paper-with-me

홈 › Papers

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

2026-06-04 · Haoyu Huang, Linlin Yang, Sheng Xu, Boyu Liu, Guodong Guo, Zhongqian Fu, Hang Zhou, Baochang Zhang arxiv

Diffusion Large Language Models (dLLMs) refine tokens iteratively but commit them irreversibly, leading to a "stability lag" where early decisions remain fragile even after being written. We reveal that Post-Training Quantization (PTQ) error easily flips these borderline decisions at the write frontier, which are then permanently locked in and amplified. To address this, we propose Frontier-Aware Instability-Reweighted Calibration (FAIR-Calib), a two-stage PTQ framework for dLLMs. Stage I probes a full-precision teacher to estimate a position prior that combines frontier hits and masked-stage reliability. Stage II performs off-policy, layer-wise calibration by minimizing a reweighted hidden-state MSE, effectively prioritizing the protection of fragile frontier states without requiring expensive end-to-end diffusion rollouts. We further theoretically justify our weighted objective as a surrogate for output KL divergence. Empirically, FAIR-Calib consistently outperforms state-of-the-art baselines on LLaDA and Dream (W4A4), significantly reducing frontier decision flips and suppressing post-commit mismatches across diverse benchmarks.

📄 PDF Abstract BibTeX arXiv:2606.06547

Code (0)

등록된 구현이 없습니다.

Similar Papers 제목 키워드 기반

Fairness Aware Reward Optimization

2026-02-08 · Ching Lam Choi, Vighnesh Subramaniam, Phillip Isola, Antonio Torralba 외 arxiv

Demographic skews in human preference data propagate systematic unfairness through reward models into aligned LLMs. We introduce Fairness Aware Reward Optimization (Faro), an in-processing framework that trains reward mo…

Fair and Calibrated Toxicity Detection with Robust Training and Abstention

2026-05-13 · Mokshit Surana arxiv

Fairness in toxicity classification involves three integrated axes: ranking, calibration, and abstention. Training-time interventions and post-hoc safety mechanisms cannot be evaluated independently because the former de…

Dual-Stage Reweighted MoE for Long-Tailed Egocentric Mistake Detection

2025-09-16 · Boyu Han, Qianqian Xu, Shilong Bao, Zhiyong Yang 외 arxiv

In this report, we address the problem of determining whether a user performs an action incorrectly from egocentric video data. To handle the challenges posed by subtle and infrequent mistakes, we propose a Dual-Stage Re…

F-WANDA: Fisher-Reweighted Post-Training Pruning for Sustainable Deployment of Large Language Models

2026-08-01 · Himanshu Mishra arxiv

One-shot post-training pruning is the most energy-frugal compression strategy for largelanguage models (LLMs), yet existing approaches trade either quality (WANDA) or compute cost (SPARSEGPT). We introduce F-WANDA, a dro…

Fairness without Demographics through Adversarially Reweighted Learning

2020-06-23 · NeurIPS 2020 12 · Preethi Lahoti, Alex Beutel, Jilin Chen, Kang Lee 외

Much of the previous machine learning (ML) fairness literature assumes that protected features such as race and sex are present in the dataset, and relies upon them to mitigate fairness concerns. However, in practice fac…

Fairness