Open science
Research code, benchmarks and systems released by the lab so others can reproduce and build on our work. More on GitHub.
A web platform that turns a plain-English request into a real Fermi-LAT gamma-ray analysis: it writes the FermiPy configuration and Python driver, checks them, runs exactly what the scientist approves and shows every step. Every model, rule, validator and data source is a pluggable module.
A small, auditable qualification test for language models used as workers in human-computation pipelines: 60 yes/no items across five domains with independently audited gold labels, plus the evaluation harness.
Complete code, data, prompts, results and analysis pipeline for the multi-phase study of metacognitive calibration and sycophancy in LLM self-assessment.
An AI system that uses Socratic dialogue and multi-perspective analysis to help people examine personally held beliefs.
Code for the CSCW paper on a culturally aware workplace tool that adapts to monochronic and polychronic work styles of crowd workers.
Dataset generation from arXiv mathematics, embedding-space experiments, automated judge agents and prompt experimentation for the NSF AIMing project.
A cross-platform tool for systematic literature reviews following the PRISMA methodology, with Semantic Scholar integration; one codebase for CLI, web, desktop and mobile.
Embedded image capture and on-device object detection for privacy- preserving analysis of workplace activity (RK3588 edge device, HDMI capture).
A Unity-based extended-reality workspace for adding and arranging virtual monitors, with a screencast server.
A web interface to generate Personal AI Assistant (PAIA) schemas.