Publications

An open-source evaluation infrastructure that connects AI conversation trajectories with user-defined goals and outcome assessments.

A framework for evaluating whether autonomous agents recognize, communicate, and act appropriately when they should abstain.

An examination of how visual uncertainty cues shape trust calibration in LLM-assisted decision making.

A proposal for traceable records that support accountability investigations of large language model behavior.

A roadmap for moving LLM evaluation beyond static leaderboards toward naturalistic, functional assessments.

A functional evaluation approach for assessing large language model behavior across languages.

An analysis of the gaps and opportunities in the tools and infrastructure used to audit AI systems.

A survey of the AI auditing landscape and the structural limits that prevent it from delivering accountability on its own.