devils-agent / reports /model-comparison.json
devildasdf's picture
Upload experimental BAIM code, research checkpoints and measured evaluations
795f737 verified
Raw
History Blame Contribute Delete
1.92 kB
{
"candidates": [
{
"label": "v000-mean-fp32",
"diagnostic_score": 0.7630678147473269,
"report": "reports/bench-v000-mean-fp32.json",
"test_joint": 1.0,
"novel_joint": 0.5583333373069763,
"p95_policy_ms": 2.8466000221669674,
"rss_bytes": 275996672,
"disk_bytes": 516312
},
{
"label": "v001-no-lexical-fp32",
"diagnostic_score": 0.7028540519792041,
"report": "reports/bench-v001-no-lexical-fp32.json",
"test_joint": 0.9979166388511658,
"novel_joint": 0.5062500238418579,
"p95_policy_ms": 2.7871999191120267,
"rss_bytes": 276901888,
"disk_bytes": 516312
},
{
"label": "v000-mean-int8",
"diagnostic_score": 0.4852571466875086,
"report": "reports/bench-v000-mean-int8.json",
"test_joint": 1.0,
"novel_joint": 0.5562499761581421,
"p95_policy_ms": 4.169900086708367,
"rss_bytes": 295170048,
"disk_bytes": 459286
},
{
"label": "v003-transformer-fp32",
"diagnostic_score": 0.22693369776324787,
"report": "reports/bench-v003-transformer-fp32.json",
"test_joint": 1.0,
"novel_joint": 0.4270833432674408,
"p95_policy_ms": 5.411299993284047,
"rss_bytes": 373432320,
"disk_bytes": 657592
},
{
"label": "v002-gru-fp32",
"diagnostic_score": 0.18824286961703973,
"report": "reports/bench-v002-gru-fp32.json",
"test_joint": 1.0,
"novel_joint": 0.3583333194255829,
"p95_policy_ms": 6.873500067740679,
"rss_bytes": 297365504,
"disk_bytes": 616488
}
],
"production_promoted": false,
"formula": "test_joint * novel_wording_joint / (p95_policy_ms * observed_python_rss_GiB)",
"reason": "Diagnostic synthetic ranking only; target-hardware, full task, recovery and security evidence missing."
}