Hua Wu

ProxyAttn: Guided Sparse Attention via Representative Heads

Enough is as good as a feast: A Comprehensive Analysis of How Reinforcement Learning Mitigates Task Conflicts in LLMs

LLMs are Single-threaded Reasoners: Demystifying the Working Mechanism of Soft Thinking

Elastic MoE: Unlocking the Inference-Time Scalability of Mixture-of-Experts

Knowledge-Level Consistency Reinforcement Learning: Dual-Fact Alignment for Long-Form Factuality

Extending RLVR to Open-Ended Tasks via Verifiable Multiple-Choice Reformulation

A2R: An Asymmetric Two-Stage Reasoning Framework for Parallel Reasoning

Semantic Energy: Detecting LLM Hallucination Beyond Entropy

Mixture of Hidden-Dimensions: Not All Hidden-States’ Dimensions are Needed in Transformer

Residual Stream Analysis of Overfitting And Structural Disruptions

MA-RLHF: Reinforcement Learning from Human Feedback with Macro Actions

Upcycling Instruction Tuning from Dense to Mixture-of-Experts via Parameter Merging

BEVWorld: A Multimodal World Model for Autonomous Driving via Unified BEV Latent Space

HFT: Half Fine-Tuning for Large Language Models