Zihan Wang

LLM agents trained with MCGPO achieve up to 5.16% higher success rates by intelligently reusing rollout data and constructing context-aware action preferences.

Robotic agents can now learn complex manipulation strategies from just a few demonstrations by treating them as evidence to be analyzed, not just training data.

Massive 100K-node routing problems can be solved in tens of seconds without learned heuristics simply by optimizing a compressed surrogate space to seed downstream solvers.

Economic interactions among agents can spontaneously generate sophisticated reasoning strategies, outperforming traditional monolithic approaches.