harmonic_residual_offset_full_curve_composite_global	te_track2g_curve_aware_full_curve_composite_global	curve_aware_harmonic_residual_offset_probe	0.001872	0.002008	0.002581	85,440	1.00 MB	High	Implemented Benchmark
residual_harmonic_gru_sequence_sparse_rcim	te_residual_harmonic_gru_sequence_remote_global_sparse_rcim	residual_harmonic_gru_sequence	0.001978	0.002112	0.002699	150,676	1.74 MB	Medium	Implemented Benchmark
wave3_harmonic_prior_residual_smooth_l1_structured_global	te_wave3_harmonic_prior_residual_smooth_l1_structured_global	wave3_harmonic_prior_residual	0.001889	0.002168	0.002763	7,168	0.10 MB	Medium	Implemented Benchmark
track2h_latent_state_hysteresis_gru_offset_residual_global	te_track2h_l_gru_offset_residual_global	latent_state_hysteresis_probe	0.002232	0.002339	0.002986	124,899	1.48 MB	Medium	Implemented Benchmark
periodic_lstm_sequence	te_periodic_lstm_sequence_remote_global	periodic_lstm_sequence	0.002526	0.002682	0.002969	210,561	2.43 MB	High	Implemented Benchmark
feedforward	te_feedforward_trial	feedforward	0.002725	0.002877	0.003835	26,113	0.32 MB	Low	Current Plain MLP Anchor
track2h_quantile_probabilistic_gaussian_nll_global	te_track2h_gaussian_nll_global	curve_aware_harmonic_residual_offset_probe	0.003267	0.003013	0.003388	85,958	1.00 MB	High	Implemented Benchmark
residual_harmonic_mlp	te_residual_h12_deep_joint_wave1_global_optuna_t0006	residual_harmonic_mlp	0.002895	0.003034	0.003550	26,266	0.32 MB	Unknown	Implemented Benchmark
periodic_mlp	te_periodic_mlp_h04_standard_global_optuna_t0010	periodic_mlp	0.002994	0.003186	0.003690	27,265	0.33 MB	Unknown	Implemented Benchmark
track2g_curve_aware_harmonic_residual_offset_raw_centered_shape_global	te_track2g_curve_aware_raw_centered_shape_global	curve_aware_harmonic_residual_offset_probe	0.003636	0.003350	0.003753	85,747	1.00 MB	Medium	Implemented Benchmark
track2h_latent_state_hysteresis_causal_tcn_offset_residual_global	te_track2h_l_causal_tcn_offset_residual_global	latent_state_hysteresis_probe	0.003543	0.003368	0.003860	97,923	1.17 MB	Medium	Implemented Benchmark
residual_harmonic_lstm_sequence_sparse_rcim	te_residual_harmonic_lstm_sequence_remote_global_sparse_rcim	residual_harmonic_lstm_sequence	0.003632	0.003368	0.003808	201,364	2.32 MB	Low	Implemented Benchmark
track2h_quantile_probabilistic_quantile_p10_p50_p90_global	te_track2h_quantile_p10_p50_p90_global	curve_aware_harmonic_residual_offset_probe	0.003606	0.003383	0.003764	86,169	1.01 MB	Medium	Implemented Benchmark
track2h_dispersion_aware_mae_robust_global	te_track2h_mae_robust_global	curve_aware_harmonic_residual_offset_probe	0.003645	0.003406	0.003807	85,747	1.00 MB	Medium	Implemented Benchmark
track2h_dispersion_aware_smooth_l1_robust_global	te_track2h_smooth_l1_robust_global	curve_aware_harmonic_residual_offset_probe	0.003641	0.003422	0.003810	85,747	1.00 MB	Medium	Implemented Benchmark
wave3_harmonic_prior_residual_pointwise_control_global	te_wave3_harmonic_prior_residual_pointwise_control_global	wave3_harmonic_prior_residual	0.003611	0.003451	0.003851	7,283	0.11 MB	Medium	Implemented Benchmark
track2g_curve_aware_harmonic_residual_offset_raw_offset_global	te_track2g_curve_aware_raw_offset_global	curve_aware_harmonic_residual_offset_probe	0.003564	0.003465	0.003829	85,747	1.00 MB	Medium	Implemented Benchmark
residual_harmonic_lstm_sequence_dense240	te_residual_harmonic_lstm_sequence_remote_global_dense240	residual_harmonic_lstm_sequence	0.003624	0.003473	0.003925	201,826	2.33 MB	Low	Implemented Benchmark
residual_harmonic_lstm_sequence_dense360	te_residual_harmonic_lstm_sequence_remote_global_dense360	residual_harmonic_lstm_sequence	0.003648	0.003477	0.003940	202,066	2.33 MB	Medium	Implemented Benchmark
lstm_sequence	te_lstm_sequence_remote_global	lstm_sequence	0.003681	0.003482	0.003948	201,345	2.32 MB	Low	Implemented Benchmark
track2h_mixture_density_heads_mdn_k2_global	te_track2h_mdn_k2_global	curve_aware_harmonic_residual_offset_probe	0.003654	0.003503	0.003938	86,802	1.01 MB	Medium	Implemented Benchmark
track2h_dispersion_aware_log_cosh_robust_global	te_track2h_log_cosh_robust_global	curve_aware_harmonic_residual_offset_probe	0.003645	0.003505	0.003935	85,747	1.00 MB	Medium	Implemented Benchmark
periodic_temporal_convolution	te_periodic_temporal_convolution_sequence_remote_global	periodic_temporal_convolution	0.003634	0.003508	0.003929	158,529	1.83 MB	Medium	Implemented Benchmark
residual_harmonic_gru_sequence_dense240	te_residual_harmonic_gru_sequence_remote_global_dense240	residual_harmonic_gru_sequence	0.003600	0.003511	0.003983	151,522	1.75 MB	Low	Implemented Benchmark
track2f_bis_clean_sequential_residual_offset_global	te_track2f_bis_clean_residual_offset_global	sequential_residual_offset_probe	0.003717	0.003528	0.004010	92,802	1.09 MB	Low	Implemented Benchmark
residual_harmonic_gru_sequence_dense360	te_residual_harmonic_gru_sequence_remote_global_dense360	residual_harmonic_gru_sequence	0.003628	0.003535	0.003999	151,762	1.75 MB	Medium	Implemented Benchmark
sequential_residual_offset_probe	te_sequential_residual_offset_probe_remote_global	sequential_residual_offset_probe	0.003783	0.003537	0.004005	92,802	1.09 MB	Low	Implemented Benchmark
track2f_bis_harmonic_residual_offset_global	te_track2f_bis_harmonic_residual_offset_global	harmonic_residual_offset_probe	0.003659	0.003538	0.003932	85,747	1.00 MB	Very Low	Implemented Benchmark
track2h_mixture_density_heads_mdn_k3_global	te_track2h_mdn_k3_global	curve_aware_harmonic_residual_offset_probe	0.003617	0.003564	0.003986	87,435	1.02 MB	Medium	Implemented Benchmark
track2g_curve_aware_harmonic_residual_offset_pointwise_control_global	te_track2g_curve_aware_pointwise_control_global	curve_aware_harmonic_residual_offset_probe	0.003607	0.003587	0.004001	85,747	1.00 MB	Medium	Implemented Benchmark
gru_sequence	te_gru_sequence_remote_global	gru_sequence	0.003707	0.003591	0.004110	151,041	1.74 MB	Low	Implemented Benchmark
temporal_convolution	te_temporal_convolution_sequence_remote_global	temporal_convolution	0.003935	0.003754	0.004266	147,009	1.70 MB	Low	Implemented Benchmark
harmonic_regression	te_harmonic_order12_linear_conditioned_recovery_global	harmonic_regression	0.003904	0.003839	0.004555	125	0.01 MB	Low	Implemented Benchmark
feedforward_recovery_micro	te_feedforward_optuna_recovery_micro_global_optuna_t0000	feedforward	0.004266	0.004164	0.005109	109,953	1.28 MB	Unknown	Implemented Benchmark
feedforward_recovery_probe_dense	te_feedforward_optuna_recovery_probe_dense_global_optuna_t0000	feedforward	0.004257	0.004602	0.005262	109,953	1.28 MB	Unknown	Implemented Benchmark

Forward Models

Family	Best Run	Model Type	Val MAE [deg]	Test MAE [deg]	Test RMSE [deg]	Params	Artifact Size	Training Cost	Current Role
tree_fw	te_hist_gbr_tabular_Fw_grid_depth6_lr008_leaf10	hist_gradient_boosting	0.002677	0.002743	0.003409	5	0.45 MB	Unknown	Implemented Benchmark
track2f_bis_harmonic_residual_offset_fw	te_track2f_bis_harmonic_residual_offset_fw	harmonic_residual_offset_probe	0.002941	0.002862	0.003334	85,747	1.00 MB	Very Low	Implemented Benchmark
harmonic_regression_fw	te_harmonic_dense360_tracking_Fw	harmonic_regression	0.002610	0.002916	0.003237	4,326	0.06 MB	Low	Implemented Benchmark
periodic_mlp_fw	te_periodic_mlp_dense240_tracking_Fw	periodic_mlp	0.002541	0.003055	0.003537	87,681	1.03 MB	Low	Implemented Benchmark
residual_harmonic_mlp_fw	te_residual_harmonic_rcim_sparse_tracking_Fw	residual_harmonic_mlp	0.002704	0.003089	0.003498	26,260	0.32 MB	Low	Implemented Benchmark
track2h_dispersion_aware_mae_robust_fw	te_track2h_mae_robust_fw	curve_aware_harmonic_residual_offset_probe	0.003258	0.003146	0.003527	85,747	1.00 MB	Low	Implemented Benchmark
track2h_quantile_probabilistic_gaussian_nll_fw	te_track2h_gaussian_nll_fw	curve_aware_harmonic_residual_offset_probe	0.003293	0.003165	0.003548	85,958	1.00 MB	Low	Implemented Benchmark
track2g_curve_aware_harmonic_residual_offset_raw_centered_shape_fw	te_track2g_curve_aware_raw_centered_shape_fw	curve_aware_harmonic_residual_offset_probe	0.003251	0.003181	0.003571	85,747	1.00 MB	Low	Implemented Benchmark
periodic_gru_sequence_fw	te_periodic_gru_sequence_remote_Fw	periodic_gru_sequence	0.003227	0.003193	0.003583	157,953	1.82 MB	Low	Implemented Benchmark
residual_harmonic_gru_sequence_fw_sparse_rcim	te_residual_harmonic_gru_sequence_remote_Fw_sparse_rcim	residual_harmonic_gru_sequence	0.003309	0.003200	0.003635	151,060	1.75 MB	Low	Implemented Benchmark
feedforward_fw	te_feedforward_stride1_high_compute_long_remote_Fw_optuna_t0008	feedforward	0.002850	0.003203	0.003787	109,953	1.28 MB	Unknown	Implemented Benchmark
residual_harmonic_gru_sequence_fw_dense240	te_residual_harmonic_gru_sequence_remote_Fw_dense240	residual_harmonic_gru_sequence	0.003270	0.003219	0.003653	151,522	1.75 MB	Low	Implemented Benchmark
residual_harmonic_lstm_sequence_fw_sparse_rcim	te_residual_harmonic_lstm_sequence_remote_Fw_sparse_rcim	residual_harmonic_lstm_sequence	0.003344	0.003234	0.003679	201,364	2.32 MB	Low	Implemented Benchmark
track2h_mixture_density_heads_mdn_k3_fw	te_track2h_mdn_k3_fw	curve_aware_harmonic_residual_offset_probe	0.003253	0.003235	0.003613	87,435	1.02 MB	Low	Implemented Benchmark
residual_harmonic_gru_sequence_fw_dense360	te_residual_harmonic_gru_sequence_remote_Fw_dense360	residual_harmonic_gru_sequence	0.003265	0.003241	0.003677	151,762	1.75 MB	Low	Implemented Benchmark
track2g_curve_aware_harmonic_residual_offset_full_curve_composite_fw	te_track2g_curve_aware_full_curve_composite_fw	curve_aware_harmonic_residual_offset_probe	0.003320	0.003260	0.003630	85,747	1.00 MB	Low	Implemented Benchmark
residual_harmonic_lstm_sequence_fw_dense240	te_residual_harmonic_lstm_sequence_remote_Fw_dense240	residual_harmonic_lstm_sequence	0.003307	0.003262	0.003706	201,826	2.33 MB	Low	Implemented Benchmark
periodic_lstm_sequence_fw	te_periodic_lstm_sequence_remote_Fw	periodic_lstm_sequence	0.003254	0.003274	0.003651	210,561	2.43 MB	Low	Implemented Benchmark
track2g_curve_aware_harmonic_residual_offset_raw_offset_fw	te_track2g_curve_aware_raw_offset_fw	curve_aware_harmonic_residual_offset_probe	0.003328	0.003279	0.003698	85,747	1.00 MB	Low	Implemented Benchmark
track2h_quantile_probabilistic_quantile_p10_p50_p90_fw	te_track2h_quantile_p10_p50_p90_fw	curve_aware_harmonic_residual_offset_probe	0.003269	0.003285	0.003668	86,169	1.01 MB	Low	Implemented Benchmark
track2h_dispersion_aware_smooth_l1_robust_fw	te_track2h_smooth_l1_robust_fw	curve_aware_harmonic_residual_offset_probe	0.003235	0.003314	0.003679	85,747	1.00 MB	Low	Implemented Benchmark
gru_sequence_fw	te_gru_sequence_remote_Fw	gru_sequence	0.003409	0.003333	0.003881	151,041	1.74 MB	Low	Implemented Benchmark
periodic_temporal_convolution_fw	te_periodic_temporal_convolution_sequence_remote_Fw	periodic_temporal_convolution	0.003321	0.003337	0.003830	158,529	1.83 MB	Low	Implemented Benchmark
track2h_mixture_density_heads_mdn_k2_fw	te_track2h_mdn_k2_fw	curve_aware_harmonic_residual_offset_probe	0.003285	0.003339	0.003721	86,802	1.01 MB	Low	Implemented Benchmark
residual_harmonic_lstm_sequence_fw_dense360	te_residual_harmonic_lstm_sequence_remote_Fw_dense360	residual_harmonic_lstm_sequence	0.003302	0.003351	0.003774	202,066	2.33 MB	Low	Implemented Benchmark
track2h_dispersion_aware_log_cosh_robust_fw	te_track2h_log_cosh_robust_fw	curve_aware_harmonic_residual_offset_probe	0.003280	0.003355	0.003708	85,747	1.00 MB	Low	Implemented Benchmark
lstm_sequence_fw	te_lstm_sequence_remote_Fw	lstm_sequence	0.003448	0.003370	0.003921	201,345	2.32 MB	Low	Implemented Benchmark
track2g_curve_aware_harmonic_residual_offset_pointwise_control_fw	te_track2g_curve_aware_pointwise_control_fw	curve_aware_harmonic_residual_offset_probe	0.003291	0.003371	0.003763	85,747	1.00 MB	Low	Implemented Benchmark
wave3_harmonic_prior_residual_pointwise_control_fw	te_wave3_harmonic_prior_residual_pointwise_control_fw	wave3_harmonic_prior_residual	0.003315	0.003382	0.003779	7,283	0.11 MB	Low	Implemented Benchmark
sequential_residual_offset_probe_fw	te_sequential_residual_offset_probe_remote_fw	sequential_residual_offset_probe	0.003380	0.003385	0.003931	92,802	1.09 MB	Low	Implemented Benchmark
track2f_bis_clean_sequential_residual_offset_fw	te_track2f_bis_clean_residual_offset_fw	sequential_residual_offset_probe	0.003474	0.003446	0.003972	92,802	1.09 MB	Low	Implemented Benchmark
track2h_latent_state_hysteresis_causal_tcn_offset_residual_fw	te_track2h_l_causal_tcn_offset_residual_fw	latent_state_hysteresis_probe	0.003565	0.003470	0.004068	97,923	1.17 MB	Low	Implemented Benchmark
wave3_harmonic_prior_residual_smooth_l1_structured_fw	te_wave3_harmonic_prior_residual_smooth_l1_structured_fw	wave3_harmonic_prior_residual	0.003310	0.003527	0.003900	7,283	0.11 MB	Low	Implemented Benchmark
track2h_latent_state_hysteresis_gru_offset_residual_fw	te_track2h_l_gru_offset_residual_fw	latent_state_hysteresis_probe	0.003468	0.003537	0.004110	125,475	1.48 MB	Low	Implemented Benchmark
temporal_convolution_fw	te_temporal_convolution_sequence_remote_Fw	temporal_convolution	0.003490	0.003611	0.004183	147,009	1.70 MB	Low	Implemented Benchmark

