{ "bench_version": 1, "track": "geometric_graph_diffusion", "model": "GeometricMessageNet", "metric_direction": "lower is better", "n_seeds": 8, "baseline": { "best_cfg": { "lr": 0.003, "alpha": 0.5 }, "sweep": [ { "cfg": { "lr": 0.001, "alpha": 0.5 }, "mean": 0.08639695774763823 }, { "cfg": { "lr": 0.003, "alpha": 0.5 }, "mean": 0.06887020450085402 }, { "cfg": { "lr": 0.01, "alpha": 0.5 }, "mean": 0.08051598258316517 } ], "full": { "mean": 0.08486439986154437, "std": 0.029581161321461158, "per_seed": [ 0.036324214190244675, 0.07466080784797668, 0.09419873356819153, 0.07029706239700317, 0.14098234474658966, 0.07441659271717072, 0.0738333985209465, 0.11420204490423203 ], "n": 8 } }, "idea": { "mean": 0.07120458083227277, "std": 0.023789005397093055, "per_seed": [ 0.048530153930187225, 0.06982574611902237, 0.07111480087041855, 0.05362492427229881, 0.1285717934370041, 0.08061950653791428, 0.05659816414117813, 0.06075155735015869 ], "n": 8 }, "comparison": { "delta_mean": -0.013659819029271603, "idea_wins": 6, "n_pairs": 8, "per_seed_diffs": [ 0.01220593973994255, -0.004835061728954315, -0.02308393269777298, -0.01667213812470436, -0.012410551309585571, 0.006202913820743561, -0.01723523437976837, -0.053450487554073334 ], "p_value": 0.0868, "mde": 0.016833120021636148, "mde_rel_pct": 19.835313805434616, "verdict": "no significant win", "system_worked": false }, "mechanism_signature": { "track_structure": "geometric graph node-field diffusion regression", "trained_model_measurements": { "baseline": { "test_mse": 0.036324214190244675, "predicted_l2": 0.22030678391456604, "observed_gradient_l2": 0.26773717999458313, "predicted_observed_corr": 0.8789346780854339 }, "idea": { "test_mse": 0.039621297270059586, "predicted_l2": 0.2222180962562561, "observed_gradient_l2": 0.24487970769405365, "predicted_observed_corr": 0.38409918113832164 }, "prediction": "distance weighting preserves more long-range influence", "confirmed": false }, "idea_sweep": [ { "cfg": { "lr": 0.001, "alpha": 0.5 }, "result": { "mean": 0.07120458083227277, "std": 0.023789005397093055, "per_seed": [ 0.048530153930187225, 0.06982574611902237, 0.07111480087041855, 0.05362492427229881, 0.1285717934370041, 0.08061950653791428, 0.05659816414117813, 0.06075155735015869 ], "n": 8 } }, { "cfg": { "lr": 0.003, "alpha": 0.5 }, "result": { "mean": 0.07151469262316823, "std": 0.05479380310064357, "per_seed": [ 0.039621297270059586, 0.06278985738754272, 0.053295787423849106, 0.06647685915231705, 0.21443338692188263, 0.05103813484311104, 0.044252004474401474, 0.040210213512182236 ], "n": 8 } }, { "cfg": { "lr": 0.01, "alpha": 0.5 }, "result": { "mean": 0.07402465306222439, "std": 0.07012354046889444, "per_seed": [ 0.03252619504928589, 0.0537540428340435, 0.06584807485342026, 0.050242867320775986, 0.2579071521759033, 0.04949888214468956, 0.042814504355192184, 0.039605505764484406 ], "n": 8 } } ], "confirmed": false }, "custom_track": { "name": "geometric_graph_diffusion", "file": "graph_track.py", "domain": "graph-neural-networks" }, "protocol_notes": { "paired_seeds": [ 0, 1, 2, 3, 4, 5, 6, 7 ], "baseline_and_idea_same_architecture": true, "only_intervention": "edge weight d^2 versus unit edge weight", "device": "cuda" } }