{ "bench_version": 1, "track": "dynamics", "model": "explicit_tanh_rnn_shared", "metric_direction": "lower is better", "n_seeds": 8, "baseline": { "best_cfg": { "lr": 0.01, "weight_decay": 0.0 }, "sweep": [ { "cfg": { "lr": 0.001, "weight_decay": 0.0 }, "mean": 0.32027048990130424 }, { "cfg": { "lr": 0.003, "weight_decay": 0.0 }, "mean": 0.01070007259841077 }, { "cfg": { "lr": 0.01, "weight_decay": 0.0 }, "mean": 0.0029869703284930438 } ], "full": { "mean": 0.0031130630668485537, "std": 0.0015298698474074236, "per_seed": [ 0.002583631779998541, 0.001341367606073618, 0.0019145020050927997, 0.006108379922807217, 0.005108166951686144, 0.002453721361234784, 0.0029577165842056274, 0.002437018323689699 ], "n": 8 } }, "idea": { "mean": 0.004186775186099112, "std": 0.0021045838143091996, "per_seed": [ 0.0009666543919593096, 0.005662341136485338, 0.0049017248675227165, 0.0013687764294445515, 0.007184116635471582, 0.0038662750739604235, 0.003245178610086441, 0.006299134343862534 ], "n": 8 }, "comparison": { "delta_mean": 0.0010737121192505583, "idea_wins": 2, "n_pairs": 8, "per_seed_diffs": [ -0.0016169773880392313, 0.00432097353041172, 0.002987222862429917, -0.004739603493362665, 0.0020759496837854385, 0.0014125537127256393, 0.0002874620258808136, 0.0038621160201728344 ], "p_value": 0.3437, "mde": 0.002543128344355023, "mde_rel_pct": 81.69215623792383, "verdict": "no significant win", "system_worked": false }, "mechanism_signature": { "prediction": "row-saturating matching gives full-rank trained controllability and improves long-horizon input access", "trained_model_measurements": [ { "seed": 0, "test_mse": 0.002090232679620385, "controllability_rank": 16, "min_singular": 6.442173137209217e-16, "max_singular": 1.0159012079238892 }, { "seed": 1, "test_mse": 0.08538427203893661, "controllability_rank": 17, "min_singular": 4.1185198621923365e-18, "max_singular": 0.7894802689552307 }, { "seed": 2, "test_mse": 0.04124434292316437, "controllability_rank": 17, "min_singular": 4.013088390668608e-20, "max_singular": 1.0363144874572754 }, { "seed": 3, "test_mse": 0.020918531343340874, "controllability_rank": 15, "min_singular": 2.736597840696797e-13, "max_singular": 0.9130778908729553 }, { "seed": 4, "test_mse": 0.1084025576710701, "controllability_rank": 15, "min_singular": 2.54601735032937e-15, "max_singular": 0.9427530765533447 }, { "seed": 5, "test_mse": 0.07313309609889984, "controllability_rank": 18, "min_singular": 1.5868783161841502e-24, "max_singular": 0.8598010540008545 }, { "seed": 6, "test_mse": 0.05488329380750656, "controllability_rank": 17, "min_singular": 6.466245724631725e-13, "max_singular": 0.9993555545806885 }, { "seed": 7, "test_mse": 0.07927680015563965, "controllability_rank": 16, "min_singular": 2.0967682614041888e-13, "max_singular": 0.8383661508560181 } ], "idea_grid": [ { "cfg": { "lr": 0.001, "weight_decay": 0.0 }, "full": { "mean": 0.4885829910635948, "std": 0.06858866720808456, "per_seed": [ 0.33161619305610657, 0.5656974911689758, 0.48498693108558655, 0.4976513087749481, 0.5585150122642517, 0.52698814868927, 0.47414085268974304, 0.4690679907798767 ], "n": 8 } }, { "cfg": { "lr": 0.003, "weight_decay": 0.0 }, "full": { "mean": 0.0581666408397723, "std": 0.033096846800440026, "per_seed": [ 0.002090232679620385, 0.08538427203893661, 0.04124434292316437, 0.020918531343340874, 0.1084025576710701, 0.07313309609889984, 0.05488329380750656, 0.07927680015563965 ], "n": 8 } }, { "cfg": { "lr": 0.01, "weight_decay": 0.0 }, "full": { "mean": 0.004186775186099112, "std": 0.0021045838143091996, "per_seed": [ 0.0009666543919593096, 0.005662341136485338, 0.0049017248675227165, 0.0013687764294445515, 0.007184116635471582, 0.0038662750739604235, 0.003245178610086441, 0.006299134343862534 ], "n": 8 } } ], "confirmed": false } }