I’m Matt Wilde. At Forum AI, I work on translating expert judgment into reliable AI evaluations, combining research on what to measure with engineering to make those evaluations useful.
Recent research:
- NewsBench: Expert-Grounded Evaluation of Epistemic Quality in AI News Reporting (first author): evaluating source quality, factuality, and neutrality in AI news reporting.
- Distilling Expert Judgment at Scale (coauthor): a methodology for translating expert judgment into automated evaluation systems.
Selected personal tools:
- gdocs-cli: a command-line tool for working with Google Docs.
- mcp_pdf_forms: tools for discovering PDF files, extracting form fields, and visualizing their layout.
Previously, I worked at Meta AI.