Pengfei Liu
Tenured Track Associate Professor
Leads the Generative AI Research (GAIR) group, working on large model training, reasoning and multimodal world models.
Pengfei Liu's lab at SJTU works on large language models, reasoning & agents and generative & foundation models. 51 papers since 2023, 33 at top venues, mostly Findings of ACL and NeurIPS. 12 PhD students and 10 master's students. 21 alumni. Applications close .
At a glance
- papers, 2023–2026
- 13/ year
- at T1 venues
- 65%
- PhD students
- 12
- MS students
- 10
- alumni listed
- 21
Papers by yearfrom DBLP, 2023–2026
- 10 papers2026 · T1 6 · collecting
- 10 papers2025 · T1 9
- 22 papers2024 · T1 12
- 9 papers2023 · T1 6
T1 venuesOtherVenues: Findings of ACL, NeurIPS, ACL, EMNLP, ICLR
Research focuspaper titles that use each topic's words
- Large language models17 papers
- Generative & foundation models11 papers
- Reasoning & agents9 papers
- large model0 papers
- model training0 papers
Paper list
From DBLP.
202610 papers
- ACL (T1)AgencyBench: Benchmarking the Frontiers of Autonomous Agents in 1M-Token Real-World Contexts (opens in a new tab)
- ECCV (T1)CiQi-Agent: Aligning Vision, Tools and Aesthetics in Multimodal Agent for Cultural Reasoning on Chinese Porcelains (opens in a new tab)
- Findings of ACLDiagnosisArena: Benchmarking Diagnostic Reasoning for Large Language Models (opens in a new tab)
- ICLR (T1)Efficient Agent Training for Computer Use (opens in a new tab)
- ICLR (T1)InnovatorBench: Evaluating Agents' Ability to Conduct Innovative AI Research (opens in a new tab)
- CHI Extended AbstractsInteraction as Intelligence: A Transparent, Fine-Grained Multi-Agent Deep Research System for Human-Agent Collaboration (opens in a new tab)
- Findings of ACLLongCLI-Bench: A Preliminary Benchmark and Study for Long-horizon Agentic Programming in Command-Line Interfaces (opens in a new tab)
- AAAI (T1)SCALE: Selective Resource Allocation for Overcoming Performance Bottlenecks in Mathematical Test-time Scaling (opens in a new tab)
- Findings of ACLSepSeq: A Training-Free Framework for Long Numerical Sequence Processing in LLMs (opens in a new tab)
- ICLR (T1)SR-Scientist: Scientific Equation Discovery With Agentic AI (opens in a new tab)
202510 papers
- EMNLP (T1)DeepResearcher: Scaling Deep Research via Reinforcement Learning in Real-world Environments (opens in a new tab)
- AAAI (T1)Evaluating Mathematical Reasoning Beyond Accuracy (opens in a new tab)
- NAACL (T1)Libra-Leaderboard: Towards Responsible AI through a Balanced Leaderboard of Safety and Capability (opens in a new tab)
- NeurIPS (T1)LIMOPro: Reasoning Refinement for Efficient and Effective Test-time Scaling (opens in a new tab)
- ICML (T1)OmniBal: Towards Fast Instruction-Tuning for Vision-Language Models via Omniverse Computation Balance (opens in a new tab)
- NeurIPS (T1)On Evaluating LLM Alignment by Evaluating LLMs as Judges (opens in a new tab)
- ICML (T1)Programming Every Example: Lifting Pre-training Data Quality Like Experts at Scale (opens in a new tab)
- ICLR (T1)Progress or Regress? Self-Improvement Reversal in Post-training (opens in a new tab)
- ACL (T1)Towards Dynamic Theory of Mind: Evaluating LLM Adaptation to Temporal Evolution of Human States (opens in a new tab)
- Findings of NAACLUnderstanding Reference Policies in Direct Preference Optimization (opens in a new tab)
202422 papers
- NeurIPS (T1)Alignment for Honesty (opens in a new tab)
- Findings of NAACLBenchmarking Generation and Evaluation Capabilities of Large Language Models for Instruction Controllable Summarization (opens in a new tab)
- CONDAData Contamination Report from the 2024 CONDA Shared Task (opens in a new tab)
- ACL (T1)Dissecting Human and LLM Preferences (opens in a new tab)
- EMNLP (T1)ECON: On the Detection and Resolution of Evidence Conflicts (opens in a new tab)
- EMNLP (T1)FRoG: Evaluating Fuzzy Reasoning of Generalized Quantifiers in LLMs (opens in a new tab)
- ICLR (T1)Generative Judge for Evaluating Alignment (opens in a new tab)
- NAACL (T1)GPTScore: Evaluate as You Desire (opens in a new tab)
- Findings of ACLInFoBench: Evaluating Instruction Following Ability in Large Language Models (opens in a new tab)
- Findings of ACLLLMCrit: Teaching Large Language Models to Use Criteria (opens in a new tab)
- NeurIPS (T1)MathPile: A Billion-Token-Scale Pretraining Corpus for Math (opens in a new tab)
- Big Data Min. Anal.MedBench: A Comprehensive, Standardized, and Reliable Benchmarking System for Evaluating Chinese Medical Large Language Models (opens in a new tab)
20239 papers
- ACL (T1)DataFinder: Scientific Dataset Recommendation from Natural Language Descriptions (opens in a new tab)
- NeurIPS (T1)FELM: Benchmarking Factuality Evaluation of Large Language Models (opens in a new tab)
- EMNLP (T1)GlobalBench: A Benchmark for Global Progress in Natural Language Processing (opens in a new tab)
- Findings of ACLMulti-Dimensional Evaluation of Text Summarization with In-Context Learning (opens in a new tab)
- ICML (T1)PAL: Program-aided Language Models (opens in a new tab)
- ACM Comput. Surv.Pre-train, Prompt, and Predict: A Systematic Survey of Prompting Methods in Natural Language Processing (opens in a new tab)
- ACL (T1)Revisiting the Gold Standard: Grounding Summarization Evaluation with Robust Human Evaluation (opens in a new tab)
- Findings of EMNLPT5Score: Discriminative Fine-tuning of Generative Evaluation Metrics (opens in a new tab)
- EMNLP (T1)Towards Interpretable and Efficient Automatic Reference-Based Summarization Evaluation (opens in a new tab)
Where alumni went21 listed
Apply to SJTUFall 2027
Computer Science PhD (international applicants)
- Round
- Self-funded
- Opens
- —
- GRE
- See program page
- English
- See program page
- Fee
- See program page
Not yet confirmed for this intake.
Similar labsnearest by research summary
- USUSC
Xiang Ren
- Large language models
- USMichigan
Huteng Dai
- Large language models
- USCMU
Daphne Ippolito
- Interpretability, evaluation & safety of LMs
- USNYU
Chen Zhao
- Reasoning & agents
- GBCambridge
Anna Korhonen
- Multilingual, translation & linguistics
- CAToronto
Ewan Dunbar
- Speech recognition & spoken language
Your notes
Save this lab to keep notes.
Data
Checked against the lab homepage.
- Collected
- Verified
- Papers
- DBLP
- Members
- Lab homepage
- Alumni
- Lab homepage
Tiers are venue classifications, not lab rankings.
SJTUcloses