Zhe Liu

Rapidly prototype and benchmark robotic navigation scenarios using simple YAML configurations, eliminating the coding barrier in simulation.

HarnessFix reveals that targeted, trace-guided repairs can significantly enhance LLM agent performance by addressing specific harness flaws rather than relying on broad modifications.

Conformal prediction for graph time series doesn't have to break down: by conditioning on low-frequency trends, you can restore exchangeability and get valid uncertainty estimates.

Flow-based VLAs can react to environmental changes ten times faster by adaptively prioritizing near-term actions during sampling, unlocking unprecedented real-time responsiveness.