{
  "required": true,
  "paths": [
    {
      "id": "gs1_kernel_advantage_loop",
      "baseline_revision": "f79fdb2ba4cd967ef2807ba43097c52985f9f17f",
      "reference_revision": "git:b3003c0183695d564ca4f7696426ac40012b235e;sha256:757d81c606f9793f259c6afd5ffe6f5e52382328560145b682d84d814da9a50e",
      "implementation_revision": "sha256:fca7ac94282999a136228f2bd6fce7944a20c4ac21959bbb8774fb93aa824202",
      "fixed_input": {
        "sha256": "ce347fb22b721e8a8097bee69d1da2f7980fb7c7d79e8efc53b00882b8cca345",
        "description": "fixed-batch.json: seed-0 first 8 GSM8K train prompts x8, sampled init rank32; trainer-forward.json: one forward on never-updated init client; no optimizer state or updates"
      },
      "precision": "numpy float64 analytic loss/full dL/dlogprobs, trainer/sampler returned logprobs, stock advantages float32",
      "normalization": "negative token sum of exp(trainer_logprob-sampler_logprob)*advantage, no token/batch divisor; prompt token advantages zero; no packing; full per-token logprob gradient",
      "errors": {
        "baseline_loss": {
          "max_abs_error": 0.0,
          "reference_max_abs": 783.4791951064552,
          "relative_l2_error": 0.0,
          "cosine_similarity": 1.0
        },
        "baseline_gradient": {
          "max_abs_error": 0.0,
          "reference_max_abs": 1.0809257631265885,
          "relative_l2_error": 0.0,
          "cosine_similarity": 1.0
        },
        "reference_loss": {
          "max_abs_error": 0.0,
          "reference_max_abs": 901.8301944397053,
          "relative_l2_error": 0.0,
          "cosine_similarity": 1.0
        },
        "reference_gradient": {
          "max_abs_error": 0.0,
          "reference_max_abs": 1.2315145668947074,
          "relative_l2_error": 0.0,
          "cosine_similarity": 1.0
        }
      }
    }
  ]
}