AI Paper Atlas

Research map / Architectures and efficiency / Quantization

FAIR-Calib: Frontier-Aware Instability-Reweighted Calibration for Post-Training Quantization of Diffusion Large Language Models

Haoyu Huang, Linlin Yang, Sheng Xu, Boyu Liu, Guodong Guo, Zhongqian Fu, Hang Zhou, Baochang Zhang

ICML 2026 · Architectures and efficiency › Quantization · approach: Post-training quantization of LLMs · 0 citations (OpenAlex)

Read the paperCheck your idea against 137,217 papers

AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.

Measured effects reported in the paper

What was changedWhereMeasuredEffect
add frontier-hit weighting to uniform PTQ (ablation)FAIR-Calib on Dream-Base · W4A4 averageavg accuracy (%): 61.76 -> 63.12Improved
add masked-stage reliability weighting to uniform PTQ (ablation)FAIR-Calib on Dream-Base · W4A4 averageavg accuracy (%): 61.76 -> 62.89Improved
combine both signals (ablation)FAIR-Calib on Dream-Base · W4A4 averageavg accuracy (%): 63.12 -> 64.64Improved
raise probing budget from 128 to 512 (ablation)FAIR-Calib on Dream-Base · W4A4 averageavg accuracy (%): 63.15 -> 64.64Improved
raise probing budget from 512 to 1024 (ablation)FAIR-Calib on Dream-Base · W4A4 averageavg accuracy (%): 64.64 -> 64.63No clear effect
use early-boost time weighting instead of late-boost (ablation)FAIR-Calib on Dream-Base · W4A4 averageavg accuracy (%): 62.58 -> 64.64Improved
use early-boost time weighting instead of uniform (ablation)FAIR-Calib on Dream-Base · W4A4 averageavg accuracy (%): 63.21 -> 64.64Improved
add lambda0 time weighting (ablation)FAIR-Calib on Dream-Base · W4A4 averageavg accuracy (%): 62.89 -> 64.64Improved
use FAIR-Calib instead of FlatQuantDream-Base · W4A4 averageavg accuracy (%): 62.08 -> 64.64Improved
use FAIR-Calib instead of FlatQuantLLaDA-Base · W4A4 averageavg accuracy (%): 59.37 -> 61.09Improved
use FAIR-Calib instead of FlatQuantLLaDA-Instruct · W3A4 averageavg accuracy (%): 69.58 -> 70.41Improved
add position-weighted MSE to QDrop-style calibrationLLaDA-Instruct · W8A8 averageavg accuracy (%): 70.66 -> 73.17Improved

Earlier methods this paper compares with

Named baselines from earlier accepted papers in this paper's results tables: better than 2 of 2 on most shared results.

Matched by method name in results tables read by AI. Authors' own tables tend to favour the newer method, and a comparison can be missed when a method is renamed.

Closest papers by meaning

Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.

Open scime draft

Get the next scime Dispatch and a monthly field update

Free. We send a confirmation link first; every email has a one-click unsubscribe. Privacy