Bounded Commuting Cochain Layer / bench_report.json

Mechanism confirmed, baseline not beaten

Raw ⬇ ZIP
  1{
  2  "bench_version": 1,
  3  "track": "simplicial_cochain_regression",
  4  "model": "cochain_mlp_tiny",
  5  "metric_direction": "lower is better",
  6  "n_seeds": 8,
  7  "baseline": {
  8    "best_cfg": {
  9      "lr": 0.001
 10    },
 11    "sweep": [
 12      {
 13        "cfg": {
 14          "lr": 0.001
 15        },
 16        "mean": 0.1318089496344328
 17      },
 18      {
 19        "cfg": {
 20          "lr": 0.003
 21        },
 22        "mean": 0.13333775848150253
 23      },
 24      {
 25        "cfg": {
 26          "lr": 0.01
 27        },
 28        "mean": 0.13918141461908817
 29      }
 30    ],
 31    "full": {
 32      "mean": 0.1284407116472721,
 33      "std": 0.01033345480907241,
 34      "per_seed": [
 35        0.13917690515518188,
 36        0.13619644939899445,
 37        0.11503438651561737,
 38        0.14294329285621643,
 39        0.12111610919237137,
 40        0.11670704931020737,
 41        0.12120074033737183,
 42        0.1351507604122162
 43      ],
 44      "n": 8
 45    }
 46  },
 47  "idea": {
 48    "mean": 0.12762724328786135,
 49    "std": 0.011713296514767995,
 50    "per_seed": [
 51      0.13896797597408295,
 52      0.13771608471870422,
 53      0.10817594081163406,
 54      0.14403699338436127,
 55      0.12192557752132416,
 56      0.11830909550189972,
 57      0.11906668543815613,
 58      0.13281959295272827
 59    ],
 60    "n": 8
 61  },
 62  "comparison": {
 63    "delta_mean": -0.0008134683594107628,
 64    "idea_wins": 4,
 65    "n_pairs": 8,
 66    "per_seed_diffs": [
 67      -0.000208929181098938,
 68      0.0015196353197097778,
 69      -0.006858445703983307,
 70      0.0010937005281448364,
 71      0.0008094683289527893,
 72      0.0016020461916923523,
 73      -0.0021340548992156982,
 74      -0.002331167459487915
 75    ],
 76    "p_value": 0.5406,
 77    "mde": 0.002417293917190717,
 78    "mde_rel_pct": 1.8820309278799117,
 79    "verdict": "no measurable effect",
 80    "system_worked": false
 81  },
 82  "math_check": {
 83    "incidence_d1_d0_max": 0.0,
 84    "compatible_idempotence": 4.0475970476012593e-16,
 85    "compatible_bound": 0.9999999999999999,
 86    "compatible_commutation_d0": 8.402395041642446e-16,
 87    "compatible_commutation_d1": 2.0014830212433607e-16,
 88    "independent_commutation_d0": 1.632993161855452,
 89    "independent_commutation_d1": 1.6329931618554514
 90  },
 91  "baseline_full_grid": [
 92    {
 93      "cfg": {
 94        "lr": 0.001
 95      },
 96      "result": {
 97        "mean": 0.12913066055625677,
 98        "std": 0.008989460805190601,
 99        "per_seed": [
100          0.13755616545677185,
101          0.13973239064216614,
102          0.11453189700841904,
103          0.13541534543037415,
104          0.12468931823968887,
105          0.11946979910135269,
106          0.12405481934547424,
107          0.1375955492258072
108        ],
109        "n": 8
110      }
111    },
112    {
113      "cfg": {
114        "lr": 0.003
115      },
116      "result": {
117        "mean": 0.1284407116472721,
118        "std": 0.01033345480907241,
119        "per_seed": [
120          0.13917690515518188,
121          0.13619644939899445,
122          0.11503438651561737,
123          0.14294329285621643,
124          0.12111610919237137,
125          0.11670704931020737,
126          0.12120074033737183,
127          0.1351507604122162
128        ],
129        "n": 8
130      }
131    },
132    {
133      "cfg": {
134        "lr": 0.01
135      },
136      "result": {
137        "mean": 0.1359273912385106,
138        "std": 0.011280773802478765,
139        "per_seed": [
140          0.14981389045715332,
141          0.1443662792444229,
142          0.11666289716959,
143          0.14588259160518646,
144          0.13578224182128906,
145          0.12386979162693024,
146          0.12721310555934906,
147          0.14382833242416382
148        ],
149        "n": 8
150      }
151    }
152  ],
153  "idea_sweep": [
154    {
155      "cfg": {
156        "lr": 0.001
157      },
158      "result": {
159        "mean": 0.1288067139685154,
160        "std": 0.010748948365935283,
161        "per_seed": [
162          0.13977062702178955,
163          0.14079467952251434,
164          0.11066274344921112,
165          0.13858631253242493,
166          0.12385569512844086,
167          0.11842944473028183,
168          0.12199791520833969,
169          0.13635629415512085
170        ],
171        "n": 8
172      }
173    },
174    {
175      "cfg": {
176        "lr": 0.003
177      },
178      "result": {
179        "mean": 0.12762724328786135,
180        "std": 0.011713296514767995,
181        "per_seed": [
182          0.13896797597408295,
183          0.13771608471870422,
184          0.10817594081163406,
185          0.14403699338436127,
186          0.12192557752132416,
187          0.11830909550189972,
188          0.11906668543815613,
189          0.13281959295272827
190        ],
191        "n": 8
192      }
193    },
194    {
195      "cfg": {
196        "lr": 0.01
197      },
198      "result": {
199        "mean": 0.13056561537086964,
200        "std": 0.012256510065250442,
201        "per_seed": [
202          0.14063149690628052,
203          0.14152514934539795,
204          0.10811122506856918,
205          0.14493317902088165,
206          0.13158296048641205,
207          0.12201656401157379,
208          0.1183496043086052,
209          0.13737474381923676
210        ],
211        "n": 8
212      }
213    }
214  ],
215  "idea_config": {
216    "lr": 0.003
217  },
218  "custom_track": {
219    "name": "simplicial_cochain_regression",
220    "file": "cochain_track.py",
221    "domain": "simplicial_geometry_pde_like"
222  },
223  "mechanism_signature": {
224    "prediction": "trained projected edge representations have lower face-boundary residual",
225    "predicted_direction": "idea < baseline",
226    "baseline_boundary_rms": 0.6453002542257309,
227    "idea_boundary_rms": 3.951021199100069e-08,
228    "reduction_fraction": 0.9999999387723595,
229    "confirmed": true
230  }
231}