Lu Wang
Associate Professor, EECS – Computer Science and Engineering
Studies natural language processing, including text summarization, trustworthy and safe language models and AI evaluation.
Lu Wang's lab at Michigan works on interpretability, evaluation & safety of LMs, large language models and trustworthy & safe ML. 60 papers since 2023, 34 at top venues, mostly EMNLP and ACL. 7 PhD students. 17 alumni, most often to Google DeepMind and NVIDIA. Applications close .
At a glance
- papers, 2023–2026
- 16/ year
- at T1 venues
- 57%
- PhD students
- 7
- MS students
- 0
- alumni listed
- 17
Papers by yearfrom the lab homepage, 2023–2026
- 19 papers2026 · T1 11 · collecting
- 14 papers2025 · T1 9
- 15 papers2024 · T1 7
- 12 papers2023 · T1 7
T1 venuesOtherVenues: EMNLP, ACL, Findings of EMNLP, Findings of ACL, ICLR
Research focuspaper titles that use each topic's words
- Large language models30 papers
- Interpretability, evaluation & safety of LMs5 papers
- summarization3 papers
- Trustworthy & safe ML0 papers
- studies natural0 papers
Paper list
From the lab homepage.
202619 papers
- ICML (T1)AdaMEM: Test-Time Adaptive Memory for Language Agents (opens in a new tab)
- COLMCan LLM Agents Discover? Evaluating Creativity on ML Engineering Tasks (opens in a new tab)
- ACL (T1)CASPER in the Machine: Insights into Character Variety in LLM-Generated Stories
- ICLR (T1)CLASH: Evaluating Language Models on Judging High-Stakes Dilemmas from Multiple Perspectives (opens in a new tab)
- EMNLP (T1)Countdown-Code: A Testbed for Studying The Emergence and Generalization of Reward Hacking in RLVR (opens in a new tab)
- Findings of ACLDo Language Models Think Consistently? A Study of Value Preferences Across Varying Response Lengths (opens in a new tab)
- ACL (T1)Do LLMs Really Need 10+ Thoughts for "Find the Time 1000 Days Later"? Towards Structural Understanding of LLM Overthinking (opens in a new tab)
- COLMEvaluating the Retrieval Robustness of Large Language Models (opens in a new tab)
- ICLR (T1)ExpertLongBench: Benchmarking Language Models on Expert-Level Long-Form Generation Tasks with Structured Checklists (opens in a new tab)
- Findings of EMNLPGaming the Judge: Unfaithful Chain-of-Thought Can Undermine Agent Evaluation (opens in a new tab)
- ICML (T1)LiveOIBench: Can Large Language Models Outperform Human Contestants in Informatics Olympiads? (opens in a new tab)
- ICLR (T1)LLMs as Rules Oracles: Exploring Real-World Multimodal Reasoning in Tabletop Strategy Game Environments
202514 papers
- Proceedings of the Conference on Empirical Methods in Natural Language Processing (EMNLP), short paperAnswer Convergence as a Signal for Early Stopping in Reasoning (opens in a new tab)
- ACL (T1)Efficient Ensemble for Fine-tuning Language Models on Multiple Datasets (opens in a new tab)
- TrustNLPEvaluating Design Choices in Verifiable Generation with Open-source Models (opens in a new tab)
- ACL (T1)FactBench: A Dynamic Benchmark for In-the-Wild Language Model Factuality Evaluation (opens in a new tab)
- AACLFINDR: A Fast Influential Data Selector for NL2Code Pretraining (opens in a new tab)
- EMNLP (T1)Linear-Time Demonstration Selection for In-Context Learning via Gradient Estimation (opens in a new tab)
- Findings of ACLMDBench: A Synthetic Multi-Document Reasoning Benchmark Generated with Knowledge Guidance (opens in a new tab)
- Proceedings of the Annual Conference on Neural Information Processing Systems (NeurIPS), Datasets & Benchmarks TrackMLRC-Bench: Can Language Agents Solve Machine Learning Research Challenges? (opens in a new tab)
- ACL (T1)On Many-Shot In-Context Learning for Long-Context Evaluation (opens in a new tab)
- EMNLP (T1)PRIME: Large Language Model Personalization with Cognitive Memory and Thought Processes (opens in a new tab)
- EMNLP (T1)Structured Moral Reasoning in Language Models: A Value-Grounded Evaluation Framework (opens in a new tab)
