{
  "schema_version": 2,
  "dataset": "fashion-mnist-qa",
  "evaluation_tier": "l1_public_images",
  "source": "https://github.com/zalandoresearch/fashion-mnist",
  "train_examples": 2000,
  "validation_examples": 1000,
  "test_examples": 1000,
  "seed": 42,
  "steps_per_trial": 150,
  "baseline": {
    "architecture": "micro_vlm_linear",
    "validation_accuracy": 0.424,
    "test_accuracy": 0.421
  },
  "champion": {
    "architecture": "micro_vlm_query",
    "validation_accuracy": 0.501,
    "validation_shuffled_image_accuracy": 0.109,
    "validation_blank_image_accuracy": 0.1,
    "validation_visual_dependency_delta": 0.392,
    "test_accuracy": 0.485,
    "test_shuffled_image_accuracy": 0.107,
    "test_blank_image_accuracy": 0.1
  },
  "scope": "Small public-image object-QA experiment; not open-ended VQA or a paper-scale VLM reproduction.",
  "manifest_ref": "multimodal-models:fashion-mnist-qa-seed42",
  "evaluation_protocol": {
    "tier": "l1_mechanism",
    "seeds": [
      42
    ],
    "formal_comparison": false,
    "claim_policy": "single/few-seed smoke result; do not claim a stable improvement"
  },
  "provenance": {
    "artifact_path": "docs/multimodal-models/metrics/fashion-mnist-qa-seed42.json",
    "historical_migration": "historical-metrics-v2-2026-08-09",
    "original_code_commit": "not recorded",
    "dataset_fingerprint": "not recorded in historical artifact"
  }
}