Upload README.md with huggingface_hub
Browse files
README.md
ADDED
|
@@ -0,0 +1,33 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
---
|
| 2 |
+
license: cc-by-4.0
|
| 3 |
+
language: [en]
|
| 4 |
+
tags: [psychometrics, quantization, IRT, DIF, LLM-evaluation]
|
| 5 |
+
---
|
| 6 |
+
|
| 7 |
+
# quant-dif-raw v1
|
| 8 |
+
|
| 9 |
+
Raw response data for **"Quantization as a Measurement Invariance Problem: A
|
| 10 |
+
Differential Item Functioning Analysis of Psychometric Scales in Compressed
|
| 11 |
+
LLMs"**.
|
| 12 |
+
|
| 13 |
+
- 189,000 logged administrations of the 50-item IPIP Big-Five Factor Markers
|
| 14 |
+
to 7 frozen model configurations (Qwen2.5-3B fp16/q8_0/q4_0, Qwen2.5-7B
|
| 15 |
+
q8_0/q4_0, Llama-3.2-3B q8_0/q4_0), served by Ollama v0.31.1.
|
| 16 |
+
- Conditions: main experiment (200 personas × 50 items), no-persona
|
| 17 |
+
distribution sub-experiment (40 samples/item), paraphrase-bank rerun,
|
| 18 |
+
temperature 0.2/1.0 subsets.
|
| 19 |
+
- Every JSONL record carries full provenance: model tag, precision label,
|
| 20 |
+
persona/item ids, per-call seed, temperature, GGUF weight SHA-256, engine
|
| 21 |
+
version, timestamp, raw response text, parsed 1–5 score, and top-5 token
|
| 22 |
+
log-probabilities.
|
| 23 |
+
- Bundle also contains the human-verified paraphrase item bank, the 200-persona
|
| 24 |
+
pool (deterministic, seed 42), and the model manifest.
|
| 25 |
+
|
| 26 |
+
`quant-dif-raw-v1.tar.gz` SHA-256:
|
| 27 |
+
`5b03248483fb60871a802331614e11af9b4ecb2c44321e1c9e1ccf580f36ef96`
|
| 28 |
+
|
| 29 |
+
Analysis code and preregistration: https://github.com/XuebinMa/quant-dif
|
| 30 |
+
(preregistration frozen at commit `a1a8582` before any data collection).
|
| 31 |
+
|
| 32 |
+
Note: IPIP item texts (public domain, ipip.ori.org) appear inside prompt/record
|
| 33 |
+
fields as administered.
|