{
  "schema_version": 2,
  "manifest_ref": "multimodal-models:video-checkpoint-benchmark",
  "benchmark": "Video-MME-v2",
  "benchmark_revision": "6e4bebb03202e1ddbf3d37703e560e51c5aa2d64",
  "evaluation_tier": "l2_public_benchmark_subset",
  "seeds": [
    42,
    43,
    44
  ],
  "evaluated_examples": 4,
  "prediction_source": "public_checkpoint",
  "metadata": {
    "dataset_source": "https://huggingface.co/datasets/MME-Benchmarks/Video-MME-v2",
    "annotations_sha256": "cde9651e53833a77b5d8830ac4cdbd251caf72c92c0ac7ad133ae9c85074ef0a",
    "subset_protocol": "all four linked questions for official video 001",
    "model_id": "HuggingFaceTB/SmolVLM2-256M-Video-Instruct",
    "model_revision": "067788b187b95ebe7b2e040b3e4299e342e5b8fd",
    "checkpoint_committed": false,
    "num_frames": 32,
    "frame_sampling": "uniform with source fps and original timestamps",
    "decoding": "deterministic, option letter only",
    "resume_replay_verified": true,
    "runtime": {
      "platform": "Linux x86_64",
      "resolved_device": "cuda",
      "accelerator": "NVIDIA A30"
    }
  },
  "seed_results": [
    {
      "seed": 42,
      "accuracy": 0.0,
      "parse_rate": 1.0
    },
    {
      "seed": 43,
      "accuracy": 0.0,
      "parse_rate": 1.0
    },
    {
      "seed": 44,
      "accuracy": 0.0,
      "parse_rate": 1.0
    }
  ],
  "aggregate_metrics": {
    "accuracy_mean": 0.0,
    "accuracy_std": 0.0,
    "accuracy_ci95_radius": 0.0,
    "parse_rate_mean": 1.0
  },
  "formal_comparison": true,
  "claim_policy": "four-question deterministic checkpoint smoke on one official video; not the full Video-MME-v2 accuracy or grouped nonlinear leaderboard score",
  "evaluation_protocol": {
    "tier": "l2_public_benchmark_subset",
    "seeds": [
      42,
      43,
      44
    ],
    "formal_comparison": true,
    "claim_policy": "four-question deterministic checkpoint smoke on one official video; not the full Video-MME-v2 accuracy or grouped nonlinear leaderboard score"
  },
  "provenance": {
    "created_at": "2026-08-20",
    "artifact_path": "docs/multimodal-models/metrics/video-mme-v2-smolvlm2-a30-4x3.json",
    "dataset_fingerprint": "sha256:cde9651e53833a77b5d8830ac4cdbd251caf72c92c0ac7ad133ae9c85074ef0a",
    "config_digest": "a06969d28fdc670ee7bfd52a8d43f9e79cf6d143a92fdcd456ada5a481cd5b13",
    "checkpoint_path": "local snapshot (not committed)"
  }
}
