Maarten Sap
Assistant Professor
Studies social commonsense reasoning and the social impact and safety of language models.
Maarten Sap's lab at CMU works on language & society and large language models. 101 papers since 2023, 51 at top venues, mostly EMNLP and ACL. 7 PhD students. 2 alumni, most often to Google DeepMind. Applications close .
At a glance
- papers, 2023–2026
- 26/ year
- at T1 venues
- 50%
- PhD students
- 7
- MS students
- 0
- alumni listed
- 2
Papers by yearfrom the lab homepage, 2023–2026
- 38 papers2026 · T1 15 · collecting
- 34 papers2025 · T1 15
- 17 papers2024 · T1 13
- 12 papers2023 · T1 8
T1 venuesOtherVenues: EMNLP, ACL, ICLR, FAccT, AAAI/ACM AIES
Research focuspaper titles that use each topic's words
- Large language models43 papers
- Language & society17 papers
- social14 papers
- commonsense reasoning1 paper
- social impact0 papers
Paper list
From the lab homepage.
202638 papers
- CSCW (T1)"I Just Don't Want My Work Being Fed Into The AI Blender": Queer Artists on Refusing and Resisting Generative AI (opens in a new tab)
- IASEAI1-2-3 Check: Enhancing Contextual Privacy in LLM via Multi-Agent Reasoning (opens in a new tab)
- ICLR (T1)Ambig-SWE: Interactive Agents to Overcome Underspecificity in Software Engineering (opens in a new tab)
- CHI (T1)Black LLMirror: User (Self) Perceptions in Black American English Interactions with LLMs (opens in a new tab)
- COLMCCBENCH: Assessing LLM Cultural Competence via Implicitly Signaled Norms using Health Queries (opens in a new tab)
- COLMCognitive Chain-of-Thought: Structured Multimodal Reasoning about Social Situations (opens in a new tab)
- EACLCommon Sense or Ableism? Rethinking Commonsense Reasoning Through the Lens of Disability (opens in a new tab)
- EMNLP (T1)Critical or Compliant? The Double-Edged Sword of Reasoning in Chain-of-Thought Explanations (opens in a new tab)
- Digital Hate ReviewEvaluating Large Language Models for Antisemitic Incident Classification (opens in a new tab)
- Findings of EMNLPEVALUESTEER: Measuring Reward Model Steerability Towards Values and Preference (opens in a new tab)
- AAAI/ACM AIESExamining the Effect of Explanations of AI Privacy Redaction in AI-mediated Interactions (opens in a new tab)
- Findings of EMNLPFortifying Hallucination Detection Under Distribution Shift in Short-Form QA
202534 papers
- NAACL (T1)AI-LieDar: Examine the Trade-off Between Utility and Truthfulness in LLM Agents (opens in a new tab)
- COLMALFA: Aligning LLMs to Ask Good Questions: A Case Study in Clinical Reasoning (opens in a new tab)
- Nature Machine Intelligence (T1)An Empirical Investigation of Machines’ Capabilities for Moral Judgment with the Delphi Experiment (opens in a new tab)
- NeurIPS D&BArtificial Hivemind: The Open-Ended Homogeneity of Language Models (and Beyond) (opens in a new tab)
- CVPR (T1)AutoPresent: Designing Structured Visuals from Scratch (opens in a new tab)
- ACL (T1)BIG5-CHAT: Shaping LLM Personalities Through Training on Human-Grounded Data (opens in a new tab)
- Multilingual and Equitable Language Technologies (MELT) Workshop at COLMBreaking mBad! Supervised Fine-tuning for Cross-Lingual Detoxification (opens in a new tab)
- MELT WorkshopDisparities in LLM Reasoning Accuracy and Explanations: A Case Study on African American English (opens in a new tab)
- KDD WorkshopExploring Big Five Personality and AI Capability Effects in LLM-Simulated Negotiation Dialogues (opens in a new tab)
- COLMFluid Language Model Benchmarking (opens in a new tab)
- COLMHAICOSYSTEM: An Ecosystem for Sandboxing Safety Risks in Human-AI Interactions (opens in a new tab)
- Findings of EMNLPLet Them Down Easy! Contextual Effects of LLM Guardrails on User Perceptions and Preferences (opens in a new tab)
