Gradient Orthogonalization

Latest papers

11 papers3 papers in the last four weeks, against 1 the four weeks before. 0.0% of all new papers.

Revisiting Over-refusal in Safety-aligned Large Language Models: A Representation and Optimization Perspective
Oct 29, 2026Chao Chen, Pengyang Shao, Wei Qin +3Gradient Orthogonalization Over-Refusal
Hefei University of Technology Hefei, China · National University of Singapore Singapore, Singapore
MuonGS: Pretraining LLMs with Relaxed Weight-Norm Control
Oct 27, 2026Athanasios Glentis, Dawei Li, Mingyi HongGradient Orthogonalization Muon Optimizer
University of Minnesota
Polar Switch: Fast Muon Orthogonalization with Polynomial Circuits
Oct 19, 2026Zhehang Du, Weijie SuGradient Orthogonalization Orthogonal Optimization
University of Pennsylvania
Muon Under Gradient Noise and the Limits of Orthogonalization Near Optima
Sep 26, 2026Xiaohui XieMuon Optimizer Gradient Orthogonalization
Department of Computer Science University of California, Irvine
Spectral Gradient Orthogonalization Improves Differentially Private Training at Scale
Aug 18, 2026Sabari Shanmugam, Nick Barnes, Kerry TaylorPrivacy-Preserving ML Differential Privacy
School of Computing, Australian National University, Canberra, Australia
FreqOrtho-SR: Frequency-Guided Orthogonal Expert Learning for Real-World Image Super-Resolution
Jun 27, 2026Minh Son Hoang, Dinh Phu Tran, Quyen Nguyen Duc +2Diffusion Models for Image Restoration Gradient Orthogonalization
School of Computing, KAIST, Republic of Korea
FOGO: Forgetting-aware Orthogonalization Optimizer
Jun 9, 2026Toan Nguyen, Yang Liu, Trung Le +2Deep Learning Optimization Gradient Orthogonalization
School of Computer Science and Engineering University of New South Wales Sydney, Australia · Department of Data Science & AI Monash University Melbourne, Australia · DEVCOM Army Research Laboratory USA
Octopus: History-Free Gradient Orthogonalization for Continual Learning in Multimodal Large Language Models
May 14, 2026Yuehao Liu, Shanyan Guan, Weijia Zhang +4Continual Learning for LLMs Gradient Orthogonalization
MoE Key Lab of Artificial Intelligence, AI Institute, Shanghai Jiao Tong University · vivo Mobile Communication Co., Ltd.
Retain-Neutral Surrogates for Min-Max Unlearning
May 7, 2026Junhao Cai, Dohun Kim, Dowon Kim +4Gradient Orthogonalization Machine Unlearning
Korea University
Subspace Control: Turning Constrained Model Steering into Controllable Spectral Optimization
Apr 5, 2026Yancheng Huang, Changsheng Wang, Chongyu Fan +7Gradient Orthogonalization Spectral Optimization
OPTML, Michigan State University · MIT-IBM Watson AI Lab, IBM Research · University of Minnesota +2
UNO: Unlearning via Orthogonalization in Generative models
Jun 5, 2025Pinak Mandal, Georg A. GottwaldGradient Orthogonalization Machine Unlearning
University of Sydney, Australia

Oct 29, 2026J/K move · Enter open · S save

Revisiting Over-refusal in Safety-aligned Large Language Models: A Representation and Optimization Perspective

Chao Chen, Pengyang Shao, Wei Qin, Shijie Hao+2

Hefei University of Technology Hefei, China · National University of Singapore Singapore, Singapore

Gradient Orthogonalization Over-Refusal Refusal Behavior in Language Models Safety-Preserving Fine-Tuning of Language Models

Open paper AskPDF arXiv

Over-refusal is a critical side effect of safety alignment in large language models (LLMs), causing harmless but superficially risky queries to be mistakenly rejected. Despite the effectiveness of existing mitigation methods, why safety-aligned LLMs may still reject harmless queries under correct safety supervision remains insufficiently understood. To address this gap, we analyze over-refusal through the lens of representation and optimization. We make two observations: 1) the representations of harmful and over-refusal samples are visually distinguishable at certain intermediate layers but increasingly overlap in deeper layers; and 2) harmful-sample gradients exhibit high intra-class similarity and concentrate in a low-rank subspace. Moreover, they are more strongly aligned with refusal-target gradients than with correct helpful-target gradients on over-refusal prompts. This imbalanced gradient coupling can bias over-refusal samples toward refusal, while the reduced apparent representational separation in deeper layers may make it more difficult for the model to maintain a clear boundary between harmful and over-refusal samples. To this end, we propose GOAL (Gradient Orthogonalization with contrAstive Learning), a safety-alignment framework for mitigating over-refusal. Specifically, GOAL consists of two components: 1) to prevent imbalanced gradient coupling from biasing the optimization of over-refusal samples toward refusal, we orthogonalize over-refusal gradients against the low-rank refusal subspace extracted from harmful gradients; and 2) to explicitly enhance the separation between harmful and over-refusal representations, we introduce a contrastive learning objective. Extensive experiments across multiple models and benchmarks demonstrate that GOAL effectively mitigates over-refusal while preserving safety performance and general capabilities.