Стенд-проверка (stage-2) · промт оператора:

(универсальный)

Ответ агента:

{ "worked": true, "confidence": 8, "verdict": "On the registered poisson_dirichlet PDE track, the learned positive monotone radial warp improved test MSE over the identically sized raw-coordinate MLP. The best idea setting (lr=0.003, 25 epochs) achieved mean MSE 0.0002235653 versus the tuned baseline mean 0.0002440518 across 8 paired seeds, with delta_mean=-2.04866e-05 and permutation p=0.01475; this satisfies the harness criterion for a significant win. The trained-model signature also confirmed finite trained behaviour and lower spatial roughness for the warped system, although the signature is qualitative rather than a direct PDE exponent fit.", "metrics": { "baseline": "poisson_dirichlet, mlp_med, tuned lr=0.003 and 25 epochs; 8-seed mean MSE 0.00024405181284237187, std 1.798621374535186e-05.", "idea": "poisson_dirichlet, same field architecture plus learned positive normalized monotone warp with q=2; lr=0.003 and 25 epochs; 8-seed mean MSE 0.00022356524641509168, std 1.1563454829705651e-05.", "paired_delta": "-2.0486566427280195e-05; idea wins 7/8 paired seeds; permutation p=0.01475." }, "how_to_run": "/home/maxwelhelp/main/bin/python3 bench_warp_registered.py", "files": [ "bench_warp_registered.py", "bench_report.json" ], "limitations": "The registered poisson_dirichlet track is a 1-D forcing-to-solution Dirichlet regression benchmark rather than the proposed full 2-D L-shaped Deep Ritz/PINN with explicit transformed derivatives, Jacobian-weighted energy, and unrolled outer optimization. The warp is applied to the predicted spatial output grid rather than implemented as a full polar PDE coordinate transform. Only three learning rates and 25 training epochs were tested; no larger models, longer budgets, fixed s^2 comparator, or coefficient/boundary-condition shifts were evaluated.", "bench_report": { "bench_version": 1, "track": "poisson_dirichlet", "model": "mlp_med", "metric_direction": "lower is better", "n_seeds": 8, "baseline": { "best_cfg": { "lr": 0.003, "epochs": 25 }, "sweep": [ { "cfg": { "lr": 0.001, "epochs": 25 }, "mean": 0.000594377372181043 }, { "cfg": { "lr": 0.002, "epochs": 25 }, "mean": 0.0003386351017979905 }, { "cfg": { "lr": 0.003, "epochs": 25 }, "mean": 0.00025070174524444155 } ], "full": { "mean": 0.00024405181284237187, "std": 1.798621374535186e-05, "per_seed": [ 0.00022197705402504653, 0.0002670164976734668, 0.0002739553165156394, 0.00023985811276361346, 0.00021926779299974442, 0.0002489262551534921, 0.00023994325601961464, 0.00024147021758835763 ], "n": 8 } }, "idea": { "mean": 0.00022356524641509168, "std": 1.1563454829705651e-05, "per_seed": [ 0.00020604842575266957, 0.00024056619440671057, 0.00022221256222110242, 0.00021581824694294482, 0.00022475924924947321, 0.00024225283414125443, 0.0002205111231887713, 0.0002163533354178071 ], "n": 8 }, "comparison": { "delta_mean": -2.0486566427280195e-05, "idea_wins": 7, "n_pairs": 8, "per_seed_diffs": [ -1.5928628272376955e-05, -2.6450303266756237e-05, -5.174275429453701e-05, -2.4039865820668638e-05, 5.491456249728799e-06, -6.673421012237668e-06, -1.9432132830843336e-05, -2.5116882170550525e-05 ], "p_value": 0.01475, "mde": 1.3898879546086865e-05, "mde_rel_pct": 5.695052777609921, "verdict": "idea better (significant)", "system_worked": true }, "mechanism_signature": { "prediction": "learned monotone warp should remain non-folding while redistributing trained field resolution", "predicted_q": 2.0, "confirmed": true, "custom_track": { "name": "poisson_dirichlet", "file": "registered bench/custom_tracks/poisson_dirichlet.py", "domain": "pde" }, "idea_sweep": [ { "cfg": { "lr": 0.001, "epochs": 25 }, "mean": 0.0004552114551188424, "std": 2.025566535985663e-05 }, { "cfg": { "lr": 0.002, "epochs": 25 }, "mean": 0.00028170261066406965, "std": 1.3204764213516373e-05 }, { "cfg": { "lr": 0.003, "epochs": 25 }, "mean": 0.00022356524641509168, "std": 1.1563454829705651e-05 } ], "baseline_trained_behaviour": [ { "trained_boundary_abs_mean": 0.008487517014145851, "trained_spatial_roughness": 0.014843669719994068 }, { "trained_boundary_abs_mean": 0.008709563873708248, "trained_spatial_roughness": 0.01600482687354088 }, { "trained_boundary_abs_mean": 0.010176915675401688, "trained_spatial_roughness": 0.016539592295885086 }, { "trained_boundary_abs_mean": 0.007808547466993332, "trained_spatial_roughness": 0.015132329426705837 }, { "trained_boundary_abs_mean": 0.006689203903079033, "trained_spatial_roughness": 0.01441388949751854 }, { "trained_boundary_abs_mean": 0.010704973712563515, "trained_spatial_roughness": 0.015490293502807617 }, { "trained_boundary_abs_mean": 0.00736059807240963, "trained_spatial_roughness": 0.014637845568358898 }, { "trained_boundary_abs_mean": 0.008381623774766922, "trained_spatial_roughness": 0.015110178850591183 } ], "idea_trained_behaviour": [ { "trained_boundary_abs_mean": 0.008834951557219028, "trained_spatial_roughness": 0.013183629140257835 }, { "trained_boundary_abs_mean": 0.009743912145495415, "trained_spatial_roughness": 0.013243976049125195 }, { "trained_boundary_abs_mean": 0.009452861733734608, "trained_spatial_roughness": 0.013008011505007744 }, { "trained_boundary_abs_mean": 0.009347931481897831, "trained_spatial_roughness": 0.01307782344520092 }, { "trained_boundary_abs_mean": 0.007658765651285648, "trained_spatial_roughness": 0.012795394286513329 }, { "trained_boundary_abs_mean": 0.010028958320617676, "trained_spatial_roughness": 0.012233818881213665 }, { "trained_boundary_abs_mean": 0.007676774635910988, "trained_spatial_roughness": 0.012434136122465134 }, { "trained_boundary_abs_mean": 0.008619775995612144, "trained_spatial_roughness": 0.012422935105860233 } ] } }, "system_verdict": "worked", "practical_verdict": "helps", "mechanism_ok": 1, "system_judged": true }