2-1B-Instruct with 3-bit AWQ using WikiText-2 (WT2) as calibration data.
Lower reconstruction loss can degrade quantized LLM performance; robust refinement fixes it
DRQ refines integer codes to minimize worst-case loss over activation distribution shifts, improving six PTQ methods without inference overhead.
Top University
Yanlong Zhao · Xiaoyuan Cheng · Huihang Liu · Baihua He · Xinyu Zhang · Harrison Bo Hua Zhu · +3 more
University of Science and Technology of China · University College London · Shanghai University of Finance and Economics · AMSS, Chinese Academy of Sciences · University of Copenhagen
Research Digest··2 min read
The authors show that minimizing reconstruction loss on calibration data during weight-only post-training quantization (PTQ) can hurt model performance, even on the very same data.
Why this paper
From AMSS, Chinese Academy of Sciences and 7 others
In one line
Distributionally robust quantization (DRQ) minimizes worst-case reconstruction loss under constrained activation shifts, avoiding the calibration-loss trap.
What we could check
- ·No code link found
- ·No weights link found
- ·No dataset link found
- ·No compute details found
- ·No stated limitations found
- ·No benchmark numbers found
Observed from the paper text and links we have. Absence here means we did not find it, not that it does not exist.
§