Since February 2025, I've been based in London pursuing my PhD journey! Fun Fact: My giraffe icon isn't random - I actually have a long neck! See for yourself and meet other talented researchers on our team page. Follow This Guy on Linkedin

About Me

• Hello, I'm Wei Liu (刘维). Welcome to my blog (0 views). Here are my Email, Linkedin, Github, and Google Scholar.
• My previous experience
  • 2014-2018: Bachelor of Communication Engineering in BUPT
  • 2018-2021: Master of Computer Engineering in CIST Lab@BUPT
  • 2021-2023: Application Research, Tencent
  • 2023.8-2025.1: Working at THUNLP with Prof. Zhiyuan Liu and Prof. Chen Qian with a focus on LLM Multi-Agent System.
  • 2025.2 Proud to be a PhD advised by Prof. Yulan He and a member of KCLNLP!.
• Recent News
  • 2025.5 Checkout KCLNLP's amazing works here, with 15 papers accepted by ACL 2025 and 3 papers accepted by ICML 2025!.
  • Check out NOVER, a novel verifier-free reinforcement learning framework for training Large Reasoning Model. Train your own R1-Zero-like reasoning model on ANY DATA!

Research Interests

• Natural Language Generation, especially on Compressing and Summarizing Languages.
• Inference Time Scaling in LLMs.
• Develop robust, safe, efficient, and human-centric LLM Multi-Agent System.
• Served as reviewer for
  • ACL(2021,2022,2024)
  • EMNLP(2020,2023,2024,2025)
  • NeurIPS(2024,2025)
  • ICLR(2024)
  • CVPR(2025)

Industrial Experience

• At Tencent, I aim to improve the performance of News Feed Recommendations and Advertising.
  • Improving the NLU ability for News Feed Recommendation.
  • Resolving the mismatch between commercial inclinations and content interests for Wechat Ads.
  • Stability, Warm-Up, Efficiency-Quality Tradeoff, Interpretability & Explainability on Large Recommendation System.
  • Diverse user interest modeling.

Publications

• Reasoning in LLMs:
  • paper code NOVER: Incentive Training for Language Models via Verifier-Free Reinforcement Learning. arXiv
• Multi-Agents powered by LLMs:
  • paper code Autonomous Agents for Collaborative Task under Information Asymmetry. NeurIPS 2024
  • paper code Scaling Large-Language-Model-based Multi-Agent Collaboration. ICLR 2025
  • paper code Communicative Agents for Software Development. ACL 2024
  • paper code Experiential Co-Learning of Software-Developing Agents. ACL 2024
  • paper code Iterative Experience Refinement of Software-Developing Agents.arXiv
  • paper code Multi-Agent Software Development through Cross-Team Collaboration. ACL 2025
  • paper code EcoLANG: Efficient and Effective Agent Communication Language Induction for Social Simulation.arXiv
• More Accurate and Controllable Keyphrase Prediction:
  • paper code UniKeyphrase: A Unified Extraction and Generation Framework for Keyphrase Prediction. ACL 2021 findings
  • paper code Fast and Constrained Absent Keyphrase Generation by Prompt-Based Learning. AAAI 2022
• More Comprehensive and Factual Summarization:
  • paper code In Conclusion Not Repetition: Comprehensive Abstractive Summarization with Diversified Attention Based on Determinantal Point Processes. CoNLL 2021
  • paper code Subjective Bias in Abstractive Summarization.arXiv
  • paper code CO2Sum: Contrastive Learning for Factual-Consistent Abstractive Summarization.arXiv
  • paper code Multi-lingual Wikipedia Summarization and Title Generation On Low Resource Corpus. RANLP 2019 workshop
  • paper A Multi-View Abstractive Summarization Model Jointly Considering Semantics and Sentiment. CCIS 2018
  • paper CIST@CLSciSumm-19: Automatic Scientific Paper Summarization with Citances and Facets. SIGIR 2019 workhop
  • paper CIST@CL-SciSumm 2020, LongSumm 2020: Automatic Scientific Document Summarization. EMNLP 2020 workshop