About Me

I am a Senior Research Scientist on the Alignment team at FAIR, Meta. I work with Dr. Jason Weston on Self- and Co-Improving (with humans) AI. I also collaborate with the TBD lab at Meta for self-improving Muse family of models, achieving SOTA on various professional tasks.

Previously, I obtained a PhD in Computer Science from the University of North Carolina at Chapel Hill, advised by Prof. Mohit Bansal. My PhD was supported by a Google PhD Fellowship and a Rebecca and Munroe Cobey Fellowship. A list of my publications (grouped by topics) can be found here:

Recent News

• June 2026: Autodata is out on arXiv.
• Jan 2026: Three papers accepted to ICLR 2026: J1, OptimalThinkingBench and Hybrid Reinforcement.
• Oct 2025: Three new pre-prints on better post-training methods: Hybrid Reinforcement, Self-aggregation, and Improving Diversity.
• Aug 2025: Bridging Offline and Online RL and OptimalThinkingBench are out on arXiv.
• May 2025: J1, new paper on RL recipes for training LLM-as-a-Judge, is now on arXiv.
• May 2025: Organizing the RAM2 workshop, co-located with COLM 2025. Submit your papers soon!
• May 2025: EvalPlanner is accepted to ICML 2025.
• January 2025: EvalPlanner, my first work out of FAIR, is now on arXiv.
• January 2025: System-1.x is accepted to ICLR 2025.
• August 2024: Joined FAIR at Meta as a Research Scientist.
• July 2024: System-1.x, my final PhD paper, is out on arXiv.
• May 2024: ReConcile is accepted to ACL 2024.
• May 2024: MAGDi is accepted to ICML 2024.
• April 2024: Defended a thesis on “Multi-step Reasoning over Natural Language”. Check it out here.
• March 2024: Branch-Solve-Merge is accepted to NAACL 2024.
• February 2024: New pre-print on Structured Distillation from Multi-agent Interaction Graphs.