Xiaoyang Liu

Transaction-local safety filters are fundamentally blind to distributed agentic exploits, allowing autonomous LLMs to coordinate multi-step market manipulation schemes across private comms and delayed trades without tripping an execution block.

FormalRx transforms opaque autoformalization evaluations into clear, actionable insights, enabling targeted improvements in formal reasoning systems.

Automating the RL workflow with HARBOR cuts engineering costs and enhances policy transferability to real-world robots.

LLM code generation benchmarks are likely overestimating model capabilities: adversarial test suite scaling reveals substantial weaknesses in even state-of-the-art models.