Refusal Localizes, the Damage Relocates: Safety Layers Under Few-Sample Fine-Tuning Paper • 2610.00320 • Published Sep 29
Where Activation Sparsity and KV-Cache Sparsity Cross in LLM Decoding Paper • 2609.33889 • Published Sep 27
Answer-Conditioned Chains of Thought Degrade Verifiable-Reasoning Distillation in Large Language Models Paper • 2607.14552 • Published Jul 16
Beyond Penalizing Mistakes: Stabilizing Efficiency Training in Large Reasoning Models via Adaptive Correct-Only Rewards Paper • 2606.22716 • Published Jun 21
MixLID: Token-Level Language Identification in Mixed-Language Text Paper • 2610.16601 • Published Oct 12 • 6
RenderRank: Learning to Rerank Text with Compressed Visual Tokens Paper • 2609.35069 • Published Sep 28 • 7
Does On-Policy Distillation Really Distill? From Noisy Teacher to Self-Improvement Paper • 2608.31046 • Published Aug 31 • 86
Language Chain in Alignment: Cross-lingual Ranking Preference Optimization Paper • 2608.23149 • Published Aug 24 • 8
Language Chain in Alignment: Cross-lingual Ranking Preference Optimization Paper • 2608.23149 • Published Aug 24 • 8
Language Chain in Alignment: Cross-lingual Ranking Preference Optimization Paper • 2608.23149 • Published Aug 24 • 8
LAMAR: An Open Language-Aware Multilingual Alignment Reranker Paper • 2607.22042 • Published Jul 24 • 17
Skin-Deep: A Geometric Diagnostic for Alignment Fragility in Large Language Model Representations Paper • 2606.22676 • Published Jun 21 • 1
Improving Semantic Proximity in Information Retrieval through Cross-Lingual Alignment Paper • 2604.05684 • Published Apr 7 • 9
CLEAR: Cross-Lingual Enhancement in Alignment via Reverse-training Paper • 2604.05821 • Published Apr 7 • 3
LAMAR: An Open Language-Aware Multilingual Alignment Reranker Paper • 2607.22042 • Published Jul 24 • 17