Liangming Pan
Assistant Professor
Researches natural language processing and large language models, focusing on reasoning and interpretability of LLMs.
Liangming Pan's lab at Peking works on large language models, reasoning & agents and trustworthy & safe ML. 59 papers since 2023, 50 at top venues, mostly EMNLP and ACL. Applications close .
At a glance
- papers, 2023–2026
- 15/ year
- at T1 venues
- 85%
- PhD students
- —
- MS students
- —
- alumni listed
- —
Papers by yearfrom the lab homepage, 2023–2026
- 12 papers2026 · T1 12 · collecting
- 15 papers2025 · T1 13
- 17 papers2024 · T1 15
- 15 papers2023 · T1 10
T1 venuesOtherVenues: EMNLP, ACL, ICML, NeurIPS, AAAI
Research focuspaper titles that use each topic's words
- Large language models29 papers
- Reasoning & agents21 papers
- interpretability1 paper
- Trustworthy & safe ML0 papers
- researches natural0 papers
Paper list
From the lab homepage.
202612 papers
- TMLR (T1)AlignSAE: Concept-Aligned Sparse Autoencoders (opens in a new tab)
- NeurIPS (T1)Do Latent-CoT Models Think Step-by-Step? A Mechanistic Study on Sequential Reasoning Tasks (opens in a new tab)
- ACL (T1)LEDOM: Reverse Language Model (opens in a new tab)
- ICML (T1)Mechanistic Data Attribution: Tracing the Training Origins of Interpretable LLM Units (opens in a new tab)
- ICML (T1)OpenIKLR: Bridging the Reasoning Gap in Open-World Scenarios via Iterative Premise Completion (opens in a new tab)
- EMNLP (T1)ProEvent: An Event-centric Benchmark for Proactive Agents (opens in a new tab)
- EMNLP (T1)The Mirage of Explainability: A Survey on Chain-of-Thought Faithfulness in Large Language Models (opens in a new tab)
- ACL (T1)Towards a Mechanistic Understanding of Large Reasoning Models: A Survey of Training, Inference, and Failures (opens in a new tab)
- EMNLP (T1)Towards a Mechanistic Understanding of Propositional Logical Reasoning in Large Language Models (opens in a new tab)
- ACL (T1)Towards Intrinsic Interpretability of Large Language Models:A Survey of Design Principles and Architectures (opens in a new tab)
- EMNLP (T1)Where Steering Signals Come From: Activation Source Selection in Activation Steering (opens in a new tab)
- EMNLP (T1)Wrong Prediction, Right Answer: Recovering Evidence from Collapsed LLM Sequence Scores (opens in a new tab)
202515 papers
- ACL (T1)AntiLeak-Bench: Preventing Data Contamination by Automatically Constructing Benchmarks with Updated Real-World Knowledge (opens in a new tab)
- ACL (T1)Aristotle: Mastering Logical Reasoning with A Logic-Complete Decompose-Search-Resolve Framework (opens in a new tab)
- NAACL (T1)CausalEval: Towards Better Causal Reasoning in Language Models (opens in a new tab)
- AAAI (T1)Combating Multimodal LLM Hallucination via Bottom-Up Holistic Reasoning (opens in a new tab)
- EMNLP (T1)ConciseRL: Conciseness-Guided Reinforcement Learning for Efficient Reasoning Models (opens in a new tab)
- ACL (T1)Gödel Agent: A Self-Referential Agent Framework for Recursive Self-Improvement (opens in a new tab)
- NeurIPS (T1)How does Transformer Learn Implicit Reasoning? (opens in a new tab)
- EMNLP (T1)How Is LLM Reasoning Distracted by Irrelevant Context? An Analysis Using a Controlled Benchmark (opens in a new tab)
- ACL (T1)InductionBench: LLMs Fail in the Simplest Complexity Class (opens in a new tab)
- NAACL (T1)Investigating the Transferability of Code Repair for Low-Resource Programming Languages (opens in a new tab)
- NeurIPS (T1)MuSLR: Multimodal Symbolic Logical Reasoning (opens in a new tab)
- ACL (T1)RuleArena: A Benchmark for Rule-Guided Reasoning with LLMs in Real-World Scenarios (opens in a new tab)
202417 papers
