Hang Zhang

Generating realistic 3D environments from satellite imagery in under 10 minutes could revolutionize how we visualize and interact with our planet.

Autonomous agents struggle to retain instructions when burdened with retrieving information from the open web, exposing a critical retrieval-reasoning trade-off.

Achieve open-world 3D segmentation without manual annotation by decoupling object discovery from semantic grounding.

VLMs often fail at spatial reasoning because they either ignore visual cues or exhibit unstable reasoning, but a novel process-shaping framework can fix this.

Seemingly strong NLI checkers can actually *hurt* medical RAG training by collapsing the RL gradient or triggering reward-hacking cascades like ultra-short answers and search avoidance.

Unlike naive approaches that cause flickering and visual artifacts, 4D-GSW embeds robust watermarks into dynamic 3D scenes by respecting the physics of motion.

Forget real-world video datasets: training VLMs on just 7.7K synthetic videos with temporal primitives beats 165K real-world examples, unlocking surprisingly effective transfer learning for video reasoning.

LLMs can escape the trap of converging on popular but incorrect answers in unsupervised RLVR by temporarily "unlearning" and exploring diverse response options.

Smart glasses powered by web-native AI agents can now outperform commercial solutions in assistive tasks, offering a practical path to always-on, context-aware help for users navigating daily life.

Autonomous driving gets a human-like reasoning boost: MindDriver uses progressive multimodal reasoning to bridge the gap between semantic understanding and physical trajectory planning.