Xiang-Xiang Chu

Real-world professional tasks demand agents that learn from exploration and feedback, a capability unlocked by a novel engine that reconstructs environments to force agents through the full problem-solving lifecycle.

Stateagent transforms video generation by effectively tracking and inferring world states, leading to a 54% improvement in coherence across segments.

Achieving synchronized audio-video generation at 2K resolution with a compact 7B model could revolutionize content creation and accessibility in multimedia applications.

Self-summarization in LLMs can enhance reasoning coherence and reduce context exhaustion, leading to a 4% performance boost with shorter rollouts.

Fine-grained decision-making in agentic RL can boost performance by nearly 4 points, challenging the reliance on coarse heuristics.

Bootstrapping LLM agents to co-evolve as both agent and environment can lead to significant performance gains, with an average improvement of over 4% on complex tasks.

Forget maps: LLMs can learn end-to-end transit route planning directly from data, even grounding GPS coordinates without explicit mapping.

LLM agents can now learn from *everyone's* experience, not just their own, leading to system-wide improvements without requiring additional user effort.