quant-dif-raw / README.md
StevenShield's picture
Upload README.md with huggingface_hub
a52b535 verified
|
Raw
History Blame Contribute Delete
1.42 kB
metadata
license: cc-by-4.0
language:
  - en
tags:
  - psychometrics
  - quantization
  - IRT
  - DIF
  - LLM-evaluation

quant-dif-raw v1

Raw response data for "Quantization as a Measurement Invariance Problem: A Differential Item Functioning Analysis of Psychometric Scales in Compressed LLMs".

  • 189,000 logged administrations of the 50-item IPIP Big-Five Factor Markers to 7 frozen model configurations (Qwen2.5-3B fp16/q8_0/q4_0, Qwen2.5-7B q8_0/q4_0, Llama-3.2-3B q8_0/q4_0), served by Ollama v0.31.1.
  • Conditions: main experiment (200 personas × 50 items), no-persona distribution sub-experiment (40 samples/item), paraphrase-bank rerun, temperature 0.2/1.0 subsets.
  • Every JSONL record carries full provenance: model tag, precision label, persona/item ids, per-call seed, temperature, GGUF weight SHA-256, engine version, timestamp, raw response text, parsed 1–5 score, and top-5 token log-probabilities.
  • Bundle also contains the human-verified paraphrase item bank, the 200-persona pool (deterministic, seed 42), and the model manifest.

quant-dif-raw-v1.tar.gz SHA-256: 5b03248483fb60871a802331614e11af9b4ecb2c44321e1c9e1ccf580f36ef96

Analysis code and preregistration: https://github.com/XuebinMa/quant-dif (preregistration frozen at commit a1a8582 before any data collection).

Note: IPIP item texts (public domain, ipip.ori.org) appear inside prompt/record fields as administered.