- EMNLP (T1)SYNC: A Synthetic Long-Context Understanding Benchmark for Controlled Comparisons of Model Capabilities (opens in a new tab)
202415 papers
- LRECAnalyzing Occupational Distribution Representation in Japanese Language Models (opens in a new tab)
- NAACL (T1)AWESOME: GPU Memory-constrained Long Document Summarization using Memory Mechanism and Global Salient Content (opens in a new tab)
- EMNLP (T1)Closing the Loop: Learning to Generate Writing Feedback via Language Model Simulated Student Revisions (opens in a new tab)
- Findings of ACLEnhanced Language Model Truthfulness with Learnable Intervention and Uncertainty Expression (opens in a new tab)
- EMNLP (T1)Enhancing Language Model Factuality via Activation-Based Confidence Calibration and Guided Decoding (opens in a new tab)
- ICLR (T1)LitCab: Lightweight Language Model Calibration over Short- and Long-form Responses (opens in a new tab)
- ACL (T1)MIDGARD: Self-Consistency Using Minimum Description Length for Structured Commonsense Reasoning (opens in a new tab)
- NAACL (T1)MOKA: Moral Knowledge Augmentation for Moral Event Extraction (opens in a new tab)
- Findings of EMNLPNarrative-of-Thought: Improving Temporal Reasoning of Large Language Models via Recounted Narratives (opens in a new tab)
- NAACL (T1)PELMS: Pre-training for Effective Low-Shot Multi-Document Summarization (opens in a new tab)
- Findings of EMNLPScalable Fine-tuning from Multiple Data Sources: A First-Order Approximation Approach (opens in a new tab)
- Findings of EMNLPShoes-ACOSI: A Dataset for Aspect-Based Sentiment Analysis with Implicit Opinion Extraction (opens in a new tab)
202312 papers
- EMNLP (T1)All Things Considered: Detecting Partisan Events from News Media with Cross-Article Comparison (opens in a new tab)
- ACL (T1)BOLT: Fast Energy-based Controlled Text Generation with Tunable Biases (opens in a new tab)
- EMNLP (T1)Cross-Cultural Analysis of Human Values, Morals, and Biases in Folk Tales (opens in a new tab)
- Findings of EMNLPCrossing the Aisle: Unveiling Partisan and Counter-Partisan Events in News Reporting (opens in a new tab)
- ME-FoMoExploring Demonstration Ensembling for In-context Learning (opens in a new tab)
- ACL (T1)Few-shot Reranking for Multi-hop QA via Language Model Prompting (opens in a new tab)
- TACL (T1)General then Personal: Decoupling and Pre-training for Personalized Headline Generation (opens in a new tab)
- Findings of EMNLPGRACE: Discriminator-Guided Chain-of-Thought Reasoning (opens in a new tab)
- EMNLP (T1)Merging Generated and Retrieved Knowledge for Open-Domain QA (opens in a new tab)
- CHI (T1)ReadingQuizMaker: A Human-NLP Collaborative System to Support Instructors Design High Quality Reading Quiz Questions (opens in a new tab)
- WNUWord Category Arcs in Literature Across Languages and Genres (opens in a new tab)
- Findings of EMNLPYou Are What You Annotate: Towards Better Models through Annotator Representations (opens in a new tab)
Where alumni went17 listed, 29% to industry
Industry 2Academia 5
Most often to
- Google DeepMind
- NVIDIA
Also to
- EPFL
- UIUC
Apply to MichiganFall 2027
Michigan CSE Computer Science PhD
- Round
- Single round
- Opens
- —
- GRE
- See program page
- English
- See program page
- Fee
- See program page
Not yet confirmed for this intake.
Similar labsnearest by research summary
- USUCSD
Prithviraj Ammanabrolu
- Reasoning & agents
- USStanford
Noah Goodman
- Large language models
- GBEdinburgh
Wenda Li
- Reasoning & neuro-symbolic AI
- SGNUS
Tat-Seng Chua
- Large language models
- USStanford
Tatsunori Hashimoto
- Interpretability, evaluation & safety of LMs
- CNTsinghua
Qingyao Ai
- Large language models
Your notes
Save this lab to keep notes.
Data
Checked against the lab homepage.
- Collected
- Verified
- Papers
- The lab homepage
- Members
- Lab homepage
- Alumni
- Lab homepage
Tiers are venue classifications, not lab rankings.
Michigancloses