Ran Xu

Web agents can now achieve state-of-the-art performance and robust test-time scaling by learning to self-verify their actions, bypassing the need for costly external judges during deployment.

Coding agents can now achieve up to 15% higher task completion rates by treating feedback as a persistent note that's followed until resolved, not just a one-off evaluation.

An on-policy expert-correction pipeline is developed, automated by a meta-level MLE agent, that localizes the failing turn in the weaker model's own rollout and asks the expert to rewrite only that turn, which preserves the model's planning style and combines the gains of harness evolution and model adaptation.

Evolving the harness of LLM agents can yield a 17-point performance boost without compromising their ability to generalize across tasks.