There is a newer version of the record available.

Published October 27, 2026 | Version v1.0

Quantization Does Not Undo Abliteration: A Measurement Study of Refusal in Compressed Language Models (artifact)

  • 1. University of Nevada, Las Vegas
  • 2. West Career and Technical Academy

Description

Artifact for the ARTMAN 2026 (ACSAC workshop) paper. Evaluation prompts, judge verdicts for all 46,600 generations (response text replaced by SHA-256), human annotation labels, checkpoint-differencing and rank-sweep measurements, and the code that produces every table and figure in the paper. Raw model responses are distributed as a separate restricted file on this record.

Files

yxk14/quant-abliteration-v1.0.zip

Files (11.6 MB)

Name Size
md5:de5947b61dbe49f6fd1a90afb304f280
11.6 MB Preview Download

Additional details

Related works