Jacky Kwok

I am a third-year PhD student in Computer Science at Stanford University. My research focuses on developing scalable verifiers for agents (with Prof. Azalia Mirhoseini and Ion Stoica) and robots (with Prof. Marco Pavone and Chelsea Finn). Recent works include LLM-as-a-Verifier, RoboMonkey, and CoVer-VLA.

Before joining Stanford, I received my Bachelor's and Master's degrees in Computer Science at UC Berkeley, where I conducted research at the Sky Computing Lab, Berkeley AI Research (BAIR), and iCyPhy Center. My master's thesis focused on efficient and reliable systems for reinforcement learning and robotics under the guidance of Ion Stoica and Edward Ashford Lee.

Email  /  LinkedIn  /  X  /  Projects  /  Google Scholar

Selected Publications

A general-purpose verification framework that provides fine-grained feedback by scaling scoring granularity, repeated verification, and criteria decompositions. SOTA on Terminal-Bench, SWE-Bench, and other agentic benchmarks.

Introduces a contrastive-based verifier for vision–language–action alignment and a hierarchical test-time verification pipeline that couples high-level prompt optimization with low-level action chunk selection for VLAs

Proposes an action preference learning framework for VLAs and characterizes embodied test-time scaling laws

Master's Thesis on efficient and reliable systems for reinforcement learning and robotics

A high-performance middleware for robots, achieving up to ~100x lower latency than ROS2

A system to run, manage, and scale AI workloads on any AI infrastructure

Miscellanea

Template