·
DataBubble
  • Home
  • Models
  • News
  • Compare
  • Boards
  • Pricing
  • About
  • Newsletter
  • Methodology
  • Contact
Latest
PT$^3$: Prototype-Guided Self-Supervised Pretraining for Generalizable mmWave Human Point Cloud Representation7h◆Right Numbers, Wrong Questions: Evaluating LLMs on Underspecified Financial Queries7h◆"Close enough?" An investigation into multiplication errors in LLMs7h◆Weaken to Strengthen: Weakening-Induced On-Policy Self-Distillation for Large Reasoning Models7h◆SocialArena: Evaluating Social Intelligence in Large Language Models7h◆ATLAS-Bench: An Agentic Tool-Use Benchmark for Local Area Search7h◆Does Scaling Toward AGI Depend on What System-1 Lets System-2 Execute?7h◆Integrated Feature Selection for classification and Representation Learning with Group-Lasso Siamese Networks7h◆PAPERBENCHX: Can AI Reproduce Science Across Disciplines?7h◆Finding Is Not Returning, Returning Is Not Reconstructing: A Multi-Dimensional Evaluation of Agent Social Re-entry7h◆Long-Term Agent Memory through Complementary Curation and Consolidation7h◆Revisiting Over-refusal in Safety-aligned Large Language Models: A Representation and Optimization Perspective7h◆EvoTeam: Multi-Agent Collaboration via Evolving Teams7h◆AegisFlow: Training-Free Non-Myopic Path-Safe Guided Flow Matching7h◆KARMA: Training-Free Knowledge-Atomic Harness Evolution for Industrial Advertising Agents7h◆Efficient Learning of Mechanism Designs and Compositionality7h◆Distributional Reward Calibration and Budget-Neutral Credit Assignment for Multimodal RLVR7h◆Neighborhood Elites: Quality-Diversity Search for Uncertainty Quantification in Question Answering7h◆Multistep and Multivariate Time Series Explainability with GradientSHAP: A Wind Power Case Study7h◆Beyond Selected Breakthroughs: Evaluating Language Models on Randomly Sampled Mathematical Conjectures7h◆PT$^3$: Prototype-Guided Self-Supervised Pretraining for Generalizable mmWave Human Point Cloud Representation7h◆Right Numbers, Wrong Questions: Evaluating LLMs on Underspecified Financial Queries7h◆"Close enough?" An investigation into multiplication errors in LLMs7h◆Weaken to Strengthen: Weakening-Induced On-Policy Self-Distillation for Large Reasoning Models7h◆SocialArena: Evaluating Social Intelligence in Large Language Models7h◆ATLAS-Bench: An Agentic Tool-Use Benchmark for Local Area Search7h◆Does Scaling Toward AGI Depend on What System-1 Lets System-2 Execute?7h◆Integrated Feature Selection for classification and Representation Learning with Group-Lasso Siamese Networks7h◆PAPERBENCHX: Can AI Reproduce Science Across Disciplines?7h◆Finding Is Not Returning, Returning Is Not Reconstructing: A Multi-Dimensional Evaluation of Agent Social Re-entry7h◆Long-Term Agent Memory through Complementary Curation and Consolidation7h◆Revisiting Over-refusal in Safety-aligned Large Language Models: A Representation and Optimization Perspective7h◆EvoTeam: Multi-Agent Collaboration via Evolving Teams7h◆AegisFlow: Training-Free Non-Myopic Path-Safe Guided Flow Matching7h◆KARMA: Training-Free Knowledge-Atomic Harness Evolution for Industrial Advertising Agents7h◆Efficient Learning of Mechanism Designs and Compositionality7h◆Distributional Reward Calibration and Budget-Neutral Credit Assignment for Multimodal RLVR7h◆Neighborhood Elites: Quality-Diversity Search for Uncertainty Quantification in Question Answering7h◆Multistep and Multivariate Time Series Explainability with GradientSHAP: A Wind Power Case Study7h◆Beyond Selected Breakthroughs: Evaluating Language Models on Randomly Sampled Mathematical Conjectures7h◆
News/Revisiting Over-refusal in Safety-aligned Large Language Models: A Representation and Optimization Perspective
arxiv
PublishedOctober 31, 2026 at 4:00 AM

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

Source
arxiv.orgfull article ↗
Read on arxiv→
Publisher summary· verbatim

arXiv:2610.39802v1 Announce Type: new Abstract: 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 ma

Stay posted· Newsletter

Three short emails a week: Monday, Wednesday and Friday. Top movers, new models and the most-referenced story.

// no spam · unsubscribe one-click · free forever

Your email is used only to send The Bubble Brief. Privacy policy

Discussion
Source
↗
arxiv
Read original ↗All from arxiv →

No replies yet. Be first.

Source
↗
arxiv
Read original ↗All from arxiv →

Related coverage

More from ARXIV
arxivPT$^3$: Prototype-Guided Self-Supervised Pretraining for Generalizable mmWave Human Point Cloud Representation7harxivRight Numbers, Wrong Questions: Evaluating LLMs on Underspecified Financial Queries7harxiv"Close enough?" An investigation into multiplication errors in LLMs7harxivWeaken to Strengthen: Weakening-Induced On-Policy Self-Distillation for Large Reasoning Models7h
The Bubble Brief
MON · WED · FRI

Read AI insights every Monday, Wednesday and Friday — top movers, new models, story of the week.

// no spam · unsubscribe one-click · free forever

Your email is used only to send The Bubble Brief. Privacy policy

Originally published on arxiv ↗
PrivacyTermsRefundsLegal notice|AboutMethodologyContact

© 2026 Databubble

Built by Marouane Gazouzi · Readyo, AI interview coach

HomeModelsNews