Backward Models

Family	Best Run	Model Type	Val MAE [deg]	Test MAE [deg]	Test RMSE [deg]	Params	Artifact Size	Training Cost	Current Role
periodic_gru_sequence_bw	te_periodic_gru_sequence_remote_Bw	periodic_gru_sequence	0.002523	0.002344	0.002747	157,953	1.82 MB	Medium	Implemented Benchmark
periodic_lstm_sequence_bw	te_periodic_lstm_sequence_remote_Bw	periodic_lstm_sequence	0.002432	0.002556	0.002953	210,561	2.43 MB	Medium	Implemented Benchmark
track2h_mixture_density_heads_mdn_k2_bw	te_track2h_mdn_k2_bw	curve_aware_harmonic_residual_offset_probe	0.002914	0.002658	0.003198	86,802	1.01 MB	Medium	Implemented Benchmark
track2h_mixture_density_heads_mdn_k3_bw	te_track2h_mdn_k3_bw	curve_aware_harmonic_residual_offset_probe	0.002775	0.002721	0.003250	87,435	1.02 MB	Medium	Implemented Benchmark
track2h_quantile_probabilistic_quantile_p10_p50_p90_bw	te_track2h_quantile_p10_p50_p90_bw	curve_aware_harmonic_residual_offset_probe	0.003436	0.002927	0.003519	86,169	1.01 MB	Medium	Implemented Benchmark
tree_bw	te_hist_gbr_tabular_Bw_grid_depth6_lr008_leaf10	hist_gradient_boosting	0.002681	0.002954	0.003749	5	0.45 MB	Unknown	Implemented Benchmark
track2h_quantile_probabilistic_gaussian_nll_bw	te_track2h_gaussian_nll_bw	curve_aware_harmonic_residual_offset_probe	0.003298	0.002998	0.003608	85,958	1.00 MB	Medium	Implemented Benchmark
residual_harmonic_mlp_bw	te_residual_harmonic_rcim_sparse_tracking_Bw	residual_harmonic_mlp	0.002953	0.003042	0.003548	26,260	0.32 MB	Low	Implemented Benchmark
track2h_dispersion_aware_smooth_l1_robust_bw	te_track2h_smooth_l1_robust_bw	curve_aware_harmonic_residual_offset_probe	0.003372	0.003074	0.003662	85,747	1.00 MB	Medium	Implemented Benchmark
feedforward_bw	te_feedforward_stride1_high_compute_long_remote_Bw_optuna_t0005	feedforward	0.003018	0.003099	0.003630	167,937	1.95 MB	Unknown	Implemented Benchmark
periodic_mlp_bw	te_periodic_mlp_h04_standard_Bw_optuna_t0006	periodic_mlp	0.002907	0.003233	0.003792	27,777	0.34 MB	Unknown	Implemented Benchmark
track2f_bis_harmonic_residual_offset_bw	te_track2f_bis_harmonic_residual_offset_bw	harmonic_residual_offset_probe	0.003555	0.003336	0.003935	85,747	1.00 MB	Very Low	Implemented Benchmark
wave3_harmonic_prior_residual_pointwise_control_bw	te_wave3_harmonic_prior_residual_pointwise_control_bw	wave3_harmonic_prior_residual	0.003634	0.003363	0.003902	7,283	0.11 MB	Low	Implemented Benchmark
harmonic_regression_bw	te_harmonic_dense240_tracking_Bw	harmonic_regression	0.003588	0.003400	0.003886	2,886	0.04 MB	Low	Implemented Benchmark
track2h_dispersion_aware_mae_robust_bw	te_track2h_mae_robust_bw	curve_aware_harmonic_residual_offset_probe	0.003579	0.003430	0.004029	85,747	1.00 MB	Medium	Implemented Benchmark
track2g_curve_aware_harmonic_residual_offset_pointwise_control_bw	te_track2g_curve_aware_pointwise_control_bw	curve_aware_harmonic_residual_offset_probe	0.003749	0.003430	0.003945	85,747	1.00 MB	Low	Implemented Benchmark
wave3_harmonic_prior_residual_smooth_l1_structured_bw	te_wave3_harmonic_prior_residual_smooth_l1_structured_bw	wave3_harmonic_prior_residual	0.003644	0.003431	0.003953	7,283	0.11 MB	Low	Implemented Benchmark
residual_harmonic_lstm_sequence_bw_sparse_rcim	te_residual_harmonic_lstm_sequence_remote_Bw_sparse_rcim	residual_harmonic_lstm_sequence	0.003764	0.003440	0.004030	201,364	2.32 MB	Low	Implemented Benchmark
track2g_curve_aware_harmonic_residual_offset_raw_centered_shape_bw	te_track2g_curve_aware_raw_centered_shape_bw	curve_aware_harmonic_residual_offset_probe	0.003740	0.003465	0.003998	85,747	1.00 MB	Medium	Implemented Benchmark
residual_harmonic_gru_sequence_bw_dense360	te_residual_harmonic_gru_sequence_remote_Bw_dense360	residual_harmonic_gru_sequence	0.003773	0.003468	0.004050	151,762	1.75 MB	Low	Implemented Benchmark
track2g_curve_aware_harmonic_residual_offset_raw_offset_bw	te_track2g_curve_aware_raw_offset_bw	curve_aware_harmonic_residual_offset_probe	0.003751	0.003471	0.003992	85,747	1.00 MB	Medium	Implemented Benchmark
track2h_dispersion_aware_log_cosh_robust_bw	te_track2h_log_cosh_robust_bw	curve_aware_harmonic_residual_offset_probe	0.003774	0.003481	0.004029	85,747	1.00 MB	Low	Implemented Benchmark
residual_harmonic_gru_sequence_bw_dense240	te_residual_harmonic_gru_sequence_remote_Bw_dense240	residual_harmonic_gru_sequence	0.003585	0.003492	0.004074	151,522	1.75 MB	Medium	Implemented Benchmark
residual_harmonic_gru_sequence_bw_sparse_rcim	te_residual_harmonic_gru_sequence_remote_Bw_sparse_rcim	residual_harmonic_gru_sequence	0.003833	0.003502	0.004061	151,060	1.75 MB	Low	Implemented Benchmark
track2g_curve_aware_harmonic_residual_offset_full_curve_composite_bw	te_track2g_curve_aware_full_curve_composite_bw	curve_aware_harmonic_residual_offset_probe	0.003803	0.003511	0.004113	85,747	1.00 MB	Medium	Implemented Benchmark
track2f_bis_clean_sequential_residual_offset_bw	te_track2f_bis_clean_residual_offset_bw	sequential_residual_offset_probe	0.003820	0.003540	0.004203	92,802	1.09 MB	Low	Implemented Benchmark
track2h_latent_state_hysteresis_gru_offset_residual_bw	te_track2h_l_gru_offset_residual_bw	latent_state_hysteresis_probe	0.003837	0.003545	0.004175	125,475	1.48 MB	Low	Implemented Benchmark
residual_harmonic_lstm_sequence_bw_dense360	te_residual_harmonic_lstm_sequence_remote_Bw_dense360	residual_harmonic_lstm_sequence	0.003729	0.003556	0.004125	202,066	2.33 MB	Medium	Implemented Benchmark
lstm_sequence_bw	te_lstm_sequence_remote_Bw	lstm_sequence	0.003815	0.003557	0.004201	201,345	2.32 MB	Low	Implemented Benchmark
residual_harmonic_lstm_sequence_bw_dense240	te_residual_harmonic_lstm_sequence_remote_Bw_dense240	residual_harmonic_lstm_sequence	0.003742	0.003605	0.004129	201,826	2.33 MB	Low	Implemented Benchmark
periodic_temporal_convolution_bw	te_periodic_temporal_convolution_sequence_remote_Bw	periodic_temporal_convolution	0.003890	0.003614	0.004163	158,529	1.83 MB	Low	Implemented Benchmark
track2h_latent_state_hysteresis_causal_tcn_offset_residual_bw	te_track2h_l_causal_tcn_offset_residual_bw	latent_state_hysteresis_probe	0.003840	0.003630	0.004312	97,923	1.17 MB	Low	Implemented Benchmark
gru_sequence_bw	te_gru_sequence_remote_Bw	gru_sequence	0.003867	0.003631	0.004297	151,041	1.74 MB	Low	Implemented Benchmark
sequential_residual_offset_probe_bw	te_sequential_residual_offset_probe_remote_bw	sequential_residual_offset_probe	0.003840	0.003638	0.004280	92,802	1.09 MB	Low	Implemented Benchmark
temporal_convolution_bw	te_temporal_convolution_sequence_remote_Bw	temporal_convolution	0.003933	0.003739	0.004369	147,009	1.70 MB	Low	Implemented Benchmark

Cross-Family Interpretation

• Current program-registry winner: te_periodic_gru_sequence_remote_global from family periodic_gru_sequence.

• Strongest current neural family: periodic_gru_sequence.

• Current plain-MLP comparison anchor: te_feedforward_trial.

• Predictive quality and deployment suitability must stay separate: the best leaderboard entry is not automatically the best TwinCAT/PLC candidate.

• Large tree artifacts should be treated cautiously even when tree-based accuracy remains strong, because model weight and memory footprint can dominate deployment feasibility.

Paper Reference Benchmark

The repository benchmark paper is reference/RCIM_ML-compensation.pdf. At the current repository state, the comparison is explicitly offline-only. A real paper-equivalent comparison still requires repository-owned online compensation tests.

Extracted Paper Targets

• Paper dataset size: 1026 operating-condition samples.

• Paper input axes: input speed, applied torque, oil temperature.

• Offline prediction target: TE-curve mean percentage error at or below 4.7% on unseen validation scenarios.

• Online robot compensation target: at least 83.6% TE RMS reduction.

• Online cycloidal compensation target: at least 94.0% TE RMS reduction and 91.7% TE max reduction.

• Paper compensation harmonics baseline: 0, 1, 39 with additional checks on 40, 78.

Paper Vs Repository

Comparison Item	Paper Reference	Repository Status	Current Verdict
Offline model-selection direction	Boosting/tree-heavy deployed harmonic predictors	Current winner te_periodic_gru_sequence_remote_global from family periodic_gru_sequence with model type periodic_gru_sequence	not_aligned
Strongest neural branch role	Neural models are evaluated, but not the primary deployed winners	Strongest repository neural family is periodic_gru_sequence and still trails the tree winner	aligned
RCIM Model-Bank Reproduction canonical closure rule	Paper Tables 3-6 replicated per target and per harmonic	Exact-paper report currently shows 0/0 harmonics fully closed, 0/0 partially closed, 0/0 still open	not_yet_met
Supporting harmonic-wise TE metric	Mean percentage error over full TE curves	Latest harmonic-wise validation reports 11.212% mean percentage error on held-out curves using harmonics 0, 1, 3, 39, 40, 78, 81, 156, 162, 240	supporting_only_not_yet_met
Online robot-profile compensation	TE RMS reduction 83.6%	No repository-owned online compensation result yet	not_yet_comparable
Online cycloidal-profile compensation	TE RMS reduction 94.0%, TE max reduction 91.7%	No repository-owned online compensation result yet	not_yet_comparable
Table 9-style end-to-end benchmark	PLC-integrated motion-profile compensation benchmark	Missing in the repository at the current state	not_yet_comparable

RCIM Model-Bank Reproduction Canonical Status

• Latest exact-paper validation summary: N/A

• Table 3 amplitude RMSE: 0/0 harmonics at or below the paper target

• Table 4 phase MAE: 0/0 harmonics at or below the paper target

• Table 5 phase RMSE: 0/0 harmonics at or below the paper target

• Target-level expected-family direction: 0/0

• Harmonic-level Table 6 closure: 0/0 fully matched, 0/0 partially matched, 0/0 still open

• Highest-priority open harmonics: N/A

Latest Harmonic-Wise Validation Support

• Latest harmonic-wise validation summary: output/validation_checks/paper_reimplementation_rcim_harmonic_wise/forward/family_exploration/rf/2026-04-13-16-00-30__track1_rf_h039_h162240_bridge_control_campaign_run/validation_summary.yaml

• Harmonic-wise test mean percentage error: 11.212%

• Target A status from the latest harmonic-wise run: not_yet_met

Online Compensation Tracking Placeholder

• Repository online compensation status: not yet available.

• When online compensation tests are implemented, update this master summary with TE RMS, TE max, and reduction percentages for both robot and cycloidal motion profiles.

• Until those tests exist, present the paper comparison as offline-only rather than end-to-end equivalent.

Gap Summary

• RCIM Model-Bank Reproduction remains open primarily because the canonical Tables 3-6 are not yet fully matched.

• Offline benchmark scope remains partially comparable rather than like-for-like.

• Not yet aligned: the current repository winner is not tree-based, while the paper deployment path is dominated by boosting/tree models.

• Neural models remain secondary in the repository (periodic_gru_sequence), which is also consistent with the paper not promoting a plain neural winner for deployment.

• End-to-end paper comparison remains not yet comparable until repository-owned online compensation tests exist.

Family-By-Family Result Breakdowns

• For multi-scope waves, family breakdowns are grouped by canonical reporting scope before the per-family ranking tables.

