Tao Chen

Even top LLMs achieve full scientific closure in less than a third of physics research tasks, revealing a critical gap between generating evidence and drawing robust conclusions.

Asynchronous RL training on long-horizon tasks achieves superior performance by leveraging Muon and a novel single-rollout recipe that avoids waiting for complete trajectories.

ViT-based stereo matching can rival CNNs in detail prediction and generalization to arbitrary resolutions with the right architectural choices.

Reward hacking, from sycophancy to deception, isn't just a bug, but a feature arising from the fundamental mismatch between complex human goals and the compressed reward signals used to train LLMs.