Wenxiao Wang

Assistant Professor@Zhejiang University·中国·OpenReview

研究方向

Natural Language Processing · Vision Transformer · Large Models · Knowledge Distillation · Neural Architecture Search (NAS) · Compression and Acceleration of CNN

Wenxiao Wang

ScalingCache: Extreme Acceleration of DiTs through Difference Scaling and Dynamic Interval Caching

HEAPr: Hessian-based Efficient Atomic Expert Pruning in Output Space

Differential Fine-Tuning Large Language Models Towards Better Diverse Reasoning Abilities

How Do Language Models Speak Languages? A Case Study on Unintended Code-Switching

IAgent: A Web Search Framework for Noise Isolation and Extended Information Access

SyncKV: A Syncopated Scheduling Approach to KV Cache Compression for Efficient Long-Context LLM Inference

AlignChat: Endowing LLMs with End-to-End Speech-to-Text Chat Capability through Token-Level Representation Alignment

Controlling Thinking Speed in Reasoning Models

GeoCAD: Local Geometry-Controllable CAD Generation with Large Language Models

Don't Take Things Out of Context: Attention Intervention for Enhancing Chain-of-Thought Reasoning in Large Language Models

Differential Fine-Tuning Large Language Models Towards Better Diverse Reasoning Abilities

Improving Complex Reasoning with Dynamic Prompt Corruption: A Soft Prompt Optimization Approach

Large Language Models Can Be More Robust Multiple Choice Selectors Through Attention Intervention

FlexCAD: Unified and Versatile Controllable CAD Generation with Fine-tuned Large Language Models

Interpretable Analysis and Reasoning Enhancement for LLMs via Cross-Generation Reasoning Trees

Simple and Fast CNN for Vision

SciPIP: An LLM-based Scientific Paper Idea Proposer