Files
zachlandesandGitHub 462aaf404e feat: add v2 benchmark and registered reproduction sample (#1)
* Kept the original raw answers while making later runs reproducible.
* Registered a variance-sized sample before the independent rerun.
* Removed article drafts and private paths from the current tree.
2026-10-14 15:36:06 -07:00

586 lines
9.9 KiB
JSON

{
"model_id": "clef-flash",
"prompts": 24,
"answers": 72,
"mean_abs_diff": 0.0026,
"max_abs_diff": 0.0147,
"choice_changes": 1,
"rows": [
{
"id": "switch|c1|o1|0",
"q": "pull",
"fp32": 0.0251,
"bf16": 0.0252,
"abs_diff": 0.0001,
"same_side": true
},
{
"id": "switch|c1|o1|0",
"q": "pull_criteria",
"fp32": 0.0278,
"bf16": 0.0274,
"abs_diff": 0.0004,
"same_side": true
},
{
"id": "switch|c1|o1|0",
"q": "stay",
"fp32": 0.94,
"bf16": 0.9417,
"abs_diff": 0.0017,
"same_side": true
},
{
"id": "switch|c1|o1|1",
"q": "pull",
"fp32": 0.3093,
"bf16": 0.3074,
"abs_diff": 0.0019,
"same_side": true
},
{
"id": "switch|c1|o1|1",
"q": "pull_criteria",
"fp32": 0.4201,
"bf16": 0.4244,
"abs_diff": 0.0043,
"same_side": true
},
{
"id": "switch|c1|o1|1",
"q": "stay",
"fp32": 0.1869,
"bf16": 0.1848,
"abs_diff": 0.0021,
"same_side": true
},
{
"id": "switch|c1|o1|5",
"q": "pull",
"fp32": 0.8941,
"bf16": 0.8911,
"abs_diff": 0.003,
"same_side": true
},
{
"id": "switch|c1|o1|5",
"q": "pull_criteria",
"fp32": 0.9157,
"bf16": 0.9137,
"abs_diff": 0.002,
"same_side": true
},
{
"id": "switch|c1|o1|5",
"q": "stay",
"fp32": 0.0429,
"bf16": 0.0442,
"abs_diff": 0.0013,
"same_side": true
},
{
"id": "switch|c1|o1|20",
"q": "pull",
"fp32": 0.9225,
"bf16": 0.9208,
"abs_diff": 0.0017,
"same_side": true
},
{
"id": "switch|c1|o1|20",
"q": "pull_criteria",
"fp32": 0.9506,
"bf16": 0.9517,
"abs_diff": 0.0011,
"same_side": true
},
{
"id": "switch|c1|o1|20",
"q": "stay",
"fp32": 0.0337,
"bf16": 0.0336,
"abs_diff": 0.0001,
"same_side": true
},
{
"id": "switch|c1|o2|0",
"q": "pull",
"fp32": 0.0281,
"bf16": 0.0284,
"abs_diff": 0.0003,
"same_side": true
},
{
"id": "switch|c1|o2|0",
"q": "pull_criteria",
"fp32": 0.0343,
"bf16": 0.0341,
"abs_diff": 0.0002,
"same_side": true
},
{
"id": "switch|c1|o2|0",
"q": "stay",
"fp32": 0.9312,
"bf16": 0.9307,
"abs_diff": 0.0005,
"same_side": true
},
{
"id": "switch|c1|o2|1",
"q": "pull",
"fp32": 0.3004,
"bf16": 0.2862,
"abs_diff": 0.0142,
"same_side": true
},
{
"id": "switch|c1|o2|1",
"q": "pull_criteria",
"fp32": 0.2872,
"bf16": 0.2802,
"abs_diff": 0.007,
"same_side": true
},
{
"id": "switch|c1|o2|1",
"q": "stay",
"fp32": 0.1434,
"bf16": 0.1456,
"abs_diff": 0.0022,
"same_side": true
},
{
"id": "switch|c1|o2|5",
"q": "pull",
"fp32": 0.9072,
"bf16": 0.9068,
"abs_diff": 0.0004,
"same_side": true
},
{
"id": "switch|c1|o2|5",
"q": "pull_criteria",
"fp32": 0.936,
"bf16": 0.9361,
"abs_diff": 0.0001,
"same_side": true
},
{
"id": "switch|c1|o2|5",
"q": "stay",
"fp32": 0.0436,
"bf16": 0.043,
"abs_diff": 0.0006,
"same_side": true
},
{
"id": "switch|c1|o2|20",
"q": "pull",
"fp32": 0.9229,
