Research
Benchmarking Reward Hack Detection in Code Environments via Contrastive Analysis [ICML 2026, main conference]
A Unified Definition of Hallucination: It's The World Model, Stupid! [ICML 2026, position paper]
DETOUR: An Interactive Benchmark for Dual-Agent Search and Reasoning [GEM Workshop at ACL 2026]
FigmaTrace: Capturing Creative Nuances in Human Figma Design Workflows [2026, open dataset release]
HalluWorld: A Controlled Benchmark for Hallucination via Reference World Models [2026, preprint]
SpeedrunBench: Challenging LLM Agents with Video Game Speedrunning [2026, under review at ICLR 2027]
MEMTRACK: Evaluating Long-Term Memory and State Tracking in Multi-Platform Dynamic Agent Environments [NeurIPS 2025, SEA Workshop]
Browsing Lost Unformed Recollections: A Benchmark for Tip-of-the-Tongue Search and Reasoning [ACL 2025, main conference]
TRAIL: Trace Reasoning and Agentic Issue Localization [2025]
Glider: Grading LLM Interactions and Decisions using Explainable Ranking [2024]
Lynx: An Open Source Hallucination Evaluation Model [2024]
FinanceBench: A New Benchmark for Financial Question Answering [2023]
SimpleSafetyTests: A Test Suite for Identifying Critical Safety Risks in Large Language Models [2023]
Step by Step to Fairness: Attributing Societal Bias in Tasems [2023]
Perturbation Augmentation for Fairer NLP [2022]
Many Episode Learning in a Modular Embodied Agent via End-to-End Interaction [2022]