Xuelong Li

Explicitly estimating wind disturbance acceleration enables quadrotors to navigate cluttered environments with unprecedented robustness, achieving 90% success in real-world fan-induced flight tests.

Uncertainty DMD is proposed, a simple uncertainty-injection framework that restores stochasticity at two key stages of AR generation: a timestep perturbation for the first chunk to increase first-chunk diversity, and a stochastic cache-writing mechanism for later chunks to preserve uncertainty in autoregressive conditioning.

Unified multimodal models can bootstrap both visual generation and perceptual understanding purely from text prompts by treating their own generative flow trajectories as self-supervised grounding signals.

RefVideo-6M revolutionizes video editing datasets by providing 5 million high-quality editing samples that prioritize visual references over flawed automatic edits.

A2I-Set enables AudioCanvas to outperform existing methods, achieving unprecedented visual expressiveness and alignment in audio-to-image generation.