"bf16": 0.9208,
"abs_diff": 0.0021,
"same_side": true
},
{
"id": "switch|c1|o2|20",
"q": "pull_criteria",
"fp32": 0.965,
"bf16": 0.9645,
"abs_diff": 0.0005,
"same_side": true
},
{
"id": "switch|c1|o2|20",
"q": "stay",
"fp32": 0.0351,
"bf16": 0.0351,
"abs_diff": 0.0,
"same_side": true
},
{
"id": "switch|w2|o1|0",
"q": "pull",
"fp32": 0.0335,
"bf16": 0.0328,
"abs_diff": 0.0007,
"same_side": true
},
{
"id": "switch|w2|o1|0",
"q": "pull_criteria",
"fp32": 0.0797,
"bf16": 0.0775,
"abs_diff": 0.0022,
"same_side": true
},
{
"id": "switch|w2|o1|0",
"q": "stay",
"fp32": 0.9083,
"bf16": 0.9099,
"abs_diff": 0.0016,
"same_side": true
},
{
"id": "switch|w2|o1|1",
"q": "pull",
"fp32": 0.1146,
"bf16": 0.1128,
"abs_diff": 0.0018,
"same_side": true
},
{
"id": "switch|w2|o1|1",
"q": "pull_criteria",
"fp32": 0.1938,
"bf16": 0.1925,
"abs_diff": 0.0013,
"same_side": true
},
{
"id": "switch|w2|o1|1",
"q": "stay",
"fp32": 0.2636,
"bf16": 0.2576,
"abs_diff": 0.006,
"same_side": true
},
{
"id": "switch|w2|o1|5",
"q": "pull",
"fp32": 0.9324,
"bf16": 0.9329,
"abs_diff": 0.0005,
"same_side": true
},
{
"id": "switch|w2|o1|5",
"q": "pull_criteria",
"fp32": 0.9447,
"bf16": 0.9442,
"abs_diff": 0.0005,
"same_side": true
},
{
"id": "switch|w2|o1|5",
"q": "stay",
"fp32": 0.0327,
"bf16": 0.0331,
"abs_diff": 0.0004,
"same_side": true
},
{
"id": "switch|w2|o1|20",
"q": "pull",
"fp32": 0.9429,
"bf16": 0.9395,
"abs_diff": 0.0034,
"same_side": true
},
{
"id": "switch|w2|o1|20",
"q": "pull_criteria",
"fp32": 0.9524,
"bf16": 0.9482,
"abs_diff": 0.0042,
"same_side": true
},
{
"id": "switch|w2|o1|20",
"q": "stay",
"fp32": 0.0293,
"bf16": 0.031,
"abs_diff": 0.0017,
"same_side": true
},
{
"id": "switch|w2|o2|0",
"q": "pull",
"fp32": 0.0335,
"bf16": 0.0328,
"abs_diff": 0.0007,
"same_side": true
},
{
"id": "switch|w2|o2|0",
"q": "pull_criteria",
"fp32": 0.0797,
"bf16": 0.0775,
"abs_diff": 0.0022,
"same_side": true
},
{
"id": "switch|w2|o2|0",
"q": "stay",
"fp32": 0.9083,
"bf16": 0.9099,
"abs_diff": 0.0016,
"same_side": true
},
{
"id": "switch|w2|o2|1",
"q": "pull",
"fp32": 0.1146,
"bf16": 0.1128,
"abs_diff": 0.0018,
"same_side": true
},
{
"id": "switch|w2|o2|1",
"q": "pull_criteria",
"fp32": 0.1938,
"bf16": 0.1925,
"abs_diff": 0.0013,
"same_side": true
},
{
"id": "switch|w2|o2|1",
"q": "stay",
"fp32": 0.2636,
"bf16": 0.2576,
"abs_diff": 0.006,
"same_side": true
},
{
"id": "switch|w2|o2|5",
"q": "pull",
"fp32": 0.9324,
"bf16": 0.9329,
"abs_diff": 0.0005,
"same_side": true
},
{
"id": "switch|w2|o2|5",
"q": "pull_criteria",
"fp32": 0.9447,
"bf16": 0.9442,
"abs_diff": 0.0005,
"same_side": true
},
{
"id": "switch|w2|o2|5",
"q": "stay",
"fp32": 0.0327,
"bf16": 0.0331,
"abs_diff": 0.0004,
"same_side": true
},
{
"id": "switch|w2|o2|20",
"q": "pull",
"fp32": 0.9429,
"bf16": 0.9395,
"abs_diff": 0.0034,
"same_side": true
},
{
"id": "switch|w2|o2|20",
"q": "pull_criteria",
"fp32": 0.9524,
"bf16": 0.9482,
"abs_diff": 0.0042,
"same_side": true
},
{
"id": "switch|w2|o2|20",
