Scholar
Tian Liang
Google Scholar ID: Z6bZcT4AAAAJ
Tencent AI Lab
Citations & Impact
Citations
1,462
H-index
13
i10-index
14
Publications
20
Co-authors
9
list available
Publications
Resume
Academic Achievements
- Published multiple papers at top venues including NeurIPS, ICML, EMNLP, ACL, and TACL. Selected works:
- DeepMath-103K: A large-scale, verifiable math reasoning dataset; >40K downloads on Hugging Face, 220+ GitHub stars.
- Encouraging Divergent Thinking via Multi-Agent Debate (EMNLP 2024): Pioneering multi-agent LLM framework; 520+ citations, 400+ GitHub stars.
- Do NOT Think That Much for 2+3=? (ICML 2025): Identified 'overthinking' in long-chain reasoning models; adopted by Kimi-1.5.
- Critical Tokens Matter (ICML 2025): Token-level contrastive estimation to enhance reasoning.
- Addressing Entity Translation Problem (ACL 2024): Tackled challenges in entity translation.
- Exploring Human-Like Translation Strategy with LLMs (TACL 2023): Investigated human-like translation in LLMs.
- CriticBench (ACL 2024): Benchmark for critique-correct reasoning in LLMs.
- ParroT (EMNLP 2023): Real-time chat translation using LLMs fine-tuned with human feedback.
- Competition awards: 2nd and 4th place in AAAI-22 Workshop shared tasks on acronym extraction and disambiguation.