Global Models

feedforward

• Best run: te_feedforward_trial

• Best test MAE: 0.002877

• Completed tracked runs: 4

• Known failed campaign attempts: 1

Rank	Run	Model Type	Test MAE [deg]	Test RMSE [deg]	Val MAE [deg]	Params	Duration	Artifact Size	Model Complexity	Training Heaviness	Campaign
1	te_feedforward_trial	feedforward	0.002877	0.003835	0.002725	26,113	3m 27s	0.32 MB	Medium	Low	polished_dataset_stage1_smoke_2026_06_21
2	te_feedforward_stride1_high_compute_long_remote_global	feedforward	0.003150	0.003603	0.003056	109,953	N/A	1.28 MB	High	Unknown	standalone_or_unknown
3	te_feedforward_stride1_high_compute_long_remote_global_optuna_t0017	feedforward	0.003208	0.003810	0.002962	43,649	N/A	0.52 MB	Medium	Unknown	standalone_or_unknown
4	te_feedforward_stride1_high_compute_long_remote_global_optuna_t0012	feedforward	0.003217	0.003847	0.003014	43,649	N/A	0.52 MB	Medium	Unknown	standalone_or_unknown

Known failed campaign attempts for this family:

• te_feedforward_trial | campaign polished_dataset_stage1_smoke_2026_06_21 | model type feedforward | error invalid literal for int() with base 10: 'auto'

feedforward_recovery_micro

• Best run: te_feedforward_optuna_recovery_micro_global_optuna_t0000

• Best test MAE: 0.004164

• Completed tracked runs: 1

• Known failed campaign attempts: 0

Rank	Run	Model Type	Test MAE [deg]	Test RMSE [deg]	Val MAE [deg]	Params	Duration	Artifact Size	Model Complexity	Training Heaviness	Campaign
1	te_feedforward_optuna_recovery_micro_global_optuna_t0000	feedforward	0.004164	0.005109	0.004266	109,953	N/A	1.28 MB	High	Unknown	standalone_or_unknown

feedforward_recovery_probe_dense

• Best run: te_feedforward_optuna_recovery_probe_dense_global_optuna_t0000

• Best test MAE: 0.004602

• Completed tracked runs: 1

• Known failed campaign attempts: 0

Rank	Run	Model Type	Test MAE [deg]	Test RMSE [deg]	Val MAE [deg]	Params	Duration	Artifact Size	Model Complexity	Training Heaviness	Campaign
1	te_feedforward_optuna_recovery_probe_dense_global_optuna_t0000	feedforward	0.004602	0.005262	0.004257	109,953	N/A	1.28 MB	High	Unknown	standalone_or_unknown

gru_sequence

• Best run: te_gru_sequence_remote_global

• Best test MAE: 0.003591

• Completed tracked runs: 1

• Known failed campaign attempts: 0

Rank	Run	Model Type	Test MAE [deg]	Test RMSE [deg]	Val MAE [deg]	Params	Duration	Artifact Size	Model Complexity	Training Heaviness	Campaign
1	te_gru_sequence_remote_global	gru_sequence	0.003591	0.004110	0.003707	151,041	8m 44s	1.74 MB	Very High	Low	wave2_temporal_model_entry_campaign_2026_05_24_11_01_15

harmonic_regression

• Best run: te_harmonic_order12_linear_conditioned_recovery_global

• Best test MAE: 0.003839

• Completed tracked runs: 7

• Known failed campaign attempts: 0

Rank	Run	Model Type	Test MAE [deg]	Test RMSE [deg]	Val MAE [deg]	Params	Duration	Artifact Size	Model Complexity	Training Heaviness	Campaign
1	te_harmonic_order12_linear_conditioned_recovery_global	harmonic_regression	0.003839	0.004555	0.003904	125	11m 27s	0.01 MB	Very Low	Low	wave1_directional_retraining_campaign_2026_05_06_16_07_16
2	te_harmonic_rcim_sparse_tracking_global	harmonic_regression	0.020767	0.022376	0.016995	114	6m 17s	0.01 MB	Very Low	Low	wave1_high_order_harmonic_tracking_campaign_2026_05_19_17_40_01
3	te_harmonic_order12_linear_conditioned_recovery_global_grid_order12_lr00005_stride5	harmonic_regression	0.020774	0.022412	0.017025	150	N/A	0.01 MB	Very Low	Unknown	standalone_or_unknown
4	te_harmonic_order12_linear_conditioned_recovery_global_grid_order12_lr0001_stride1	harmonic_regression	0.020775	0.022417	0.017013	150	N/A	0.01 MB	Very Low	Unknown	standalone_or_unknown
5	te_harmonic_order12_linear_conditioned_recovery_global	harmonic_regression	0.020779	0.022403	0.017017	150	N/A	0.01 MB	Very Low	Unknown	standalone_or_unknown
6	te_harmonic_dense360_tracking_global	harmonic_regression	0.020780	0.022399	0.016991	4,326	8m 57s	0.06 MB	Low	Low	wave1_high_order_harmonic_tracking_campaign_2026_05_19_17_40_01
7	te_harmonic_dense240_tracking_global	harmonic_regression	0.020787	0.022388	0.016989	2,886	6m 02s	0.04 MB	Low	Low	wave1_high_order_harmonic_tracking_campaign_2026_05_19_17_40_01

lstm_sequence

• Best run: te_lstm_sequence_remote_global

• Best test MAE: 0.003482

• Completed tracked runs: 1

• Known failed campaign attempts: 0

Rank	Run	Model Type	Test MAE [deg]	Test RMSE [deg]	Val MAE [deg]	Params	Duration	Artifact Size	Model Complexity	Training Heaviness	Campaign
1	te_lstm_sequence_remote_global	lstm_sequence	0.003482	0.003948	0.003681	201,345	9m 56s	2.32 MB	Very High	Low	wave2_temporal_model_entry_campaign_2026_05_24_11_01_15

periodic_gru_sequence

• Best run: te_periodic_gru_sequence_remote_global

• Best test MAE: 0.001279

• Completed tracked runs: 2

• Known failed campaign attempts: 0

Rank	Run	Model Type	Test MAE [deg]	Test RMSE [deg]	Val MAE [deg]	Params	Duration	Artifact Size	Model Complexity	Training Heaviness	Campaign
1	te_periodic_gru_sequence_remote_global	periodic_gru_sequence	0.001279	0.001638	0.001274	157,569	40m 03s	1.82 MB	Very High	High	wave2b_harmonic_temporal_hybrid_campaign_2026_05_25
2	te_periodic_gru_sequence_remote_global	periodic_gru_sequence	0.002681	0.002971	0.002507	157,953	1h 00m 14s	1.82 MB	Very High	High	wave2b_harmonic_temporal_hybrid_campaign_2026_05_25

periodic_lstm_sequence

• Best run: te_periodic_lstm_sequence_remote_global

• Best test MAE: 0.002682

• Completed tracked runs: 1

• Known failed campaign attempts: 0

Rank	Run	Model Type	Test MAE [deg]	Test RMSE [deg]	Val MAE [deg]	Params	Duration	Artifact Size	Model Complexity	Training Heaviness	Campaign
1	te_periodic_lstm_sequence_remote_global	periodic_lstm_sequence	0.002682	0.002969	0.002526	210,561	1h 11m 12s	2.43 MB	Very High	High	wave2b_harmonic_temporal_hybrid_campaign_2026_05_25

periodic_mlp

• Best run: te_periodic_mlp_h04_standard_global_optuna_t0010

• Best test MAE: 0.003186

• Completed tracked runs: 6

• Known failed campaign attempts: 0

Rank	Run	Model Type	Test MAE [deg]	Test RMSE [deg]	Val MAE [deg]	Params	Duration	Artifact Size	Model Complexity	Training Heaviness	Campaign
1	te_periodic_mlp_h04_standard_global_optuna_t0010	periodic_mlp	0.003186	0.003690	0.002994	27,265	N/A	0.33 MB	Medium	Unknown	standalone_or_unknown
2	te_periodic_mlp_h04_standard_global_optuna_t0008	periodic_mlp	0.003200	0.003798	0.003057	46,721	N/A	0.56 MB	Medium	Unknown	standalone_or_unknown
3	te_periodic_mlp_h04_standard_global_optuna_t0006	periodic_mlp	0.003233	0.003733	0.002964	27,777	N/A	0.34 MB	Medium	Unknown	standalone_or_unknown
4	te_periodic_mlp_rcim_sparse_tracking_global	periodic_mlp	0.003275	0.003726	0.002863	28,545	7h 47m 34s	0.35 MB	Medium	Very High	wave1_periodic_mlp_explicit_harmonic_tracking_campaign_2026_05_20_22_42_49
5	te_periodic_mlp_dense240_tracking_global	periodic_mlp	0.003348	0.003862	0.002962	87,681	20m 22s	1.03 MB	High	Medium	wave1_periodic_mlp_explicit_harmonic_tracking_campaign_2026_05_20_22_42_49
6	te_periodic_mlp_dense360_tracking_global	periodic_mlp	0.003401	0.003831	0.002859	118,401	50m 45s	1.38 MB	High	High	wave1_periodic_mlp_explicit_harmonic_tracking_campaign_2026_05_20_22_42_49

periodic_temporal_convolution

• Best run: te_periodic_temporal_convolution_sequence_remote_global

• Best test MAE: 0.003508

• Completed tracked runs: 1

• Known failed campaign attempts: 0

Rank	Run	Model Type	Test MAE [deg]	Test RMSE [deg]	Val MAE [deg]	Params	Duration	Artifact Size	Model Complexity	Training Heaviness	Campaign
1	te_periodic_temporal_convolution_sequence_remote_global	periodic_temporal_convolution	0.003508	0.003929	0.003634	158,529	25m 37s	1.83 MB	Very High	Medium	wave2b_harmonic_temporal_hybrid_campaign_2026_05_25

residual_harmonic_gru_sequence_dense240

• Best run: te_residual_harmonic_gru_sequence_remote_global_dense240

• Best test MAE: 0.003511

• Completed tracked runs: 1

• Known failed campaign attempts: 0

Rank	Run	Model Type	Test MAE [deg]	Test RMSE [deg]	Val MAE [deg]	Params	Duration	Artifact Size	Model Complexity	Training Heaviness	Campaign
1	te_residual_harmonic_gru_sequence_remote_global_dense240	residual_harmonic_gru_sequence	0.003511	0.003983	0.003600	151,522	13m 21s	1.75 MB	Very High	Low	wave2c_residual_harmonic_temporal_hybrid_campaign_2026_05_27

residual_harmonic_gru_sequence_dense360

• Best run: te_residual_harmonic_gru_sequence_remote_global_dense360

• Best test MAE: 0.003535

• Completed tracked runs: 1

• Known failed campaign attempts: 0

Rank	Run	Model Type	Test MAE [deg]	Test RMSE [deg]	Val MAE [deg]	Params	Duration	Artifact Size	Model Complexity	Training Heaviness	Campaign
1	te_residual_harmonic_gru_sequence_remote_global_dense360	residual_harmonic_gru_sequence	0.003535	0.003999	0.003628	151,762	21m 39s	1.75 MB	Very High	Medium	wave2c_residual_harmonic_temporal_hybrid_campaign_2026_05_27

residual_harmonic_gru_sequence_sparse_rcim

• Best run: te_residual_harmonic_gru_sequence_remote_global_sparse_rcim

• Best test MAE: 0.002112

• Completed tracked runs: 2

• Known failed campaign attempts: 0

Rank	Run	Model Type	Test MAE [deg]	Test RMSE [deg]	Val MAE [deg]	Params	Duration	Artifact Size	Model Complexity	Training Heaviness	Campaign
1	te_residual_harmonic_gru_sequence_remote_global_sparse_rcim	residual_harmonic_gru_sequence	0.002112	0.002699	0.001978	150,676	24m 21s	1.74 MB	Very High	Medium	wave2c_residual_harmonic_temporal_hybrid_campaign_2026_05_27
2	te_residual_harmonic_gru_sequence_remote_global_sparse_rcim	residual_harmonic_gru_sequence	0.003440	0.003848	0.003607	151,060	11m 44s	1.75 MB	Very High	Low	wave2c_residual_harmonic_temporal_hybrid_campaign_2026_05_27

residual_harmonic_lstm_sequence_dense240

• Best run: te_residual_harmonic_lstm_sequence_remote_global_dense240

• Best test MAE: 0.003473

• Completed tracked runs: 1

• Known failed campaign attempts: 0

Rank	Run	Model Type	Test MAE [deg]	Test RMSE [deg]	Val MAE [deg]	Params	Duration	Artifact Size	Model Complexity	Training Heaviness	Campaign
1	te_residual_harmonic_lstm_sequence_remote_global_dense240	residual_harmonic_lstm_sequence	0.003473	0.003925	0.003624	201,826	13m 54s	2.33 MB	Very High	Low	wave2c_residual_harmonic_temporal_hybrid_campaign_2026_05_27

residual_harmonic_lstm_sequence_dense360

• Best run: te_residual_harmonic_lstm_sequence_remote_global_dense360

• Best test MAE: 0.003477

• Completed tracked runs: 1

• Known failed campaign attempts: 0

Rank	Run	Model Type	Test MAE [deg]	Test RMSE [deg]	Val MAE [deg]	Params	Duration	Artifact Size	Model Complexity	Training Heaviness	Campaign
1	te_residual_harmonic_lstm_sequence_remote_global_dense360	residual_harmonic_lstm_sequence	0.003477	0.003940	0.003648	202,066	28m 49s	2.33 MB	Very High	Medium	wave2c_residual_harmonic_temporal_hybrid_campaign_2026_05_27

residual_harmonic_lstm_sequence_sparse_rcim

• Best run: te_residual_harmonic_lstm_sequence_remote_global_sparse_rcim

• Best test MAE: 0.003368

• Completed tracked runs: 1

• Known failed campaign attempts: 0

Rank	Run	Model Type	Test MAE [deg]	Test RMSE [deg]	Val MAE [deg]	Params	Duration	Artifact Size	Model Complexity	Training Heaviness	Campaign
1	te_residual_harmonic_lstm_sequence_remote_global_sparse_rcim	residual_harmonic_lstm_sequence	0.003368	0.003808	0.003632	201,364	9m 32s	2.32 MB	Very High	Low	wave2c_residual_harmonic_temporal_hybrid_campaign_2026_05_27

