RESEARCH

Human Behavior research

Peer-reviewed work from the team, and the parts of it that changed how the product is built. Each paper is reported at the level of detail we would want if we were reading it — the objective, the constraints, the tables, and the ablations that did not go the way we expected.

NeurIPS 2025

Multi-level Evaluation with LLM-based Integrated Self-Scrutiny and Auditing

Amirhossein Afsharrad, Sri Jaladi, Nima Yazdani, Ali Ansari, Seyed Shahabeddin Mousavi, Sanjay Lall

Hierarchical, calibration-first evaluation for LLM judges. Decompose a conversation into levels, learn a bias correction from human labels, and turn a weak model into a reliable judge — no fine-tuning. Bias correction alone cuts error by over 50%.