"q": "stay",
"fp32": 0.0293,
"bf16": 0.031,
"abs_diff": 0.0017,
"same_side": true
},
{
"id": "footbridge|v1|o1|0",
"q": "pull",
"fp32": 0.6567,
"bf16": 0.6619,
"abs_diff": 0.0052,
"same_side": true
},
{
"id": "footbridge|v1|o1|0",
"q": "pull_criteria",
"fp32": 0.4976,
"bf16": 0.5068,
"abs_diff": 0.0092,
"same_side": false
},
{
"id": "footbridge|v1|o1|0",
"q": "stay",
"fp32": 0.4629,
"bf16": 0.4625,
"abs_diff": 0.0004,
"same_side": true
},
{
"id": "footbridge|v1|o1|1",
"q": "pull",
"fp32": 0.7707,
"bf16": 0.7712,
"abs_diff": 0.0005,
"same_side": true
},
{
"id": "footbridge|v1|o1|1",
"q": "pull_criteria",
"fp32": 0.6358,
"bf16": 0.6352,
"abs_diff": 0.0006,
"same_side": true
},
{
"id": "footbridge|v1|o1|1",
"q": "stay",
"fp32": 0.294,
"bf16": 0.2991,
"abs_diff": 0.0051,
"same_side": true
},
{
"id": "footbridge|v1|o1|5",
"q": "pull",
"fp32": 0.7895,
"bf16": 0.7912,
"abs_diff": 0.0017,
"same_side": true
},
{
"id": "footbridge|v1|o1|5",
"q": "pull_criteria",
"fp32": 0.7089,
"bf16": 0.7106,
"abs_diff": 0.0017,
"same_side": true
},
{
"id": "footbridge|v1|o1|5",
"q": "stay",
"fp32": 0.2596,
"bf16": 0.2568,
"abs_diff": 0.0028,
"same_side": true
},
{
"id": "footbridge|v1|o1|20",
"q": "pull",
"fp32": 0.8124,
"bf16": 0.8081,
"abs_diff": 0.0043,
"same_side": true
},
{
"id": "footbridge|v1|o1|20",
"q": "pull_criteria",
"fp32": 0.7438,
"bf16": 0.7402,
"abs_diff": 0.0036,
"same_side": true
},
{
"id": "footbridge|v1|o1|20",
"q": "stay",
"fp32": 0.2107,
"bf16": 0.2108,
"abs_diff": 0.0001,
"same_side": true
},
{
"id": "footbridge|v1|o2|0",
"q": "pull",
"fp32": 0.5188,
"bf16": 0.5234,
"abs_diff": 0.0046,
"same_side": true
},
{
"id": "footbridge|v1|o2|0",
"q": "pull_criteria",
"fp32": 0.4164,
"bf16": 0.4311,
"abs_diff": 0.0147,
"same_side": true
},
{
"id": "footbridge|v1|o2|0",
"q": "stay",
"fp32": 0.4472,
"bf16": 0.4552,
"abs_diff": 0.008,
"same_side": true
},
{
"id": "footbridge|v1|o2|1",
"q": "pull",
"fp32": 0.8443,
"bf16": 0.8429,
"abs_diff": 0.0014,
"same_side": true
},
{
"id": "footbridge|v1|o2|1",
"q": "pull_criteria",
"fp32": 0.7425,
"bf16": 0.7476,
"abs_diff": 0.0051,
"same_side": true
},
{
"id": "footbridge|v1|o2|1",
"q": "stay",
"fp32": 0.2698,
"bf16": 0.2606,
"abs_diff": 0.0092,
"same_side": true
},
{
"id": "footbridge|v1|o2|5",
"q": "pull",
"fp32": 0.8802,
"bf16": 0.8804,
"abs_diff": 0.0002,
"same_side": true
},
{
"id": "footbridge|v1|o2|5",
"q": "pull_criteria",
"fp32": 0.795,
"bf16": 0.7925,
"abs_diff": 0.0025,
"same_side": true
},
{
"id": "footbridge|v1|o2|5",
"q": "stay",
"fp32": 0.2097,
"bf16": 0.2053,
"abs_diff": 0.0044,
"same_side": true
},
{
"id": "footbridge|v1|o2|20",
"q": "pull",
"fp32": 0.8787,
"bf16": 0.8791,
"abs_diff": 0.0004,
"same_side": true
},
{
"id": "footbridge|v1|o2|20",
"q": "pull_criteria",
"fp32": 0.8147,
"bf16": 0.8158,
"abs_diff": 0.0011,
"same_side": true
},
{
"id": "footbridge|v1|o2|20",
"q": "stay",
"fp32": 0.189,
"bf16": 0.1895,
"abs_diff": 0.0005,
"same_side": true
}
]
}