residual_harmonic_mlp

• Best run: te_residual_h12_deep_joint_wave1_global_optuna_t0006

• Best test MAE: 0.003034

• Completed tracked runs: 6

• Known failed campaign attempts: 0

Rank	Run	Model Type	Test MAE [deg]	Test RMSE [deg]	Val MAE [deg]	Params	Duration	Artifact Size	Model Complexity	Training Heaviness	Campaign
1	te_residual_h12_deep_joint_wave1_global_optuna_t0006	residual_harmonic_mlp	0.003034	0.003550	0.002895	26,266	N/A	0.32 MB	Medium	Unknown	standalone_or_unknown
2	te_residual_h12_deep_joint_wave1_global_optuna_t0010	residual_harmonic_mlp	0.003067	0.003568	0.002903	26,258	N/A	0.32 MB	Medium	Unknown	standalone_or_unknown
3	te_residual_h12_deep_joint_wave1	residual_harmonic_mlp	0.003152	0.003640	0.003024	26,266	N/A	0.32 MB	Medium	Unknown	standalone_or_unknown
4	te_residual_harmonic_dense240_tracking_global	residual_harmonic_mlp	0.003162	0.003598	0.002976	26,722	11m 07s	0.33 MB	Medium	Low	wave1_high_order_harmonic_tracking_campaign_2026_05_19_17_40_01
5	te_residual_harmonic_rcim_sparse_tracking_global	residual_harmonic_mlp	0.003378	0.003902	0.002969	26,260	8m 03s	0.32 MB	Medium	Low	wave1_high_order_harmonic_tracking_campaign_2026_05_19_17_40_01
6	te_residual_harmonic_dense360_tracking_global	residual_harmonic_mlp	0.003434	0.003957	0.002943	26,962	13m 52s	0.33 MB	Medium	Low	wave1_high_order_harmonic_tracking_campaign_2026_05_19_17_40_01

sequential_residual_offset_probe

• Best run: te_sequential_residual_offset_probe_remote_global

• Best test MAE: 0.003537

• Completed tracked runs: 1

• Known failed campaign attempts: 0

Rank	Run	Model Type	Test MAE [deg]	Test RMSE [deg]	Val MAE [deg]	Params	Duration	Artifact Size	Model Complexity	Training Heaviness	Campaign
1	te_sequential_residual_offset_probe_remote_global	sequential_residual_offset_probe	0.003537	0.004005	0.003783	92,802	9m 22s	1.09 MB	High	Low	track2f_offset_aware_probe_campaign_2026_06_03

temporal_convolution

• Best run: te_temporal_convolution_sequence_remote_global

• Best test MAE: 0.003754

• Completed tracked runs: 1

• Known failed campaign attempts: 0

Rank	Run	Model Type	Test MAE [deg]	Test RMSE [deg]	Val MAE [deg]	Params	Duration	Artifact Size	Model Complexity	Training Heaviness	Campaign
1	te_temporal_convolution_sequence_remote_global	temporal_convolution	0.003754	0.004266	0.003935	147,009	9m 46s	1.70 MB	High	Low	wave2_temporal_model_entry_campaign_2026_05_24_11_01_15

track2f_bis_clean_sequential_residual_offset_global

• Best run: te_track2f_bis_clean_residual_offset_global

• Best test MAE: 0.003528

• Completed tracked runs: 1

• Known failed campaign attempts: 0

Rank	Run	Model Type	Test MAE [deg]	Test RMSE [deg]	Val MAE [deg]	Params	Duration	Artifact Size	Model Complexity	Training Heaviness	Campaign
1	te_track2f_bis_clean_residual_offset_global	sequential_residual_offset_probe	0.003528	0.004010	0.003717	92,802	11m 40s	1.09 MB	High	Low	track2f_bis_harmonic_offset_probe_campaign_2026_06_04

track2f_bis_harmonic_residual_offset_global

• Best run: te_track2f_bis_harmonic_residual_offset_global

• Best test MAE: 0.003538

• Completed tracked runs: 1

• Known failed campaign attempts: 1

Rank	Run	Model Type	Test MAE [deg]	Test RMSE [deg]	Val MAE [deg]	Params	Duration	Artifact Size	Model Complexity	Training Heaviness	Campaign
1	te_track2f_bis_harmonic_residual_offset_global	harmonic_residual_offset_probe	0.003538	0.003932	0.003659	85,747	0s	1.00 MB	High	Very Low	track2f_bis_harmonic_offset_probe_campaign_2026_06_04

Known failed campaign attempts for this family:

• te_track2f_bis_harmonic_residual_offset_global | campaign track2f_bis_harmonic_offset_probe_campaign_2026_06_04 | model type harmonic_residual_offset_probe | error Unsupported Model Type for Campaign Runner | harmonic_residual_offset_probe | Supported: ['feedforward', 'gru_sequence', 'harmonic_regression', 'hist_gradient_boosting', 'lstm_sequence', 'periodic_gru_sequence', 'periodic_lstm_sequence', 'periodic_mlp', 'periodic_temporal_convolution', 'random_forest', 'residual_harmonic_gru_sequence', 'residual_harmonic_lstm_sequence', 'residual_harmonic_mlp', 'sequential_residual_offset_probe', 'temporal_convolution']

track2g_curve_aware_harmonic_residual_offset_full_curve_composite_global

• Best run: te_track2g_curve_aware_full_curve_composite_global

• Best test MAE: 0.002008

• Completed tracked runs: 2

• Known failed campaign attempts: 0

Rank	Run	Model Type	Test MAE [deg]	Test RMSE [deg]	Val MAE [deg]	Params	Duration	Artifact Size	Model Complexity	Training Heaviness	Campaign
1	te_track2g_curve_aware_full_curve_composite_global	curve_aware_harmonic_residual_offset_probe	0.002008	0.002581	0.001872	85,440	44m 57s	1.00 MB	High	High	track2g_curve_aware_training_campaign_2026_06_08
2	te_track2g_curve_aware_full_curve_composite_global	curve_aware_harmonic_residual_offset_probe	0.003345	0.003713	0.003616	85,747	32m 15s	1.00 MB	High	Medium	track2g_curve_aware_training_campaign_2026_06_08

track2g_curve_aware_harmonic_residual_offset_pointwise_control_global

• Best run: te_track2g_curve_aware_pointwise_control_global

• Best test MAE: 0.003587

• Completed tracked runs: 1

• Known failed campaign attempts: 0

Rank	Run	Model Type	Test MAE [deg]	Test RMSE [deg]	Val MAE [deg]	Params	Duration	Artifact Size	Model Complexity	Training Heaviness	Campaign
1	te_track2g_curve_aware_pointwise_control_global	curve_aware_harmonic_residual_offset_probe	0.003587	0.004001	0.003607	85,747	20m 29s	1.00 MB	High	Medium	track2g_curve_aware_training_campaign_2026_06_08

track2g_curve_aware_harmonic_residual_offset_raw_centered_shape_global

• Best run: te_track2g_curve_aware_raw_centered_shape_global

• Best test MAE: 0.003350

• Completed tracked runs: 1

• Known failed campaign attempts: 0

Rank	Run	Model Type	Test MAE [deg]	Test RMSE [deg]	Val MAE [deg]	Params	Duration	Artifact Size	Model Complexity	Training Heaviness	Campaign
1	te_track2g_curve_aware_raw_centered_shape_global	curve_aware_harmonic_residual_offset_probe	0.003350	0.003753	0.003636	85,747	22m 08s	1.00 MB	High	Medium	track2g_curve_aware_training_campaign_2026_06_08

track2g_curve_aware_harmonic_residual_offset_raw_offset_global

• Best run: te_track2g_curve_aware_raw_offset_global

• Best test MAE: 0.003465

• Completed tracked runs: 1

• Known failed campaign attempts: 0

Rank	Run	Model Type	Test MAE [deg]	Test RMSE [deg]	Val MAE [deg]	Params	Duration	Artifact Size	Model Complexity	Training Heaviness	Campaign
1	te_track2g_curve_aware_raw_offset_global	curve_aware_harmonic_residual_offset_probe	0.003465	0.003829	0.003564	85,747	32m 11s	1.00 MB	High	Medium	track2g_curve_aware_training_campaign_2026_06_08

track2h_dispersion_aware_log_cosh_robust_global

• Best run: te_track2h_log_cosh_robust_global

• Best test MAE: 0.003505

• Completed tracked runs: 1

• Known failed campaign attempts: 0

Rank	Run	Model Type	Test MAE [deg]	Test RMSE [deg]	Val MAE [deg]	Params	Duration	Artifact Size	Model Complexity	Training Heaviness	Campaign
1	te_track2h_log_cosh_robust_global	curve_aware_harmonic_residual_offset_probe	0.003505	0.003935	0.003645	85,747	18m 16s	1.00 MB	High	Medium	track2h_dispersion_aware_modeling_campaign_2026_06_10

track2h_dispersion_aware_mae_robust_global

• Best run: te_track2h_mae_robust_global

• Best test MAE: 0.003406

• Completed tracked runs: 1

• Known failed campaign attempts: 0

Rank	Run	Model Type	Test MAE [deg]	Test RMSE [deg]	Val MAE [deg]	Params	Duration	Artifact Size	Model Complexity	Training Heaviness	Campaign
1	te_track2h_mae_robust_global	curve_aware_harmonic_residual_offset_probe	0.003406	0.003807	0.003645	85,747	16m 33s	1.00 MB	High	Medium	track2h_dispersion_aware_modeling_campaign_2026_06_10

track2h_dispersion_aware_smooth_l1_robust_global

• Best run: te_track2h_smooth_l1_robust_global

• Best test MAE: 0.003422

• Completed tracked runs: 1

• Known failed campaign attempts: 0

Rank	Run	Model Type	Test MAE [deg]	Test RMSE [deg]	Val MAE [deg]	Params	Duration	Artifact Size	Model Complexity	Training Heaviness	Campaign
1	te_track2h_smooth_l1_robust_global	curve_aware_harmonic_residual_offset_probe	0.003422	0.003810	0.003641	85,747	15m 30s	1.00 MB	High	Medium	track2h_dispersion_aware_modeling_campaign_2026_06_10

track2h_latent_state_hysteresis_causal_tcn_offset_residual_global

• Best run: te_track2h_l_causal_tcn_offset_residual_global

• Best test MAE: 0.003368

• Completed tracked runs: 1

• Known failed campaign attempts: 0

Rank	Run	Model Type	Test MAE [deg]	Test RMSE [deg]	Val MAE [deg]	Params	Duration	Artifact Size	Model Complexity	Training Heaviness	Campaign
1	te_track2h_l_causal_tcn_offset_residual_global	latent_state_hysteresis_probe	0.003368	0.003860	0.003543	97,923	28m 35s	1.17 MB	High	Medium	track2h_latent_state_hysteresis_campaign_2026_06_16

track2h_latent_state_hysteresis_gru_offset_residual_global

• Best run: te_track2h_l_gru_offset_residual_global

• Best test MAE: 0.002339

• Completed tracked runs: 2

• Known failed campaign attempts: 0

Rank	Run	Model Type	Test MAE [deg]	Test RMSE [deg]	Val MAE [deg]	Params	Duration	Artifact Size	Model Complexity	Training Heaviness	Campaign
1	te_track2h_l_gru_offset_residual_global	latent_state_hysteresis_probe	0.002339	0.002986	0.002232	124,899	23m 03s	1.48 MB	High	Medium	track2h_latent_state_hysteresis_campaign_2026_06_16
2	te_track2h_l_gru_offset_residual_global	latent_state_hysteresis_probe	0.003590	0.004074	0.003717	125,475	17m 28s	1.48 MB	High	Medium	track2h_latent_state_hysteresis_campaign_2026_06_16

track2h_mixture_density_heads_mdn_k2_global

• Best run: te_track2h_mdn_k2_global

• Best test MAE: 0.003503

• Completed tracked runs: 1

• Known failed campaign attempts: 0

Rank	Run	Model Type	Test MAE [deg]	Test RMSE [deg]	Val MAE [deg]	Params	Duration	Artifact Size	Model Complexity	Training Heaviness	Campaign
1	te_track2h_mdn_k2_global	curve_aware_harmonic_residual_offset_probe	0.003503	0.003938	0.003654	86,802	20m 19s	1.01 MB	High	Medium	track2h_mixture_density_heads_campaign_2026_06_13

track2h_mixture_density_heads_mdn_k3_global

• Best run: te_track2h_mdn_k3_global

• Best test MAE: 0.003564

• Completed tracked runs: 1

• Known failed campaign attempts: 0

Rank	Run	Model Type	Test MAE [deg]	Test RMSE [deg]	Val MAE [deg]	Params	Duration	Artifact Size	Model Complexity	Training Heaviness	Campaign
1	te_track2h_mdn_k3_global	curve_aware_harmonic_residual_offset_probe	0.003564	0.003986	0.003617	87,435	20m 24s	1.02 MB	High	Medium	track2h_mixture_density_heads_campaign_2026_06_13

track2h_quantile_probabilistic_gaussian_nll_global

• Best run: te_track2h_gaussian_nll_global

• Best test MAE: 0.003013

• Completed tracked runs: 1

• Known failed campaign attempts: 0

Rank	Run	Model Type	Test MAE [deg]	Test RMSE [deg]	Val MAE [deg]	Params	Duration	Artifact Size	Model Complexity	Training Heaviness	Campaign
1	te_track2h_gaussian_nll_global	curve_aware_harmonic_residual_offset_probe	0.003013	0.003388	0.003267	85,958	1h 02m 12s	1.00 MB	High	High	track2h_quantile_probabilistic_campaign_2026_06_12

track2h_quantile_probabilistic_quantile_p10_p50_p90_global

• Best run: te_track2h_quantile_p10_p50_p90_global

• Best test MAE: 0.003383

• Completed tracked runs: 1

• Known failed campaign attempts: 0

Rank	Run	Model Type	Test MAE [deg]	Test RMSE [deg]	Val MAE [deg]	Params	Duration	Artifact Size	Model Complexity	Training Heaviness	Campaign
1	te_track2h_quantile_p10_p50_p90_global	curve_aware_harmonic_residual_offset_probe	0.003383	0.003764	0.003606	86,169	18m 38s	1.01 MB	High	Medium	track2h_quantile_probabilistic_campaign_2026_06_12

tree

• Best run: te_hist_gbr_tabular_global

• Best test MAE: 0.001753

• Completed tracked runs: 4

• Known failed campaign attempts: 0

Rank	Run	Model Type	Test MAE [deg]	Test RMSE [deg]	Val MAE [deg]	Params	Duration	Artifact Size	Model Complexity	Training Heaviness	Campaign
1	te_hist_gbr_tabular_global	hist_gradient_boosting	0.001753	0.002892	0.001591	4	1m 53s	0.44 MB	Light Artifact	Very Low	wave1_directional_retraining_campaign_2026_05_06_16_07_16
2	te_hist_gbr_tabular_global_grid_depth10_lr008_leaf10	hist_gradient_boosting	0.002782	0.003520	0.002655	5	N/A	0.48 MB	Light Artifact	Unknown	standalone_or_unknown
3	te_hist_gbr_tabular_global_grid_depth10_lr008_leaf20	hist_gradient_boosting	0.002782	0.003520	0.002655	5	N/A	0.48 MB	Light Artifact	Unknown	standalone_or_unknown
4	te_hist_gbr_tabular_global_grid_depth8_lr008_leaf10	hist_gradient_boosting	0.002830	0.003585	0.002677	5	N/A	0.50 MB	Light Artifact	Unknown	standalone_or_unknown

wave3_harmonic_prior_residual_pointwise_control_global

• Best run: te_wave3_harmonic_prior_residual_pointwise_control_global

• Best test MAE: 0.003451

• Completed tracked runs: 1

• Known failed campaign attempts: 0

Rank	Run	Model Type	Test MAE [deg]	Test RMSE [deg]	Val MAE [deg]	Params	Duration	Artifact Size	Model Complexity	Training Heaviness	Campaign
1	te_wave3_harmonic_prior_residual_pointwise_control_global	wave3_harmonic_prior_residual	0.003451	0.003851	0.003611	7,283	26m 08s	0.11 MB	Low	Medium	wave3_harmonic_prior_residual_campaign_2026_06_14

wave3_harmonic_prior_residual_smooth_l1_structured_global

• Best run: te_wave3_harmonic_prior_residual_smooth_l1_structured_global

• Best test MAE: 0.002168

• Completed tracked runs: 2

• Known failed campaign attempts: 0

Rank	Run	Model Type	Test MAE [deg]	Test RMSE [deg]	Val MAE [deg]	Params	Duration	Artifact Size	Model Complexity	Training Heaviness	Campaign
1	te_wave3_harmonic_prior_residual_smooth_l1_structured_global	wave3_harmonic_prior_residual	0.002168	0.002763	0.001889	7,168	21m 31s	0.10 MB	Low	Medium	wave3_harmonic_prior_residual_campaign_2026_06_14
2	te_wave3_harmonic_prior_residual_smooth_l1_structured_global	wave3_harmonic_prior_residual	0.003403	0.003785	0.003633	7,283	19m 38s	0.11 MB	Low	Medium	wave3_harmonic_prior_residual_campaign_2026_06_14

