Haiyang Yu

Adaptive Social Learning via Mode Policy Optimization for Language Agents

FrontierBench: Are We Only Testing Agents Under the Streetlight?

On the Role of Attention Heads in Large Language Model Safety

StructRAG: Boosting Knowledge Intensive Reasoning of LLMs via Inference-time Hybrid Information Structurization

Extend Model Merging from Fine-Tuned to Pre-Trained Large Language Models via Weight Disentanglement