{
  "dataset": "seantrott/cs_norms",
  "revision": "80ac8fc7008fc0b9b4fcaa80ffd7b006c2bba550",
  "source_file": "contextualized_sensorimotor_norms.csv",
  "source_sha256": "9e3dd772529c2376352704c246018c18f3cacc9ce26c2724fddf761d83f84735",
  "license": "cc-by-4.0",
  "raw_rows": 448,
  "unique_words": 112,
  "aggregation": "Human mean strength 0..5 divided by 5 as soft BCE targets; evaluation binarizes human mean >=2.5. Classification boundary is a project-defined proxy, not an original dataset class.",
  "task": "Prototype for sensory associations of a target word in context, covering five external senses and interoception. Not an established whole-sentence sensation-presence benchmark.",
  "input_format": "Target: WORD\nContext: SENTENCE",
  "labels": [
    "vision",
    "hearing",
    "smell",
    "taste",
    "touch",
    "interoception"
  ],
  "seed": 20261004,
  "split_method": "80/10/10 shuffled groups: normalized comment for morals; target word for sensations; deterministic rejection sampling until every label has at least 3 positives in validation and test, before any model training",
  "sizes": {
    "train": 356,
    "validation": 44,
    "test": 48
  },
  "synthetic_examples": 177,
  "synthetic_independent_seeds": 59,
  "synthetic_limitations": "Assistant labels, not human-reviewed; correlated framing variants; one-hot synthetic targets simplify naturally overlapping categories.",
  "positive_supports": {
    "train": {
      "vision": 273,
      "hearing": 74,
      "smell": 28,
      "taste": 24,
      "touch": 65,
      "interoception": 23
    },
    "validation": {
      "vision": 36,
      "hearing": 6,
      "smell": 4,
      "taste": 5,
      "touch": 8,
      "interoception": 8
    },
    "test": {
      "vision": 38,
      "hearing": 7,
      "smell": 6,
      "taste": 6,
      "touch": 13,
      "interoception": 4
    }
  },
  "files_sha256": {
    "synthetic.jsonl": "21516a46b68b9cbb4d8b8edca395b0078915ca5a951377704f1eedd2e39d247b",
    "test.jsonl": "2f49703f5ceeb241e7b9e7a2fef03b7c5088fd30a9e854019d975dd720e9e883",
    "train.jsonl": "636ca7ffb67b87f3ad7b82b3b760d194061ad0797404d49bacef11021f05e693",
    "validation.jsonl": "f45f5f06c462905cb54c61ffebcaa4f752d251df7c5817479eba20302fb72409"
  },
  "synthetic_word_overlap_removed": 3,
  "synthetic_excluded_heldout_words": [
    "cold"
  ],
  "synthetic_word_audit": "Explicit target words reserved for validation/test are excluded from synthetic training, before fitting.",
  "additional_preparation_source_hash": "3899b2211d7cd6a56641813e1a65fd39a75b0332122122a6d9aab670c42d4338"
}