Forward Models

feedforward_fw

• Best run: te_feedforward_stride1_high_compute_long_remote_Fw_optuna_t0008

• Best test MAE: 0.003203

• Completed tracked runs: 3

• Known failed campaign attempts: 0

Rank	Run	Model Type	Test MAE [deg]	Test RMSE [deg]	Val MAE [deg]	Params	Duration	Artifact Size	Model Complexity	Training Heaviness	Campaign
1	te_feedforward_stride1_high_compute_long_remote_Fw_optuna_t0008	feedforward	0.003203	0.003787	0.002850	109,953	N/A	1.28 MB	High	Unknown	standalone_or_unknown
2	te_feedforward_stride1_high_compute_long_remote_Fw_optuna_t0009	feedforward	0.003229	0.003774	0.002850	143,745	N/A	1.67 MB	High	Unknown	standalone_or_unknown
3	te_feedforward_stride1_high_compute_long_remote_Fw_optuna_t0014	feedforward	0.003232	0.003812	0.002846	43,649	N/A	0.52 MB	Medium	Unknown	standalone_or_unknown

gru_sequence_fw

• Best run: te_gru_sequence_remote_Fw

• Best test MAE: 0.003333

• Completed tracked runs: 1

• Known failed campaign attempts: 0

Rank	Run	Model Type	Test MAE [deg]	Test RMSE [deg]	Val MAE [deg]	Params	Duration	Artifact Size	Model Complexity	Training Heaviness	Campaign
1	te_gru_sequence_remote_Fw	gru_sequence	0.003333	0.003881	0.003409	151,041	6m 01s	1.74 MB	Very High	Low	wave2_temporal_model_entry_campaign_2026_05_24_11_01_15

harmonic_regression_fw

• Best run: te_harmonic_dense360_tracking_Fw

• Best test MAE: 0.002916

• Completed tracked runs: 6

• Known failed campaign attempts: 0

Rank	Run	Model Type	Test MAE [deg]	Test RMSE [deg]	Val MAE [deg]	Params	Duration	Artifact Size	Model Complexity	Training Heaviness	Campaign
1	te_harmonic_dense360_tracking_Fw	harmonic_regression	0.002916	0.003237	0.002610	4,326	7m 00s	0.06 MB	Low	Low	wave1_high_order_harmonic_tracking_campaign_2026_05_19_17_40_01
2	te_harmonic_dense240_tracking_Fw	harmonic_regression	0.002935	0.003239	0.002593	2,886	5m 56s	0.04 MB	Low	Low	wave1_high_order_harmonic_tracking_campaign_2026_05_19_17_40_01
3	te_harmonic_rcim_sparse_tracking_Fw	harmonic_regression	0.002943	0.003254	0.002566	114	5m 05s	0.01 MB	Very Low	Low	wave1_high_order_harmonic_tracking_campaign_2026_05_19_17_40_01
4	te_harmonic_order12_linear_conditioned_recovery_Fw_grid_order8_lr00005_stride5	harmonic_regression	0.003101	0.003527	0.002848	102	N/A	0.01 MB	Very Low	Unknown	standalone_or_unknown
5	te_harmonic_order12_linear_conditioned_recovery_Fw_grid_order12_lr00005_stride5	harmonic_regression	0.003102	0.003528	0.002843	150	N/A	0.01 MB	Very Low	Unknown	standalone_or_unknown
6	te_harmonic_order12_linear_conditioned_recovery_Fw_grid_order12_lr00005_stride1	harmonic_regression	0.003105	0.003534	0.002839	150	N/A	0.01 MB	Very Low	Unknown	standalone_or_unknown

lstm_sequence_fw

• Best run: te_lstm_sequence_remote_Fw

• Best test MAE: 0.003370

• Completed tracked runs: 1

• Known failed campaign attempts: 0

Rank	Run	Model Type	Test MAE [deg]	Test RMSE [deg]	Val MAE [deg]	Params	Duration	Artifact Size	Model Complexity	Training Heaviness	Campaign
1	te_lstm_sequence_remote_Fw	lstm_sequence	0.003370	0.003921	0.003448	201,345	4m 31s	2.32 MB	Very High	Low	wave2_temporal_model_entry_campaign_2026_05_24_11_01_15

periodic_gru_sequence_fw

• Best run: te_periodic_gru_sequence_remote_Fw

• Best test MAE: 0.003193

• Completed tracked runs: 1

• Known failed campaign attempts: 0

Rank	Run	Model Type	Test MAE [deg]	Test RMSE [deg]	Val MAE [deg]	Params	Duration	Artifact Size	Model Complexity	Training Heaviness	Campaign
1	te_periodic_gru_sequence_remote_Fw	periodic_gru_sequence	0.003193	0.003583	0.003227	157,953	11m 11s	1.82 MB	Very High	Low	wave2b_harmonic_temporal_hybrid_campaign_2026_05_25

periodic_lstm_sequence_fw

• Best run: te_periodic_lstm_sequence_remote_Fw

• Best test MAE: 0.003274

• Completed tracked runs: 1

• Known failed campaign attempts: 0

Rank	Run	Model Type	Test MAE [deg]	Test RMSE [deg]	Val MAE [deg]	Params	Duration	Artifact Size	Model Complexity	Training Heaviness	Campaign
1	te_periodic_lstm_sequence_remote_Fw	periodic_lstm_sequence	0.003274	0.003651	0.003254	210,561	9m 20s	2.43 MB	Very High	Low	wave2b_harmonic_temporal_hybrid_campaign_2026_05_25

periodic_mlp_fw

• Best run: te_periodic_mlp_dense240_tracking_Fw

• Best test MAE: 0.003055

• Completed tracked runs: 6

• Known failed campaign attempts: 0

Rank	Run	Model Type	Test MAE [deg]	Test RMSE [deg]	Val MAE [deg]	Params	Duration	Artifact Size	Model Complexity	Training Heaviness	Campaign
1	te_periodic_mlp_dense240_tracking_Fw	periodic_mlp	0.003055	0.003537	0.002541	87,681	13m 21s	1.03 MB	High	Low	wave1_periodic_mlp_explicit_harmonic_tracking_campaign_2026_05_20_22_42_49
2	te_periodic_mlp_rcim_sparse_tracking_Fw	periodic_mlp	0.003131	0.003578	0.002516	28,545	9m 28s	0.35 MB	Medium	Low	wave1_periodic_mlp_explicit_harmonic_tracking_campaign_2026_05_20_22_42_49
3	te_periodic_mlp_dense360_tracking_Fw	periodic_mlp	0.003155	0.003680	0.002524	118,401	12m 15s	1.38 MB	High	Low	wave1_periodic_mlp_explicit_harmonic_tracking_campaign_2026_05_20_22_42_49
4	te_periodic_mlp_h04_standard_Fw_optuna_t0008	periodic_mlp	0.003287	0.003833	0.002809	46,721	N/A	0.56 MB	Medium	Unknown	standalone_or_unknown
5	te_periodic_mlp_h04_standard_Fw_optuna_t0001	periodic_mlp	0.003294	0.003899	0.002751	27,777	N/A	0.34 MB	Medium	Unknown	standalone_or_unknown
6	te_periodic_mlp_h04_standard_Fw_optuna_t0015	periodic_mlp	0.003296	0.003924	0.002802	28,289	N/A	0.35 MB	Medium	Unknown	standalone_or_unknown

periodic_temporal_convolution_fw

• Best run: te_periodic_temporal_convolution_sequence_remote_Fw

• Best test MAE: 0.003337

• Completed tracked runs: 1

• Known failed campaign attempts: 0

Rank	Run	Model Type	Test MAE [deg]	Test RMSE [deg]	Val MAE [deg]	Params	Duration	Artifact Size	Model Complexity	Training Heaviness	Campaign
1	te_periodic_temporal_convolution_sequence_remote_Fw	periodic_temporal_convolution	0.003337	0.003830	0.003321	158,529	8m 15s	1.83 MB	Very High	Low	wave2b_harmonic_temporal_hybrid_campaign_2026_05_25

residual_harmonic_gru_sequence_fw_dense240

• Best run: te_residual_harmonic_gru_sequence_remote_Fw_dense240

• Best test MAE: 0.003219

• Completed tracked runs: 1

• Known failed campaign attempts: 0

Rank	Run	Model Type	Test MAE [deg]	Test RMSE [deg]	Val MAE [deg]	Params	Duration	Artifact Size	Model Complexity	Training Heaviness	Campaign
1	te_residual_harmonic_gru_sequence_remote_Fw_dense240	residual_harmonic_gru_sequence	0.003219	0.003653	0.003270	151,522	8m 13s	1.75 MB	Very High	Low	wave2c_residual_harmonic_temporal_hybrid_campaign_2026_05_27

residual_harmonic_gru_sequence_fw_dense360

• Best run: te_residual_harmonic_gru_sequence_remote_Fw_dense360

• Best test MAE: 0.003241

• Completed tracked runs: 1

• Known failed campaign attempts: 0

Rank	Run	Model Type	Test MAE [deg]	Test RMSE [deg]	Val MAE [deg]	Params	Duration	Artifact Size	Model Complexity	Training Heaviness	Campaign
1	te_residual_harmonic_gru_sequence_remote_Fw_dense360	residual_harmonic_gru_sequence	0.003241	0.003677	0.003265	151,762	11m 14s	1.75 MB	Very High	Low	wave2c_residual_harmonic_temporal_hybrid_campaign_2026_05_27

residual_harmonic_gru_sequence_fw_sparse_rcim

• Best run: te_residual_harmonic_gru_sequence_remote_Fw_sparse_rcim

• Best test MAE: 0.003200

• Completed tracked runs: 1

• Known failed campaign attempts: 0

Rank	Run	Model Type	Test MAE [deg]	Test RMSE [deg]	Val MAE [deg]	Params	Duration	Artifact Size	Model Complexity	Training Heaviness	Campaign
1	te_residual_harmonic_gru_sequence_remote_Fw_sparse_rcim	residual_harmonic_gru_sequence	0.003200	0.003635	0.003309	151,060	5m 07s	1.75 MB	Very High	Low	wave2c_residual_harmonic_temporal_hybrid_campaign_2026_05_27

residual_harmonic_lstm_sequence_fw_dense240

• Best run: te_residual_harmonic_lstm_sequence_remote_Fw_dense240

• Best test MAE: 0.003262

• Completed tracked runs: 1

• Known failed campaign attempts: 0

Rank	Run	Model Type	Test MAE [deg]	Test RMSE [deg]	Val MAE [deg]	Params	Duration	Artifact Size	Model Complexity	Training Heaviness	Campaign
1	te_residual_harmonic_lstm_sequence_remote_Fw_dense240	residual_harmonic_lstm_sequence	0.003262	0.003706	0.003307	201,826	7m 24s	2.33 MB	Very High	Low	wave2c_residual_harmonic_temporal_hybrid_campaign_2026_05_27

residual_harmonic_lstm_sequence_fw_dense360

• Best run: te_residual_harmonic_lstm_sequence_remote_Fw_dense360

• Best test MAE: 0.003351

• Completed tracked runs: 1

• Known failed campaign attempts: 0

Rank	Run	Model Type	Test MAE [deg]	Test RMSE [deg]	Val MAE [deg]	Params	Duration	Artifact Size	Model Complexity	Training Heaviness	Campaign
1	te_residual_harmonic_lstm_sequence_remote_Fw_dense360	residual_harmonic_lstm_sequence	0.003351	0.003774	0.003302	202,066	10m 20s	2.33 MB	Very High	Low	wave2c_residual_harmonic_temporal_hybrid_campaign_2026_05_27

residual_harmonic_lstm_sequence_fw_sparse_rcim

• Best run: te_residual_harmonic_lstm_sequence_remote_Fw_sparse_rcim

• Best test MAE: 0.003234

• Completed tracked runs: 1

• Known failed campaign attempts: 0

Rank	Run	Model Type	Test MAE [deg]	Test RMSE [deg]	Val MAE [deg]	Params	Duration	Artifact Size	Model Complexity	Training Heaviness	Campaign
1	te_residual_harmonic_lstm_sequence_remote_Fw_sparse_rcim	residual_harmonic_lstm_sequence	0.003234	0.003679	0.003344	201,364	4m 50s	2.32 MB	Very High	Low	wave2c_residual_harmonic_temporal_hybrid_campaign_2026_05_27

