Portrait of Wenhao Deng
    Wenhao Deng 邓文豪 (Man-Ho Tang)
  • The trouble maker
  • The prompt evolver
  • The agent orchestrator
  • The problem solver ← I wish I am
GitHub Google Scholar LinkedIn Twitter
Email: wenhao [dot] deng [at] foxmail [dot] com

What I do?

I do what I want to do.

Education

Papers

CIKM 2025 Exploring the Upper Limits of Text-Based Collaborative Filtering Using Large Language Models: Discoveries and Insights
Ruyu Li*, Wenhao Deng*, Yu Cheng, Zheng Yuan, Jiaqi Zhang, Fajie Yuan (* Equal contribution)
TL;DR
We investigate how far text-based collaborative filtering can go with extremly large language models, revealing key insights on leveraging language models for recommendation systems.
Preprint BuildArena: A Physics-Aligned Interactive Benchmark of LLMs for Engineering Construction
Tian Xia, Tianrun Gao, Wenhao Deng, Long Wei, Xiaowei Qian, Yixian Jiang, Chenglei Yu, Tailin Wu
TL;DR
arXiv:2510.16559 (2025)
We introduce BuildArena, the first physics-aligned interactive benchmark for language-driven engineering construction, comprehensively evaluating LLMs' capabilities for construction automation under strict physical constraints through a highly customizable agentic engineering framework with static and dynamic mechanics tasks.
Preprint Unlocking Reasoning Capabilities in LLMs via Reinforcement Learning Exploration
Wenhao Deng*, Long Wei*, Chenglei Yu*, Tailin Wu (* Equal contribution)
TL;DR
arXiv:2510.03865 (2025)
We attributes one of the limitations of Reinforcement Learning via Verifiable Reward (RLVR) to the reverse KL divergence, and propose Reward-Aware Policy Optimization with forward KL divergence to enhance the reasoning capabilities of large language models.
Preprint Towards Understanding the Capability of Large Language Models on Code Clone Detection: A Survey
Shihan Dou*, Junjie Shan*, Haoxiang Jia, Wenhao Deng, Zhiheng Xi, Wei He, Yueming Wu, Tao Gui, Yang Liu, Xuanjing Huang (* Equal contribution)
TL;DR
arXiv:2308.01191 (2023)
An early survey comprehensively examining how well large language models perform on code clone detection tasks and their capabilities.

Reviewers

*You don’t need to solve your entire life overnight, just focus on one small step at a time.*
Powered by Jekyll and Lanyon.