Leon Chen

I am an AI researcher at Stanford working on data and evaluation for AI agents.

I am a CS PhD candidate in the Stanford AI Lab, advised by Serena Yeung-Levy, Ludwig Schmidt, and Tatsunori Hashimoto.

I contribute to the team projects Terminal-Bench, OpenThoughts, OpenThoughts-Agent, and Harbor-Index. Anthropic, OpenAI, and Google report Terminal-Bench scores when they launch new models.

In 2025, I was a research intern at Meta Superintelligence Labs, where I worked on test-time scaling for unified multimodal models.

Before Stanford, I worked with Ziwei Liu at MMLab and with Alan Yuille at Johns Hopkins.

Portrait of Leon Chen

Papers

2026 SepHarbor Adapters and Harbor-Index: infrastructure and a curated meta-dataset for large-scale agentic evaluation. (blog)
2026 JunOpenThoughts-Agent: data recipes for agentic models. (project)
2026 FebUniT: unified multimodal chain-of-thought test-time scaling.
2026 JanTerminal-Bench: benchmarking agents on hard, realistic tasks in command line interfaces. (leaderboard)
2025 JunOpenThoughts: data recipes for reasoning models. (blog)
2024 JulMMLongBench-Doc: benchmarking long-context document understanding with visualizations. (project)
2023 OctLarge language models are visual reasoning coordinators. (talk)
2023 MayOtter: a multi-modal model with in-context instruction tuning. (code)

Authors, venues, and all other papers are on the Papers page.

Talks

2026 NovTalk, Recursive.
2025 JulTalk, Meta Superintelligence Labs.
2023 DecTalk, Cohere For AI.

Service