Paper 2: Prediction-Powered Inference Across Many Tasks for AI Evaluation and Social Science Research
We develop methods for statistically valid inference from many related tasks, with many cheap proxy labels (via, e.g., an LM) and only a few high-quality labels per task. 🧵
Assistant Professor of OR/Stat at MIT. Past: postdoc UC Berkley, PhD Harvard. Interested in incentive-aware ML, bandits, mechanism design & public policy. 🏃♀️
- Paper 1: Conformal Language Modeling via Posterior Sampling How can we control hallucination risk in LLM generations without sacrificing the usefulness and coherence of the response? In our new paper, we use conformal methods to calibrate the generation process itself. 🧵
- We've had some exciting papers coming out of our lab this past Spring, and 2 of them are going to be presented to workshops at #ICML26 this weekend! Next few tweets are going to be devoted to these papers.
- Rad and I are co-organizing the EC mentoring workshop this year, and we're so excited about the fantastic speakers that are joining us! Please spread the word to your students. The registration form will remain open until tomorrow EOD.📣 The EC 2026 Mentoring Workshop is happening virtually on Thursday, June 18, 10am–4:45pm EST, as part of EC's virtual Preview Week! Organized by Rad Niazadeh & Chara Podimata. Free, but registration required by June 15. 🧵
- 📢 Asu Ozdaglar and I are hiring a postdoc for next year to work on adaptive, ethically-informed decision systems, human–AI collaboration, and decision-making under uncertainty. More info here: charapodimata.com/postdoc_announ….


