Maarten Sap

PluriHarms: Benchmarking the Full Spectrum of Human Judgments on AI Harm

OpenAgentSafety: A Comprehensive Framework For Evaluating Real-World AI Agent Safety

Ambig-SWE: Interactive Agents to Overcome Underspecificity in Software Engineering

Fortifying Hallucination Detection to Out-of-Domain Data

TOM-SWE: User Mental Modeling For Software Engineering Agents

Social World Models: Universal Structured Representations for Social Reasoning

HypoVeil: A Hypothesis-Driven Pragmatic Inference-Time Control Framework for Privacy–Utility-Aware LLM-Agent Dialogue

The PIMMUR Principles: Ensuring Validity in Collective Behavior of LLM Societies

EVALUESTEER: Measuring Reward Model Steerability Towards Values and Preferences

The Delta Learning Hypothesis: Preference Tuning on Weak Data can Yield Strong Gains

Fluid Language Model Benchmarking

ALFA: Aligning LLMs to Ask Good Questions A Case Study in Clinical Reasoning

Martingale Score: An Unsupervised Metric for Bayesian Rationality in LLM Reasoning

PolyGuard: A Multilingual Safety Moderation Tool for 17 Languages

HAICOSYSTEM: An Ecosystem for Sandboxing Safety Risks in Interactive AI Agents

HAICOSYSTEM: An Ecosystem for Sandboxing Safety Risks in Human-AI Interactions

On the Resilience of LLM-Based Multi-Agent Collaboration with Faulty Agents

SafetyAnalyst: Interpretable, Transparent, and Steerable Safety Moderation for AI Behavior

On the Resilience of Multi-Agent Systems with Malicious Agents

BIG5-CHAT: Shaping LLM Personalities Through Training on Human-Grounded Data

SafetyAnalyst: Interpretable, transparent, and steerable LLM safety moderation