residual_harmonic_mlp_fw

• Best run: te_residual_harmonic_rcim_sparse_tracking_Fw

• Best test MAE: 0.003089

• Completed tracked runs: 6

• Known failed campaign attempts: 0

Rank	Run	Model Type	Test MAE [deg]	Test RMSE [deg]	Val MAE [deg]	Params	Duration	Artifact Size	Model Complexity	Training Heaviness	Campaign
1	te_residual_harmonic_rcim_sparse_tracking_Fw	residual_harmonic_mlp	0.003089	0.003498	0.002704	26,260	4m 56s	0.32 MB	Medium	Low	wave1_high_order_harmonic_tracking_campaign_2026_05_19_17_40_01
2	te_residual_h12_deep_joint_wave1_Fw_optuna_t0005	residual_harmonic_mlp	0.003168	0.003871	0.002870	34,978	N/A	0.42 MB	Medium	Unknown	standalone_or_unknown
3	te_residual_h12_deep_joint_wave1_Fw_optuna_t0006	residual_harmonic_mlp	0.003194	0.003809	0.002827	26,266	N/A	0.32 MB	Medium	Unknown	standalone_or_unknown
4	te_residual_h12_deep_joint_wave1_Fw_optuna_t0009	residual_harmonic_mlp	0.003211	0.003828	0.002794	34,970	N/A	0.42 MB	Medium	Unknown	standalone_or_unknown
5	te_residual_harmonic_dense240_tracking_Fw	residual_harmonic_mlp	0.003304	0.003773	0.002649	26,722	5m 04s	0.33 MB	Medium	Low	wave1_high_order_harmonic_tracking_campaign_2026_05_19_17_40_01
6	te_residual_harmonic_dense360_tracking_Fw	residual_harmonic_mlp	0.003568	0.004118	0.002598	26,962	6m 12s	0.33 MB	Medium	Low	wave1_high_order_harmonic_tracking_campaign_2026_05_19_17_40_01

sequential_residual_offset_probe_fw

• Best run: te_sequential_residual_offset_probe_remote_fw

• Best test MAE: 0.003385

• Completed tracked runs: 1

• Known failed campaign attempts: 0

Rank	Run	Model Type	Test MAE [deg]	Test RMSE [deg]	Val MAE [deg]	Params	Duration	Artifact Size	Model Complexity	Training Heaviness	Campaign
1	te_sequential_residual_offset_probe_remote_fw	sequential_residual_offset_probe	0.003385	0.003931	0.003380	92,802	12m 09s	1.09 MB	High	Low	track2f_offset_aware_probe_campaign_2026_06_03

temporal_convolution_fw

• Best run: te_temporal_convolution_sequence_remote_Fw

• Best test MAE: 0.003611

• Completed tracked runs: 1

• Known failed campaign attempts: 0

Rank	Run	Model Type	Test MAE [deg]	Test RMSE [deg]	Val MAE [deg]	Params	Duration	Artifact Size	Model Complexity	Training Heaviness	Campaign
1	te_temporal_convolution_sequence_remote_Fw	temporal_convolution	0.003611	0.004183	0.003490	147,009	6m 45s	1.70 MB	High	Low	wave2_temporal_model_entry_campaign_2026_05_24_11_01_15

track2f_bis_clean_sequential_residual_offset_fw

• Best run: te_track2f_bis_clean_residual_offset_fw

• Best test MAE: 0.003446

• Completed tracked runs: 1

• Known failed campaign attempts: 0

Rank	Run	Model Type	Test MAE [deg]	Test RMSE [deg]	Val MAE [deg]	Params	Duration	Artifact Size	Model Complexity	Training Heaviness	Campaign
1	te_track2f_bis_clean_residual_offset_fw	sequential_residual_offset_probe	0.003446	0.003972	0.003474	92,802	5m 16s	1.09 MB	High	Low	track2f_bis_harmonic_offset_probe_campaign_2026_06_04

track2f_bis_harmonic_residual_offset_fw

• Best run: te_track2f_bis_harmonic_residual_offset_fw

• Best test MAE: 0.002862

• Completed tracked runs: 1

• Known failed campaign attempts: 1

Rank	Run	Model Type	Test MAE [deg]	Test RMSE [deg]	Val MAE [deg]	Params	Duration	Artifact Size	Model Complexity	Training Heaviness	Campaign
1	te_track2f_bis_harmonic_residual_offset_fw	harmonic_residual_offset_probe	0.002862	0.003334	0.002941	85,747	0s	1.00 MB	High	Very Low	track2f_bis_harmonic_offset_probe_campaign_2026_06_04

Known failed campaign attempts for this family:

• te_track2f_bis_harmonic_residual_offset_fw | campaign track2f_bis_harmonic_offset_probe_campaign_2026_06_04 | model type harmonic_residual_offset_probe | error Unsupported Model Type for Campaign Runner | harmonic_residual_offset_probe | Supported: ['feedforward', 'gru_sequence', 'harmonic_regression', 'hist_gradient_boosting', 'lstm_sequence', 'periodic_gru_sequence', 'periodic_lstm_sequence', 'periodic_mlp', 'periodic_temporal_convolution', 'random_forest', 'residual_harmonic_gru_sequence', 'residual_harmonic_lstm_sequence', 'residual_harmonic_mlp', 'sequential_residual_offset_probe', 'temporal_convolution']

track2g_curve_aware_harmonic_residual_offset_full_curve_composite_fw

• Best run: te_track2g_curve_aware_full_curve_composite_fw

• Best test MAE: 0.003260

• Completed tracked runs: 1

• Known failed campaign attempts: 0

Rank	Run	Model Type	Test MAE [deg]	Test RMSE [deg]	Val MAE [deg]	Params	Duration	Artifact Size	Model Complexity	Training Heaviness	Campaign
1	te_track2g_curve_aware_full_curve_composite_fw	curve_aware_harmonic_residual_offset_probe	0.003260	0.003630	0.003320	85,747	10m 35s	1.00 MB	High	Low	track2g_curve_aware_training_campaign_2026_06_08

track2g_curve_aware_harmonic_residual_offset_pointwise_control_fw

• Best run: te_track2g_curve_aware_pointwise_control_fw

• Best test MAE: 0.003371

• Completed tracked runs: 1

• Known failed campaign attempts: 0

Rank	Run	Model Type	Test MAE [deg]	Test RMSE [deg]	Val MAE [deg]	Params	Duration	Artifact Size	Model Complexity	Training Heaviness	Campaign
1	te_track2g_curve_aware_pointwise_control_fw	curve_aware_harmonic_residual_offset_probe	0.003371	0.003763	0.003291	85,747	11m 40s	1.00 MB	High	Low	track2g_curve_aware_training_campaign_2026_06_08

track2g_curve_aware_harmonic_residual_offset_raw_centered_shape_fw

• Best run: te_track2g_curve_aware_raw_centered_shape_fw

• Best test MAE: 0.003181

• Completed tracked runs: 1

• Known failed campaign attempts: 0

Rank	Run	Model Type	Test MAE [deg]	Test RMSE [deg]	Val MAE [deg]	Params	Duration	Artifact Size	Model Complexity	Training Heaviness	Campaign
1	te_track2g_curve_aware_raw_centered_shape_fw	curve_aware_harmonic_residual_offset_probe	0.003181	0.003571	0.003251	85,747	10m 48s	1.00 MB	High	Low	track2g_curve_aware_training_campaign_2026_06_08

track2g_curve_aware_harmonic_residual_offset_raw_offset_fw

• Best run: te_track2g_curve_aware_raw_offset_fw

• Best test MAE: 0.003279

• Completed tracked runs: 1

• Known failed campaign attempts: 0

Rank	Run	Model Type	Test MAE [deg]	Test RMSE [deg]	Val MAE [deg]	Params	Duration	Artifact Size	Model Complexity	Training Heaviness	Campaign
1	te_track2g_curve_aware_raw_offset_fw	curve_aware_harmonic_residual_offset_probe	0.003279	0.003698	0.003328	85,747	7m 42s	1.00 MB	High	Low	track2g_curve_aware_training_campaign_2026_06_08

track2h_dispersion_aware_log_cosh_robust_fw

• Best run: te_track2h_log_cosh_robust_fw

• Best test MAE: 0.003355

• Completed tracked runs: 1

• Known failed campaign attempts: 0

Rank	Run	Model Type	Test MAE [deg]	Test RMSE [deg]	Val MAE [deg]	Params	Duration	Artifact Size	Model Complexity	Training Heaviness	Campaign
1	te_track2h_log_cosh_robust_fw	curve_aware_harmonic_residual_offset_probe	0.003355	0.003708	0.003280	85,747	7m 56s	1.00 MB	High	Low	track2h_dispersion_aware_modeling_campaign_2026_06_10

track2h_dispersion_aware_mae_robust_fw

• Best run: te_track2h_mae_robust_fw

• Best test MAE: 0.003146

• Completed tracked runs: 1

• Known failed campaign attempts: 0

Rank	Run	Model Type	Test MAE [deg]	Test RMSE [deg]	Val MAE [deg]	Params	Duration	Artifact Size	Model Complexity	Training Heaviness	Campaign
1	te_track2h_mae_robust_fw	curve_aware_harmonic_residual_offset_probe	0.003146	0.003527	0.003258	85,747	7m 09s	1.00 MB	High	Low	track2h_dispersion_aware_modeling_campaign_2026_06_10

track2h_dispersion_aware_smooth_l1_robust_fw

• Best run: te_track2h_smooth_l1_robust_fw

• Best test MAE: 0.003314

• Completed tracked runs: 1

• Known failed campaign attempts: 0

Rank	Run	Model Type	Test MAE [deg]	Test RMSE [deg]	Val MAE [deg]	Params	Duration	Artifact Size	Model Complexity	Training Heaviness	Campaign
1	te_track2h_smooth_l1_robust_fw	curve_aware_harmonic_residual_offset_probe	0.003314	0.003679	0.003235	85,747	7m 42s	1.00 MB	High	Low	track2h_dispersion_aware_modeling_campaign_2026_06_10

track2h_latent_state_hysteresis_causal_tcn_offset_residual_fw

• Best run: te_track2h_l_causal_tcn_offset_residual_fw

• Best test MAE: 0.003470

• Completed tracked runs: 1

• Known failed campaign attempts: 0

Rank	Run	Model Type	Test MAE [deg]	Test RMSE [deg]	Val MAE [deg]	Params	Duration	Artifact Size	Model Complexity	Training Heaviness	Campaign
1	te_track2h_l_causal_tcn_offset_residual_fw	latent_state_hysteresis_probe	0.003470	0.004068	0.003565	97,923	5m 24s	1.17 MB	High	Low	track2h_latent_state_hysteresis_campaign_2026_06_16

track2h_latent_state_hysteresis_gru_offset_residual_fw

• Best run: te_track2h_l_gru_offset_residual_fw

• Best test MAE: 0.003537

• Completed tracked runs: 1

• Known failed campaign attempts: 0

Rank	Run	Model Type	Test MAE [deg]	Test RMSE [deg]	Val MAE [deg]	Params	Duration	Artifact Size	Model Complexity	Training Heaviness	Campaign
1	te_track2h_l_gru_offset_residual_fw	latent_state_hysteresis_probe	0.003537	0.004110	0.003468	125,475	10m 34s	1.48 MB	High	Low	track2h_latent_state_hysteresis_campaign_2026_06_16

track2h_mixture_density_heads_mdn_k2_fw

• Best run: te_track2h_mdn_k2_fw

• Best test MAE: 0.003339

• Completed tracked runs: 1

• Known failed campaign attempts: 0

Rank	Run	Model Type	Test MAE [deg]	Test RMSE [deg]	Val MAE [deg]	Params	Duration	Artifact Size	Model Complexity	Training Heaviness	Campaign
1	te_track2h_mdn_k2_fw	curve_aware_harmonic_residual_offset_probe	0.003339	0.003721	0.003285	86,802	9m 01s	1.01 MB	High	Low	track2h_mixture_density_heads_campaign_2026_06_13

track2h_mixture_density_heads_mdn_k3_fw

• Best run: te_track2h_mdn_k3_fw

• Best test MAE: 0.003235

• Completed tracked runs: 1

• Known failed campaign attempts: 0

Rank	Run	Model Type	Test MAE [deg]	Test RMSE [deg]	Val MAE [deg]	Params	Duration	Artifact Size	Model Complexity	Training Heaviness	Campaign
1	te_track2h_mdn_k3_fw	curve_aware_harmonic_residual_offset_probe	0.003235	0.003613	0.003253	87,435	8m 40s	1.02 MB	High	Low	track2h_mixture_density_heads_campaign_2026_06_13

track2h_quantile_probabilistic_gaussian_nll_fw

• Best run: te_track2h_gaussian_nll_fw

• Best test MAE: 0.003165

• Completed tracked runs: 1

• Known failed campaign attempts: 0

Rank	Run	Model Type	Test MAE [deg]	Test RMSE [deg]	Val MAE [deg]	Params	Duration	Artifact Size	Model Complexity	Training Heaviness	Campaign
1	te_track2h_gaussian_nll_fw	curve_aware_harmonic_residual_offset_probe	0.003165	0.003548	0.003293	85,958	13m 18s	1.00 MB	High	Low	track2h_quantile_probabilistic_campaign_2026_06_12

track2h_quantile_probabilistic_quantile_p10_p50_p90_fw

• Best run: te_track2h_quantile_p10_p50_p90_fw

• Best test MAE: 0.003285

• Completed tracked runs: 1

• Known failed campaign attempts: 0

Rank	Run	Model Type	Test MAE [deg]	Test RMSE [deg]	Val MAE [deg]	Params	Duration	Artifact Size	Model Complexity	Training Heaviness	Campaign
1	te_track2h_quantile_p10_p50_p90_fw	curve_aware_harmonic_residual_offset_probe	0.003285	0.003668	0.003269	86,169	8m 53s	1.01 MB	High	Low	track2h_quantile_probabilistic_campaign_2026_06_12

tree_fw

• Best run: te_hist_gbr_tabular_Fw_grid_depth6_lr008_leaf10

• Best test MAE: 0.002743

• Completed tracked runs: 3

• Known failed campaign attempts: 0

Rank	Run	Model Type	Test MAE [deg]	Test RMSE [deg]	Val MAE [deg]	Params	Duration	Artifact Size	Model Complexity	Training Heaviness	Campaign
1	te_hist_gbr_tabular_Fw_grid_depth6_lr008_leaf10	hist_gradient_boosting	0.002743	0.003409	0.002677	5	N/A	0.45 MB	Very Low	Unknown	standalone_or_unknown
2	te_hist_gbr_tabular_Fw_grid_depth6_lr008_leaf20	hist_gradient_boosting	0.002743	0.003409	0.002677	5	N/A	0.45 MB	Very Low	Unknown	standalone_or_unknown
3	te_hist_gbr_tabular_Fw	hist_gradient_boosting	0.002845	0.003476	0.002666	5	N/A	0.50 MB	Very Low	Unknown	standalone_or_unknown

wave3_harmonic_prior_residual_pointwise_control_fw

• Best run: te_wave3_harmonic_prior_residual_pointwise_control_fw

• Best test MAE: 0.003382

• Completed tracked runs: 1

• Known failed campaign attempts: 0

Rank	Run	Model Type	Test MAE [deg]	Test RMSE [deg]	Val MAE [deg]	Params	Duration	Artifact Size	Model Complexity	Training Heaviness	Campaign
1	te_wave3_harmonic_prior_residual_pointwise_control_fw	wave3_harmonic_prior_residual	0.003382	0.003779	0.003315	7,283	7m 10s	0.11 MB	Low	Low	wave3_harmonic_prior_residual_campaign_2026_06_14

wave3_harmonic_prior_residual_smooth_l1_structured_fw

• Best run: te_wave3_harmonic_prior_residual_smooth_l1_structured_fw

• Best test MAE: 0.003527

• Completed tracked runs: 1

• Known failed campaign attempts: 0

Rank	Run	Model Type	Test MAE [deg]	Test RMSE [deg]	Val MAE [deg]	Params	Duration	Artifact Size	Model Complexity	Training Heaviness	Campaign
1	te_wave3_harmonic_prior_residual_smooth_l1_structured_fw	wave3_harmonic_prior_residual	0.003527	0.003900	0.003310	7,283	7m 28s	0.11 MB	Low	Low	wave3_harmonic_prior_residual_campaign_2026_06_14

