4 papers · 1 filter
CopT: Contrastive On-Policy Thinking with Continuous Spaces for General and Agentic Reasoning
Dachuan Shi, Hanlin Zhu, Xiangchi Yuan +4
Chain-of-thought (CoT) is a standard approach for eliciting reasoning capabilities from large language models (LLMs). However, the common CoT paradigm treats thinking as a prerequi…
Not All Heads Matter: A Head-Level KV Cache Compression Method with Integrated Retrieval and Reasoning
Yu Fu, Zefan Cai, Abedelkadir Asi +3
Key-Value (KV) caching is a common technique to enhance the computational efficiency of Large Language Models (LLMs), but its memory overhead grows rapidly with input length. Prior…
Cross-Task Defense: Instruction-Tuning LLMs for Content Safety
Yu Fu, Wen Xiao, Jia Chen +4
Recent studies reveal that Large Language Models (LLMs) face challenges in balancing safety with utility, particularly when processing long texts for NLP tasks like summarization a…
Safety Alignment in NLP Tasks: Weakly Aligned Summarization as an In-Context Attack
Yu Fu, Yufei Li, Wen Xiao +2
Recent developments in balancing the usefulness and safety of Large Language Models (LLMs) have raised a critical question: Are mainstream NLP tasks adequately aligned with safety…