Xin Chen

LLMs can be trained to negotiate like expert agents, extracting significantly higher surpluses by strategically exploring buyer markets rather than fixating on immediate bids.

VLMs struggle with raw medical data, achieving only a 48.6% success rate in standardization, revealing a critical gap in their clinical applicability.

Users can now generate and control high-quality videos in real-time using just their voice, revolutionizing interactive content creation.

Synthetic heart sounds generated by a diffusion model retain classification accuracy but miss critical abnormal acoustic features, highlighting a gap in current generation techniques.

Agricultural VLN agents struggle with instruction mistakes, showing a staggering 57% drop in success rates when faced with realistic errors.

Humanoid states, not low-level actions, are the key to unlocking text-driven control, enabling a diffusion model to generate more natural and semantically aligned behaviors.

Current AI models for liver fibrosis staging can match expert radiologists in some settings, but real-world clinical deployment is still hampered by data heterogeneity and label imbalance.

Forget centralized parameter servers: Totoro$^+$'s decentralized architecture lets you run massive federated learning applications simultaneously on edge networks, scaling gracefully and adapting to network churn.

Stop LLMs from drifting to English when reasoning in other languages: language-adaptive RL can guide them to stay consistent without sacrificing performance.

Open-source TTS gets a serious upgrade with Fish Audio S2, offering instruction-following control via natural language and production-ready streaming performance.