Backward Models

feedforward_bw

• Best run: te_feedforward_stride1_high_compute_long_remote_Bw_optuna_t0005

• Best test MAE: 0.003099

• Completed tracked runs: 3

• Known failed campaign attempts: 0

Rank	Run	Model Type	Test MAE [deg]	Test RMSE [deg]	Val MAE [deg]	Params	Duration	Artifact Size	Model Complexity	Training Heaviness	Campaign
1	te_feedforward_stride1_high_compute_long_remote_Bw_optuna_t0005	feedforward	0.003099	0.003630	0.003018	167,937	N/A	1.95 MB	Very High	Unknown	standalone_or_unknown
2	te_feedforward_stride1_high_compute_long_remote_Bw_optuna_t0013	feedforward	0.003106	0.003700	0.002989	109,953	N/A	1.28 MB	High	Unknown	standalone_or_unknown
3	te_feedforward_stride1_high_compute_long_remote_Bw_optuna_t0016	feedforward	0.003173	0.003818	0.002901	167,937	N/A	1.95 MB	Very High	Unknown	standalone_or_unknown

gru_sequence_bw

• Best run: te_gru_sequence_remote_Bw

• Best test MAE: 0.003631

• Completed tracked runs: 1

• Known failed campaign attempts: 0

Rank	Run	Model Type	Test MAE [deg]	Test RMSE [deg]	Val MAE [deg]	Params	Duration	Artifact Size	Model Complexity	Training Heaviness	Campaign
1	te_gru_sequence_remote_Bw	gru_sequence	0.003631	0.004297	0.003867	151,041	6m 29s	1.74 MB	Very High	Low	wave2_temporal_model_entry_campaign_2026_05_24_11_01_15

harmonic_regression_bw

• Best run: te_harmonic_dense240_tracking_Bw

• Best test MAE: 0.003400

• Completed tracked runs: 6

• Known failed campaign attempts: 0

Rank	Run	Model Type	Test MAE [deg]	Test RMSE [deg]	Val MAE [deg]	Params	Duration	Artifact Size	Model Complexity	Training Heaviness	Campaign
1	te_harmonic_dense240_tracking_Bw	harmonic_regression	0.003400	0.003886	0.003588	2,886	5m 00s	0.04 MB	Low	Low	wave1_high_order_harmonic_tracking_campaign_2026_05_19_17_40_01
2	te_harmonic_dense360_tracking_Bw	harmonic_regression	0.003403	0.003866	0.003637	4,326	6m 43s	0.06 MB	Low	Low	wave1_high_order_harmonic_tracking_campaign_2026_05_19_17_40_01
3	te_harmonic_rcim_sparse_tracking_Bw	harmonic_regression	0.003406	0.003894	0.003570	114	5m 56s	0.01 MB	Very Low	Low	wave1_high_order_harmonic_tracking_campaign_2026_05_19_17_40_01
4	te_harmonic_order12_linear_conditioned_recovery_Bw_grid_order8_lr0002_stride5	harmonic_regression	0.003494	0.004081	0.003638	102	N/A	0.01 MB	Very Low	Unknown	standalone_or_unknown
5	te_harmonic_order12_linear_conditioned_recovery_Bw_grid_order8_lr00005_stride1	harmonic_regression	0.003497	0.004053	0.003743	102	N/A	0.01 MB	Very Low	Unknown	standalone_or_unknown
6	te_harmonic_order12_linear_conditioned_recovery_Bw_grid_order12_lr00005_stride5	harmonic_regression	0.003506	0.004063	0.003729	150	N/A	0.01 MB	Very Low	Unknown	standalone_or_unknown

lstm_sequence_bw

• Best run: te_lstm_sequence_remote_Bw

• Best test MAE: 0.003557

• Completed tracked runs: 1

• Known failed campaign attempts: 0

Rank	Run	Model Type	Test MAE [deg]	Test RMSE [deg]	Val MAE [deg]	Params	Duration	Artifact Size	Model Complexity	Training Heaviness	Campaign
1	te_lstm_sequence_remote_Bw	lstm_sequence	0.003557	0.004201	0.003815	201,345	6m 29s	2.32 MB	Very High	Low	wave2_temporal_model_entry_campaign_2026_05_24_11_01_15

periodic_gru_sequence_bw

• Best run: te_periodic_gru_sequence_remote_Bw

• Best test MAE: 0.002344

• Completed tracked runs: 1

• Known failed campaign attempts: 0

Rank	Run	Model Type	Test MAE [deg]	Test RMSE [deg]	Val MAE [deg]	Params	Duration	Artifact Size	Model Complexity	Training Heaviness	Campaign
1	te_periodic_gru_sequence_remote_Bw	periodic_gru_sequence	0.002344	0.002747	0.002523	157,953	31m 26s	1.82 MB	Very High	Medium	wave2b_harmonic_temporal_hybrid_campaign_2026_05_25

periodic_lstm_sequence_bw

• Best run: te_periodic_lstm_sequence_remote_Bw

• Best test MAE: 0.002556

• Completed tracked runs: 1

• Known failed campaign attempts: 0

Rank	Run	Model Type	Test MAE [deg]	Test RMSE [deg]	Val MAE [deg]	Params	Duration	Artifact Size	Model Complexity	Training Heaviness	Campaign
1	te_periodic_lstm_sequence_remote_Bw	periodic_lstm_sequence	0.002556	0.002953	0.002432	210,561	35m 21s	2.43 MB	Very High	Medium	wave2b_harmonic_temporal_hybrid_campaign_2026_05_25

periodic_mlp_bw

• Best run: te_periodic_mlp_h04_standard_Bw_optuna_t0006

• Best test MAE: 0.003233

• Completed tracked runs: 6

• Known failed campaign attempts: 0

Rank	Run	Model Type	Test MAE [deg]	Test RMSE [deg]	Val MAE [deg]	Params	Duration	Artifact Size	Model Complexity	Training Heaviness	Campaign
1	te_periodic_mlp_h04_standard_Bw_optuna_t0006	periodic_mlp	0.003233	0.003792	0.002907	27,777	N/A	0.34 MB	Medium	Unknown	standalone_or_unknown
2	te_periodic_mlp_h04_standard_Bw_optuna_t0007	periodic_mlp	0.003239	0.003820	0.002933	28,289	N/A	0.35 MB	Medium	Unknown	standalone_or_unknown
3	te_periodic_mlp_h04_standard_Bw_optuna_t0010	periodic_mlp	0.003248	0.003817	0.002963	27,265	N/A	0.33 MB	Medium	Unknown	standalone_or_unknown
4	te_periodic_mlp_rcim_sparse_tracking_Bw	periodic_mlp	0.003398	0.003922	0.003011	28,545	9m 57s	0.35 MB	Medium	Low	wave1_periodic_mlp_explicit_harmonic_tracking_campaign_2026_05_20_22_42_49
5	te_periodic_mlp_dense240_tracking_Bw	periodic_mlp	0.003417	0.004005	0.003041	87,681	20m 05s	1.03 MB	High	Medium	wave1_periodic_mlp_explicit_harmonic_tracking_campaign_2026_05_20_22_42_49
6	te_periodic_mlp_dense360_tracking_Bw	periodic_mlp	0.003424	0.004006	0.003072	118,401	20m 33s	1.38 MB	High	Medium	wave1_periodic_mlp_explicit_harmonic_tracking_campaign_2026_05_20_22_42_49

periodic_temporal_convolution_bw

• Best run: te_periodic_temporal_convolution_sequence_remote_Bw

• Best test MAE: 0.003614

• Completed tracked runs: 1

• Known failed campaign attempts: 0

Rank	Run	Model Type	Test MAE [deg]	Test RMSE [deg]	Val MAE [deg]	Params	Duration	Artifact Size	Model Complexity	Training Heaviness	Campaign
1	te_periodic_temporal_convolution_sequence_remote_Bw	periodic_temporal_convolution	0.003614	0.004163	0.003890	158,529	8m 25s	1.83 MB	Very High	Low	wave2b_harmonic_temporal_hybrid_campaign_2026_05_25

residual_harmonic_gru_sequence_bw_dense240

• Best run: te_residual_harmonic_gru_sequence_remote_Bw_dense240

• Best test MAE: 0.003492

• Completed tracked runs: 1

• Known failed campaign attempts: 0

Rank	Run	Model Type	Test MAE [deg]	Test RMSE [deg]	Val MAE [deg]	Params	Duration	Artifact Size	Model Complexity	Training Heaviness	Campaign
1	te_residual_harmonic_gru_sequence_remote_Bw_dense240	residual_harmonic_gru_sequence	0.003492	0.004074	0.003585	151,522	19m 40s	1.75 MB	Very High	Medium	wave2c_residual_harmonic_temporal_hybrid_campaign_2026_05_27

residual_harmonic_gru_sequence_bw_dense360

• Best run: te_residual_harmonic_gru_sequence_remote_Bw_dense360

• Best test MAE: 0.003468

• Completed tracked runs: 1

• Known failed campaign attempts: 0

Rank	Run	Model Type	Test MAE [deg]	Test RMSE [deg]	Val MAE [deg]	Params	Duration	Artifact Size	Model Complexity	Training Heaviness	Campaign
1	te_residual_harmonic_gru_sequence_remote_Bw_dense360	residual_harmonic_gru_sequence	0.003468	0.004050	0.003773	151,762	13m 22s	1.75 MB	Very High	Low	wave2c_residual_harmonic_temporal_hybrid_campaign_2026_05_27

residual_harmonic_gru_sequence_bw_sparse_rcim

• Best run: te_residual_harmonic_gru_sequence_remote_Bw_sparse_rcim

• Best test MAE: 0.003502

• Completed tracked runs: 1

• Known failed campaign attempts: 0

Rank	Run	Model Type	Test MAE [deg]	Test RMSE [deg]	Val MAE [deg]	Params	Duration	Artifact Size	Model Complexity	Training Heaviness	Campaign
1	te_residual_harmonic_gru_sequence_remote_Bw_sparse_rcim	residual_harmonic_gru_sequence	0.003502	0.004061	0.003833	151,060	6m 18s	1.75 MB	Very High	Low	wave2c_residual_harmonic_temporal_hybrid_campaign_2026_05_27

residual_harmonic_lstm_sequence_bw_dense240

• Best run: te_residual_harmonic_lstm_sequence_remote_Bw_dense240

• Best test MAE: 0.003605

• Completed tracked runs: 1

• Known failed campaign attempts: 0

Rank	Run	Model Type	Test MAE [deg]	Test RMSE [deg]	Val MAE [deg]	Params	Duration	Artifact Size	Model Complexity	Training Heaviness	Campaign
1	te_residual_harmonic_lstm_sequence_remote_Bw_dense240	residual_harmonic_lstm_sequence	0.003605	0.004129	0.003742	201,826	10m 18s	2.33 MB	Very High	Low	wave2c_residual_harmonic_temporal_hybrid_campaign_2026_05_27

residual_harmonic_lstm_sequence_bw_dense360

• Best run: te_residual_harmonic_lstm_sequence_remote_Bw_dense360

• Best test MAE: 0.003556

• Completed tracked runs: 1

• Known failed campaign attempts: 0

Rank	Run	Model Type	Test MAE [deg]	Test RMSE [deg]	Val MAE [deg]	Params	Duration	Artifact Size	Model Complexity	Training Heaviness	Campaign
1	te_residual_harmonic_lstm_sequence_remote_Bw_dense360	residual_harmonic_lstm_sequence	0.003556	0.004125	0.003729	202,066	15m 59s	2.33 MB	Very High	Medium	wave2c_residual_harmonic_temporal_hybrid_campaign_2026_05_27

residual_harmonic_lstm_sequence_bw_sparse_rcim

• Best run: te_residual_harmonic_lstm_sequence_remote_Bw_sparse_rcim

• Best test MAE: 0.003440

• Completed tracked runs: 1

• Known failed campaign attempts: 0

Rank	Run	Model Type	Test MAE [deg]	Test RMSE [deg]	Val MAE [deg]	Params	Duration	Artifact Size	Model Complexity	Training Heaviness	Campaign
1	te_residual_harmonic_lstm_sequence_remote_Bw_sparse_rcim	residual_harmonic_lstm_sequence	0.003440	0.004030	0.003764	201,364	7m 48s	2.32 MB	Very High	Low	wave2c_residual_harmonic_temporal_hybrid_campaign_2026_05_27

residual_harmonic_mlp_bw

• Best run: te_residual_harmonic_rcim_sparse_tracking_Bw

• Best test MAE: 0.003042

• Completed tracked runs: 6

• Known failed campaign attempts: 0

Rank	Run	Model Type	Test MAE [deg]	Test RMSE [deg]	Val MAE [deg]	Params	Duration	Artifact Size	Model Complexity	Training Heaviness	Campaign
1	te_residual_harmonic_rcim_sparse_tracking_Bw	residual_harmonic_mlp	0.003042	0.003548	0.002953	26,260	6m 07s	0.32 MB	Medium	Low	wave1_high_order_harmonic_tracking_campaign_2026_05_19_17_40_01
2	te_residual_harmonic_dense360_tracking_Bw	residual_harmonic_mlp	0.003068	0.003545	0.002826	26,962	14m 01s	0.33 MB	Medium	Low	wave1_high_order_harmonic_tracking_campaign_2026_05_19_17_40_01
3	te_residual_h12_deep_joint_wave1_Bw_optuna_t0007	residual_harmonic_mlp	0.003162	0.003862	0.002948	34,962	N/A	0.42 MB	Medium	Unknown	standalone_or_unknown
4	te_residual_h12_deep_joint_wave1_Bw_optuna_t0012	residual_harmonic_mlp	0.003180	0.003642	0.002979	43,026	N/A	0.52 MB	Medium	Unknown	standalone_or_unknown
5	te_residual_harmonic_dense240_tracking_Bw	residual_harmonic_mlp	0.003188	0.003717	0.002861	26,722	8m 25s	0.33 MB	Medium	Low	wave1_high_order_harmonic_tracking_campaign_2026_05_19_17_40_01
6	te_residual_h12_deep_joint_wave1_Bw_optuna_t0013	residual_harmonic_mlp	0.003195	0.003636	0.003051	43,026	N/A	0.52 MB	Medium	Unknown	standalone_or_unknown