- TMLR (T1)A Survey on Data Selection for Language Models (opens in a new tab)
- EMNLP (T1)A Survey on Detection of LLMs-Generated Content (opens in a new tab)
- EMNLP (T1)AKEW: Assessing Knowledge Editing in the Wild (opens in a new tab)
- TACL (T1)Automatically Correcting Large Language Models: Surveying the Landscape of Diverse Automated Correction Strategies (opens in a new tab)
- EMNLP (T1)Factcheck-Bench: Fine-Grained Evaluation Benchmark for Automatic Fact-checkers (opens in a new tab)
- ACL (T1)Faithful Logical Reasoning via Symbolic Chain-of-Thought (opens in a new tab)
- ACL (T1)Knowledge of Knowledge: Exploring Known-Unknowns Uncertainty with Large Language Models (opens in a new tab)
- Annual Conference on Neural Information Processing Systems (NeurIPS)MMLongBench-Doc: Benchmarking Long-context Document Understanding with Visualizations (opens in a new tab)
- ACL (T1)Modeling Dynamic Topics in Chain-Free Fashion by Evolution-Tracking Contrastive Learning and Unassociated Word Exclusion (opens in a new tab)
- EMNLP (T1)MultiAgent Collaboration Attack: Investigating Adversarial Attacks in Large Language Model Collaborations via Debate (opens in a new tab)
- ACL (T1)Pride and Prejudice: LLM Amplifies Self-Bias in Self-Refinement (opens in a new tab)
- EMNLP (T1)SciAgent: Tool-augmented Language Models for Scientific Reasoning (opens in a new tab)
202315 papers
- IJCNLP-AACLAttacking Open-domain Question Answering by Injecting Misinformation (opens in a new tab)
- EMNLP (T1)Doolittle: Benchmarks and Corpora for Academic Writing Formalization (opens in a new tab)
- NeurIPS WorkshopEfficient Online Data Mixing For Language Model Pre-Training (opens in a new tab)
- ACL (T1)Fact-Checking Complex Claims with Program-Guided Reasoning (opens in a new tab)
- IJCNLP-AACLFollowupQG: Towards Information-Seeking Follow-up Question Generation (opens in a new tab)
- WWW (T1)Hashtag-Guided Low-Resource Tweet Classification (opens in a new tab)
- AAAI (T1)InfoCTM: A Mutual Information Maximization Perspective of Cross-Lingual Topic Modeling (opens in a new tab)
- EMNLP (T1)INSTRUCTSCORE: Towards Explainable Text Generation Evaluation with Automatic Feedback (opens in a new tab)
- IJCNLP-AACLInvestigating Zero- and Few-shot Generalization in Fact Verification (opens in a new tab)
- EMNLP (T1)Logic-LM: Empowering Large Language Models with Symbolic Solvers for Faithful Logical Reasoning (opens in a new tab)
- EMNLP (T1)MAF: Multi-Aspect Feedback for Improving Reasoning in Large Language Models (opens in a new tab)
- ACL (T1)Modeling What-to-ask and How-to-ask for Answer-unaware Conversational Question Generation (opens in a new tab)
Where alumni went
Alumni are not listed on the lab homepage.
Apply to PekingFall 2027
Computer Science PhD (international applicants)
- Round
- Main
- Opens
- —
- GRE
- See program page
- English
- See program page
- Fee
- See program page
Not yet confirmed for this intake.
Similar labsnearest by research summary
- CNPeking
Zhihong Deng
16 T1 - SGNTU
Wenya Wang
- Reasoning & agents
- USUC Berkeley
Alane Suhr
20 T1 - GBEdinburgh
Adam Lopez
- Multilingual, translation & linguistics
- GBCambridge
Anna Korhonen
- Multilingual, translation & linguistics
- USCMU
Daphne Ippolito
- Interpretability, evaluation & safety of LMs
Your notes
Save this lab to keep notes.
Data
Some numbers could not be verified:
- The member list could not be found on the lab site.
- The alumni list could not be found on the lab site.
- Collected
- Verified
- Papers
- The lab homepage
- Members
- Not collected
- Alumni
- Not collected
Tiers are venue classifications, not lab rankings.
Pekingcloses