news

Aug 2026 New preprint: PACT: Can Enterprise AI Assistants Be Trusted Under Pressure? A novel benchmark testing whether AI assistants follow enterprise rules and regulations despite pressures. Check out the website, paper, and code!
Jul 2026 Why Do AI Agents Break Rules? How Framing, Context, and Social Signals Shape Compliance accepted to AIES 2026 (AAAI/ACM Conference on AI, Ethics, and Society)!
Jul 2026 Where Does Social Reasoning Come From? Capability Provenance in Language Models accepted to Conference on Language Modeling (COLM) 2026! This was a collaboration between Mark Riedl’s lab and Eleuther AI on tracing social reasoning capabilities in LLMs to specific pretraining data sources using gradient-based attribution.
Mar 2026 Two papers accepted to the Human-Centered Explainable AI (HCXAI) 2026 workshop at CHI: Explainable Model Routing for Agentic Workflows and Counterfactual Explanations for Agentic Workflows as spotlight presentations!
May 2025 FLaME (Holistic Finance Language Model Evaluation) accepted to ACL Findings 2025!
Apr 2025 Trust by Design: Skill Profiles for Transparent, Cost-Aware LLM Routing (BELLA) accepted as a poster at MLSys YPS 2025!
Mar 2025 Excited to be joining Two Sigma Investments as a Software Engineering Intern on Modeling Insights this summer (2025)!