Zhiyuan Liu

5 papers Tsinghua University

• A*-Thought: Efficient Reasoning via Bidirectional Compression for Low-Resource Settings
• Learning to Focus: Causal Attention Distillation via Gradient‐Guided Token Pruning
• Multi-Agent Collaboration via Evolving Orchestration
• ParamMute: Suppressing Knowledge-Critical FFNs for Faithful Retrieval-Augmented Generation
• The Overthinker's DIET: Cutting Token Calories with DIfficulty-AwarE Training