Yizhou Wang

Explicitly inferring mental states under uncertainty leads to more thoughtful dialogue than unrestricted information access, challenging conventional wisdom in AI interactions.

Mimicking clinical diagnostic thinking with counterfactual reasoning boosts medical video diagnosis accuracy by up to 10.2%.

Hallucinated citations and logical inconsistencies in AI-generated scientific introductions can be significantly reduced by explicitly modeling and optimizing a logic-reasoning graph.

A hierarchical agent that separates visual and textual contexts drastically improves multi-step reasoning on complex charts, outperforming monolithic MLLMs.

Embodied agents can now exhibit coherent, long-horizon, self-directed behavior by reasoning about abstract value trade-offs, a capability previously absent in instruction-following or needs-driven approaches.

Achieve zero-shot cross-embodiment visual tracking by dynamically adapting control policies to inferred embodiment constraints, eliminating the need for per-robot training.

Forget generic assistants – EgoSelf learns your habits from your first-person view data to predict your future interactions.

Video LLMs don't just get details wrong, they fundamentally distort motion and fabricate entire events, demanding a new approach to evaluation and mitigation.

Imagine populating any 3D environment with digital humans that spontaneously navigate and interact, driven only by visual input and goals.

Forget hand-crafted heuristics: this new dynamics-aware policy learns to exploit contact forces in cluttered environments, outperforming traditional methods by 25% in simulation and showing impressive sim-to-real transfer.

MLLMs that ace standard Referring Expression Comprehension benchmarks still stumble when faced with images designed to eliminate shortcuts, revealing a surprising lack of robust visual reasoning.