Haotian Wang

Existing models mismanage tool use, but Beacon achieves a balance that enhances performance on complex tasks while preserving accuracy on simpler ones.

RefCaptioner not only outperforms existing models in video captioning but also enables precise grounding of visual elements to multiple reference images, enhancing factual accuracy.

Current video generation models face a critical trade-off between faithfully executing keyframes and producing natural-looking videos, with performance degrading under increased keyframe density.

DRIFT not only sets a new state-of-the-art in self-improvement for language models but also redefines how we can dynamically adapt learning strategies based on problem difficulty.

PLAN-S achieves a 42% reduction in collision rates while providing diverse driving style adaptations through a novel cost map approach.

Current audio-visual generation models struggle to maintain coherence and alignment when scaling to minute-long content, a problem exposed by the new LongAV-Compass benchmark.

Achieve 9.97% higher accuracy in cross-domain human activity recognition while simultaneously reducing computation by 6.4x with a new sensor data tokenization and attention mechanism.

Forget tedious human-in-the-loop DAgger – WM-DAgger uses World Models to synthesize corrective actions for robots, achieving impressive 93% success with just 5 demos.

Multi-resolution decomposition and diffusion models can boost time series forecasting accuracy by up to 10% over existing methods.

Forcing networks to perform well under varying sparsity constraints during training can surprisingly boost generalization, outperforming standard dense training.