sequential_residual_offset_probe_bw

• Best run: te_sequential_residual_offset_probe_remote_bw

• Best test MAE: 0.003638

• Completed tracked runs: 1

• Known failed campaign attempts: 0

Rank	Run	Model Type	Test MAE [deg]	Test RMSE [deg]	Val MAE [deg]	Params	Duration	Artifact Size	Model Complexity	Training Heaviness	Campaign
1	te_sequential_residual_offset_probe_remote_bw	sequential_residual_offset_probe	0.003638	0.004280	0.003840	92,802	7m 07s	1.09 MB	High	Low	track2f_offset_aware_probe_campaign_2026_06_03

temporal_convolution_bw

• Best run: te_temporal_convolution_sequence_remote_Bw

• Best test MAE: 0.003739

• Completed tracked runs: 1

• Known failed campaign attempts: 0

Rank	Run	Model Type	Test MAE [deg]	Test RMSE [deg]	Val MAE [deg]	Params	Duration	Artifact Size	Model Complexity	Training Heaviness	Campaign
1	te_temporal_convolution_sequence_remote_Bw	temporal_convolution	0.003739	0.004369	0.003933	147,009	8m 12s	1.70 MB	High	Low	wave2_temporal_model_entry_campaign_2026_05_24_11_01_15

track2f_bis_clean_sequential_residual_offset_bw

• Best run: te_track2f_bis_clean_residual_offset_bw

• Best test MAE: 0.003540

• Completed tracked runs: 1

• Known failed campaign attempts: 0

Rank	Run	Model Type	Test MAE [deg]	Test RMSE [deg]	Val MAE [deg]	Params	Duration	Artifact Size	Model Complexity	Training Heaviness	Campaign
1	te_track2f_bis_clean_residual_offset_bw	sequential_residual_offset_probe	0.003540	0.004203	0.003820	92,802	9m 37s	1.09 MB	High	Low	track2f_bis_harmonic_offset_probe_campaign_2026_06_04

track2f_bis_harmonic_residual_offset_bw

• Best run: te_track2f_bis_harmonic_residual_offset_bw

• Best test MAE: 0.003336

• Completed tracked runs: 1

• Known failed campaign attempts: 1

Rank	Run	Model Type	Test MAE [deg]	Test RMSE [deg]	Val MAE [deg]	Params	Duration	Artifact Size	Model Complexity	Training Heaviness	Campaign
1	te_track2f_bis_harmonic_residual_offset_bw	harmonic_residual_offset_probe	0.003336	0.003935	0.003555	85,747	0s	1.00 MB	High	Very Low	track2f_bis_harmonic_offset_probe_campaign_2026_06_04

Known failed campaign attempts for this family:

• te_track2f_bis_harmonic_residual_offset_bw | campaign track2f_bis_harmonic_offset_probe_campaign_2026_06_04 | model type harmonic_residual_offset_probe | error Unsupported Model Type for Campaign Runner | harmonic_residual_offset_probe | Supported: ['feedforward', 'gru_sequence', 'harmonic_regression', 'hist_gradient_boosting', 'lstm_sequence', 'periodic_gru_sequence', 'periodic_lstm_sequence', 'periodic_mlp', 'periodic_temporal_convolution', 'random_forest', 'residual_harmonic_gru_sequence', 'residual_harmonic_lstm_sequence', 'residual_harmonic_mlp', 'sequential_residual_offset_probe', 'temporal_convolution']

track2g_curve_aware_harmonic_residual_offset_full_curve_composite_bw

• Best run: te_track2g_curve_aware_full_curve_composite_bw

• Best test MAE: 0.003511

• Completed tracked runs: 1

• Known failed campaign attempts: 0

Rank	Run	Model Type	Test MAE [deg]	Test RMSE [deg]	Val MAE [deg]	Params	Duration	Artifact Size	Model Complexity	Training Heaviness	Campaign
1	te_track2g_curve_aware_full_curve_composite_bw	curve_aware_harmonic_residual_offset_probe	0.003511	0.004113	0.003803	85,747	15m 23s	1.00 MB	High	Medium	track2g_curve_aware_training_campaign_2026_06_08

track2g_curve_aware_harmonic_residual_offset_pointwise_control_bw

• Best run: te_track2g_curve_aware_pointwise_control_bw

• Best test MAE: 0.003430

• Completed tracked runs: 1

• Known failed campaign attempts: 0

Rank	Run	Model Type	Test MAE [deg]	Test RMSE [deg]	Val MAE [deg]	Params	Duration	Artifact Size	Model Complexity	Training Heaviness	Campaign
1	te_track2g_curve_aware_pointwise_control_bw	curve_aware_harmonic_residual_offset_probe	0.003430	0.003945	0.003749	85,747	14m 29s	1.00 MB	High	Low	track2g_curve_aware_training_campaign_2026_06_08

track2g_curve_aware_harmonic_residual_offset_raw_centered_shape_bw

• Best run: te_track2g_curve_aware_raw_centered_shape_bw

• Best test MAE: 0.003465

• Completed tracked runs: 1

• Known failed campaign attempts: 0

Rank	Run	Model Type	Test MAE [deg]	Test RMSE [deg]	Val MAE [deg]	Params	Duration	Artifact Size	Model Complexity	Training Heaviness	Campaign
1	te_track2g_curve_aware_raw_centered_shape_bw	curve_aware_harmonic_residual_offset_probe	0.003465	0.003998	0.003740	85,747	15m 37s	1.00 MB	High	Medium	track2g_curve_aware_training_campaign_2026_06_08

track2g_curve_aware_harmonic_residual_offset_raw_offset_bw

• Best run: te_track2g_curve_aware_raw_offset_bw

• Best test MAE: 0.003471

• Completed tracked runs: 1

• Known failed campaign attempts: 0

Rank	Run	Model Type	Test MAE [deg]	Test RMSE [deg]	Val MAE [deg]	Params	Duration	Artifact Size	Model Complexity	Training Heaviness	Campaign
1	te_track2g_curve_aware_raw_offset_bw	curve_aware_harmonic_residual_offset_probe	0.003471	0.003992	0.003751	85,747	15m 22s	1.00 MB	High	Medium	track2g_curve_aware_training_campaign_2026_06_08

track2h_dispersion_aware_log_cosh_robust_bw

• Best run: te_track2h_log_cosh_robust_bw

• Best test MAE: 0.003481

• Completed tracked runs: 1

• Known failed campaign attempts: 0

Rank	Run	Model Type	Test MAE [deg]	Test RMSE [deg]	Val MAE [deg]	Params	Duration	Artifact Size	Model Complexity	Training Heaviness	Campaign
1	te_track2h_log_cosh_robust_bw	curve_aware_harmonic_residual_offset_probe	0.003481	0.004029	0.003774	85,747	10m 56s	1.00 MB	High	Low	track2h_dispersion_aware_modeling_campaign_2026_06_10

track2h_dispersion_aware_mae_robust_bw

• Best run: te_track2h_mae_robust_bw

• Best test MAE: 0.003430

• Completed tracked runs: 1

• Known failed campaign attempts: 0

Rank	Run	Model Type	Test MAE [deg]	Test RMSE [deg]	Val MAE [deg]	Params	Duration	Artifact Size	Model Complexity	Training Heaviness	Campaign
1	te_track2h_mae_robust_bw	curve_aware_harmonic_residual_offset_probe	0.003430	0.004029	0.003579	85,747	18m 22s	1.00 MB	High	Medium	track2h_dispersion_aware_modeling_campaign_2026_06_10

track2h_dispersion_aware_smooth_l1_robust_bw

• Best run: te_track2h_smooth_l1_robust_bw

• Best test MAE: 0.003074

• Completed tracked runs: 1

• Known failed campaign attempts: 0

Rank	Run	Model Type	Test MAE [deg]	Test RMSE [deg]	Val MAE [deg]	Params	Duration	Artifact Size	Model Complexity	Training Heaviness	Campaign
1	te_track2h_smooth_l1_robust_bw	curve_aware_harmonic_residual_offset_probe	0.003074	0.003662	0.003372	85,747	28m 21s	1.00 MB	High	Medium	track2h_dispersion_aware_modeling_campaign_2026_06_10

track2h_latent_state_hysteresis_causal_tcn_offset_residual_bw

• Best run: te_track2h_l_causal_tcn_offset_residual_bw

• Best test MAE: 0.003630

• Completed tracked runs: 1

• Known failed campaign attempts: 0

Rank	Run	Model Type	Test MAE [deg]	Test RMSE [deg]	Val MAE [deg]	Params	Duration	Artifact Size	Model Complexity	Training Heaviness	Campaign
1	te_track2h_l_causal_tcn_offset_residual_bw	latent_state_hysteresis_probe	0.003630	0.004312	0.003840	97,923	11m 52s	1.17 MB	High	Low	track2h_latent_state_hysteresis_campaign_2026_06_16

track2h_latent_state_hysteresis_gru_offset_residual_bw

• Best run: te_track2h_l_gru_offset_residual_bw

• Best test MAE: 0.003545

• Completed tracked runs: 1

• Known failed campaign attempts: 0

Rank	Run	Model Type	Test MAE [deg]	Test RMSE [deg]	Val MAE [deg]	Params	Duration	Artifact Size	Model Complexity	Training Heaviness	Campaign
1	te_track2h_l_gru_offset_residual_bw	latent_state_hysteresis_probe	0.003545	0.004175	0.003837	125,475	14m 01s	1.48 MB	High	Low	track2h_latent_state_hysteresis_campaign_2026_06_16

track2h_mixture_density_heads_mdn_k2_bw

• Best run: te_track2h_mdn_k2_bw

• Best test MAE: 0.002658

• Completed tracked runs: 1

• Known failed campaign attempts: 0

Rank	Run	Model Type	Test MAE [deg]	Test RMSE [deg]	Val MAE [deg]	Params	Duration	Artifact Size	Model Complexity	Training Heaviness	Campaign
1	te_track2h_mdn_k2_bw	curve_aware_harmonic_residual_offset_probe	0.002658	0.003198	0.002914	86,802	33m 06s	1.01 MB	High	Medium	track2h_mixture_density_heads_campaign_2026_06_13

track2h_mixture_density_heads_mdn_k3_bw

• Best run: te_track2h_mdn_k3_bw

• Best test MAE: 0.002721

• Completed tracked runs: 1

• Known failed campaign attempts: 0

Rank	Run	Model Type	Test MAE [deg]	Test RMSE [deg]	Val MAE [deg]	Params	Duration	Artifact Size	Model Complexity	Training Heaviness	Campaign
1	te_track2h_mdn_k3_bw	curve_aware_harmonic_residual_offset_probe	0.002721	0.003250	0.002775	87,435	26m 49s	1.02 MB	High	Medium	track2h_mixture_density_heads_campaign_2026_06_13

track2h_quantile_probabilistic_gaussian_nll_bw

• Best run: te_track2h_gaussian_nll_bw

• Best test MAE: 0.002998

• Completed tracked runs: 1

• Known failed campaign attempts: 0

Rank	Run	Model Type	Test MAE [deg]	Test RMSE [deg]	Val MAE [deg]	Params	Duration	Artifact Size	Model Complexity	Training Heaviness	Campaign
1	te_track2h_gaussian_nll_bw	curve_aware_harmonic_residual_offset_probe	0.002998	0.003608	0.003298	85,958	27m 33s	1.00 MB	High	Medium	track2h_quantile_probabilistic_campaign_2026_06_12

track2h_quantile_probabilistic_quantile_p10_p50_p90_bw

• Best run: te_track2h_quantile_p10_p50_p90_bw

• Best test MAE: 0.002927

• Completed tracked runs: 1

• Known failed campaign attempts: 0

Rank	Run	Model Type	Test MAE [deg]	Test RMSE [deg]	Val MAE [deg]	Params	Duration	Artifact Size	Model Complexity	Training Heaviness	Campaign
1	te_track2h_quantile_p10_p50_p90_bw	curve_aware_harmonic_residual_offset_probe	0.002927	0.003519	0.003436	86,169	26m 33s	1.01 MB	High	Medium	track2h_quantile_probabilistic_campaign_2026_06_12

tree_bw

• Best run: te_hist_gbr_tabular_Bw_grid_depth6_lr008_leaf10

• Best test MAE: 0.002954

• Completed tracked runs: 3

• Known failed campaign attempts: 0

Rank	Run	Model Type	Test MAE [deg]	Test RMSE [deg]	Val MAE [deg]	Params	Duration	Artifact Size	Model Complexity	Training Heaviness	Campaign
1	te_hist_gbr_tabular_Bw_grid_depth6_lr008_leaf10	hist_gradient_boosting	0.002954	0.003749	0.002681	5	N/A	0.45 MB	Very Low	Unknown	standalone_or_unknown
2	te_hist_gbr_tabular_Bw_grid_depth6_lr008_leaf20	hist_gradient_boosting	0.002954	0.003749	0.002681	5	N/A	0.45 MB	Very Low	Unknown	standalone_or_unknown
3	te_hist_gbr_tabular_Bw_grid_depth8_lr008_leaf10	hist_gradient_boosting	0.003002	0.003809	0.002650	5	N/A	0.44 MB	Very Low	Unknown	standalone_or_unknown

wave3_harmonic_prior_residual_pointwise_control_bw

• Best run: te_wave3_harmonic_prior_residual_pointwise_control_bw

• Best test MAE: 0.003363

• Completed tracked runs: 1

• Known failed campaign attempts: 0

Rank	Run	Model Type	Test MAE [deg]	Test RMSE [deg]	Val MAE [deg]	Params	Duration	Artifact Size	Model Complexity	Training Heaviness	Campaign
1	te_wave3_harmonic_prior_residual_pointwise_control_bw	wave3_harmonic_prior_residual	0.003363	0.003902	0.003634	7,283	14m 45s	0.11 MB	Low	Low	wave3_harmonic_prior_residual_campaign_2026_06_14

wave3_harmonic_prior_residual_smooth_l1_structured_bw

• Best run: te_wave3_harmonic_prior_residual_smooth_l1_structured_bw

• Best test MAE: 0.003431

• Completed tracked runs: 1

• Known failed campaign attempts: 0

Rank	Run	Model Type	Test MAE [deg]	Test RMSE [deg]	Val MAE [deg]	Params	Duration	Artifact Size	Model Complexity	Training Heaviness	Campaign
1	te_wave3_harmonic_prior_residual_smooth_l1_structured_bw	wave3_harmonic_prior_residual	0.003431	0.003953	0.003644	7,283	13m 56s	0.11 MB	Low	Low	wave3_harmonic_prior_residual_campaign_2026_06_14

Source Of Truth

• Live backlog: doc/running/te_model_live_backlog.md

• Active campaign state: doc/running/active_training_campaign.yaml

• Program registry: output/registries/program/current_best_solution.yaml

• Family registries root: output/registries/families

• Training campaign root: output/training_campaigns

• Training run root: output/training_runs

• Paper reference report: doc/reports/analysis/RCIM Paper Reference Benchmark.md

This document is repository-generated. Regenerate it after new campaign results so the cross-family snapshot stays aligned with the canonical registries and campaign artifacts.