202417 papers
- ICLR (T1)Can LLMs Keep a Secret? Testing Privacy Implications of Language Models via Contextual Integrity Theory (opens in a new tab)
- EACLClever Hans or Neural Theory of Mind? Stress Testing Social Reasoning in Large Language Models (opens in a new tab)
- CHI (T1)Counterspeakers’ Perspectives: Unveiling Barriers and AI Needs in the Fight against Online Hate (opens in a new tab)
- JMIR Mental HealthEmpathy Towards AI vs Human Experiences: The Role of Transparency in Mental Health and Social Support Chatbot Design (opens in a new tab)
- EMNLP (T1)HEART-felt Narratives: Tracing Empathy and Narrative Style in Personal Stories with LLMs (opens in a new tab)
- ACL (T1)Is the Pope Catholic? Yes, the Pope is Catholic. Generative Evaluation of Non-Literal Intent Resolution in LLMs (opens in a new tab)
- EMNLP (T1)Is This the Real Life? Is This Just Fantasy? The Misleading Success of Simulating Social Interactions With LLMs (opens in a new tab)
- ICLR (T1)Leftover-Lunch: Advantage-based Offline Reinforcement Learning for Language Models (opens in a new tab)
- AAAI/ACM AIESParticip-AI: A Democratic Surveying Framework for Anticipating Future AI Use Cases, Harms and Benefits (opens in a new tab)
- COLMPolygloToxicityPrompts: Multilingual Evaluation of Neural Toxic Degeneration in Large Language Models (opens in a new tab)
- ACL (T1)Relying on the Unreliable: The Impact of Language Models' Reluctance to Express Uncertainty (opens in a new tab)
- ACL (T1)SOTOPIA-$\pi$: Interactive Learning of Socially Intelligent Language Agents (opens in a new tab)
202312 papers
- EMNLP (T1)``Don't Take This Out of Context!'' On the Need for Contextual Models and Evaluations for Stylistic Rewriting (opens in a new tab)
- Findings of EMNLPBeyond Denouncing Hate: Strategies for Countering Implied Biases and Stereotypes in Language (opens in a new tab)
- EMNLP (T1)BiasX: ``Thinking Slow'' in Toxic Content Moderation with Explanations of Implied Social Biases (opens in a new tab)
- Findings of ACLCOBRA Frames: Contextual Reasoning about Effects and Harms of Offensive Statements (opens in a new tab)
- ACL (T1)Detoxifying Text with MaRCo: Controllable Revision with Experts and Anti-Experts (opens in a new tab)
- EMNLP (T1)FANToM: A Benchmark for Stress-testing Machine Theory of Mind in Interactions (opens in a new tab)
- ACL (T1)From Dogwhistles to Bullhorns: Unveiling Coded Rhetoric with Language Models (opens in a new tab)
- EMNLP (T1)Modeling Empathic Similarity in Personal Narratives (opens in a new tab)
- ACL (T1)NLPositionality: Characterizing Design Biases of Datasets and Models (opens in a new tab)
- FAccTQueer In AI: A Case Study in Community-Led Participatory AI (opens in a new tab)
- ACL demonstrationsRiveter: Measuring Power and Social Dynamics Between Entities (opens in a new tab)
- EMNLP (T1)SODA: Million-scale Dialogue Distillation with Social Commonsense Contextualization (opens in a new tab)
Where alumni went2 listed, 50% to industry
Industry 1Academia 1
Most often to
- Google DeepMind
Also to
- Carnegie Mellon University
Apply to CMUFall 2027
CMU SCS Computer Science PhD
- Round
- Single round
- Opens
- GRE
- See program page
- English
- TOEFL / IELTS / Duolingo (F-1/J-1 non-native speakers)
- Fee
- USD 100 (USD 80 by early deadline)
Confirmed on the program page.
Similar labsnearest by research summary
- SGNTU
Jordan Boyd-Graber
- Large language models
- GBEdinburgh
Bjorn Ross
- Language & society
- USGeorgia Tech
Mark Riedl
- Human-AI collaboration & trust
- CHETH Zurich
Mennatallah El-Assady
- Human-AI collaboration & trust
- CAToronto
Ebrahim Bagheri
- Knowledge, QA & retrieval
- SGNTU
Yewen Pu
- Reasoning & agents
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.
CMUcloses