About Me
- Hello, I'm Wei Liu (刘维). Here are my Email, Github and Google
Scholar.
- 2014-2018: Bachelor of Communication Engineering in BUPT
- 2018-2021: Master of Computer Engineering in CIST Lab@BUPT
- 2021-2023: NLP Researcher, Tencent
- 2023.8: I will be working as a RA at THUNLP, in collaboration with Prof. Zhiyuan Liu
- 🙋🏻♂ Now actively seeking Ph.D. opportunities (2024 Spring/Fall) on NLP/LLM!
Research Interests
- I focus on exploring the intelligence in deep natural language processing models when compressing and summarizing languages. I believe that in the process of language compression lies the birth of knowledge and intelligence.
- I also started to explore the next stage of intelligence in natural
languages by studying LLM multi-agent behaviors in a complex
environment or society. It is still an early exploration and I
am interested in several potential research directions, including:
- let LLMs build and use complicated tools with multi-agent collaboration
- self-organization and automatic job differentiation in multi-agent collaboration
- alignment in multi-agent activities
- improve the communication efficiency among agents
- make multi-agent activities improve LLMs
Research Details
- More Comprehensive and Factual Summarization:
- Introduce Determinantal Point Processes to solve the attention degeneration in Summarization[1].
- Discover the subjective bias in public summarization datasets which leads to text degeneration[2].
- Design a fine-tuning schema based on mutual information that minimizes hallucination in summarization[3].
- Improve the sentiment consistency in abstractive summarization through a memory-based approach[4].
- Scientific Paper Summarization, Multi-lingual Lay Summarization, Long Document Summarization[5,6,7].
- More Accurate and Controllable Keyphrase Prediction:
- Consider Keyphrase Ranking as an MRC task that better leverages PLM to improve performance. (Patent-only)
- Develop a unified present/absent Keyphrase Prediction method[8].
- Explore Controllable Keyphrase Generation as an early attempt at prompt engineering[9].
- Multi-Agents powered by LLMs:
- Release a repo for building complicated tools within a LLM multi-agent virtual environment: ChatDev[10]
Industrial Experience
- At Tencent, I aim to improve the performance of News Feed
Recommendations and Advertising.
- Improve the NLU ability for News Feed Recommendation by more accurate and controllable keyphrase prediction.
- Introducing non-commercial behaviors into advertising modeling through graph modeling.
- Explore stable and end-to-end feature quantization methods for Advertising models.
- Diverse user interest modeling in a diffusion-based way[In Submission].
- Achieve better tradeoffs between single-tower and two-tower models during the recall/pre-rank stage in Advertising[In Submission].
Publications
|
[1] In Conclusion Not Repetition: Comprehensive Abstractive Summarization with Diversified Attention Based on Determinantal Point Processes |
| CoNLL 2019 Long Paper | |
| Lei Li, Wei Liu, Marina Litvak, Natalia Vanetik, Zuying Huang | |
| code paper | |
|
[2] Subjective Bias in Abstractive Summarization |
| Arxiv Preprint | |
| Lei Li, Wei Liu, Marina Litvak, Natalia Vanetik, Jiacheng Pei, Yinan Liu, Siya Qi | |
| code paper | |
|
[3] CO2Sum: Contrastive Learning for Factual-Consistent Abstractive Summarization |
| Arxiv Preprint | |
| Wei Liu, Huanqin Wu, Wenjing Mu, Zhen Li, Tao Chen, Dan Nie | |
| code paper | |
|
[4] A Multi-View Abstractive Summarization Model Jointly Considering Semantics and Sentiment |
| CCIS 2018 Long Paper | |
| Moye Chen, Lei Li, Wei Liu | |
| paper | |
|
[5] CIST@CLSciSumm-19: Automatic Scientific Paper Summarization with Citances and Facets |
| SIGIR 2019 Shared Task | |
| Lei Li, Yingqi Zhu, Yang Xie, Zuying Huang, Wei Liu, Xingyuan Li, Yinan Liu | |
| paper | |
|
[6] Multi-lingual Wikipedia Summarization and Title Generation On Low Resource Corpus |
| RANLP 2019 Shared Task | |
| Wei Liu, Lei Li, Zuying Huang, Yinan Liu | |
| code paper | |
|
[7] CIST@CL-SciSumm 2020, LongSumm 2020: Automatic Scientific Document Summarization |
| EMNLP 2020 Shared Task | |
| Lei Li, Yang Xie, Wei Liu, Yinan Liu, Yafei Jiang, Siya Qi, Xingyuan Li | |
| paper | |
|
[8] UniKeyphrase: A Unified Extraction and Generation Framework for Keyphrase Prediction |
| ACL 2021 Findings Long Paper | |
| Huanqin Wu, Wei Liu, Lei Li, Dan Nie, Tao Chen, Feng Zhang, Di Wang | |
| code paper | |
|
[9] Fast and Constrained Absent Keyphrase Generation by Prompt-Based Learning |
| AAAI 2022 Long Paper | |
| Huanqin Wu, Baijiaxin Ma, Wei Liu, Tao Chen, Dan Nie | |
| code paper | |
|
[10] Communicative Agents for Software Development |
| Arxiv | |
| Chen Qian, Xin Cong, Wei Liu, Cheng Yang, Weize Chen, Yusheng Su, Yufan Dang, Jiahao Li, Juyuan Xu, Dahai Li, Zhiyuan Liu, Maosong Sun | |
| code paper |