Tsinghua AI

A frozen world-model planner's performance can be drastically improved by adapting its latent rollouts to controller-induced execution shifts, rather than directly substituting predicted trajectories.

Robots can now learn complex manipulation from single human videos by extracting object trajectories, bypassing the need for teleoperation and achieving near-perfect task success.

Consolidating diverse VLA skills into one policy is now achievable with a two-level on-policy distillation that leverages student-generated action flows and teacher velocity supervision, outperforming prior methods by up to 18.3% on complex tasks.

Ablation shock propels micro-LEDs with sub-micrometer precision and 50 $\unicode{x03BC}$m gap tolerance, shattering previous mass transfer bottlenecks.

Targeted data augmentation and joint RL/OPD training unlock significantly stronger long-horizon multimodal search capabilities for agents.

Auxiliary future-vision prediction targets, not just action prediction, unlock superior robot policy performance even when action supervision is halved.

A single GPU can now plan complex, contact-rich robotic manipulation trajectories for thousands of parallel worlds, achieving near-80% success on challenging in-hand rotation tasks with unprecedented speed.

Direct regression policies can match generative model performance in robot learning by learning input-dependent residual scales and robustly handling heavy-tailed action distributions.

Purely autoregressive VLAs no longer require awkward hybrid diffusion heads to achieve fine-grained robot control: structuring discrete action tokenization as causally annealed flow matching recovers continuous diffusion precision while preserving end-to-end token efficiency.

Mobile manipulation policies do not need joint action spaces to coordinate effectively: decoupling arm and base streams via masked joint attention and training-time world imagination drives state-of-the-art coordination across 17 distinct embodiments.

Frozen autonomous policies can survive out-of-distribution dynamics without retraining by turning streaming conformal prediction residuals directly into adaptive safety projection margins.

World action models do not need to render clean future video to generalize: conditioning on a single forward pass of pure noise captures virtually all the robustness benefits of multi-step diffusion at a fraction of the compute.

Generative world action models can run up to 3× faster in closed-loop control without fine-tuning or task degradation by dynamically repairing and reusing intermediate computational states instead of recomputing them from scratch.

Drawing predicted human gaze crosshairs directly onto input images more than doubles real-world bimanual VLA success (from 26.3% to 56.0%) without requiring runtime eye-tracking hardware or policy architecture changes.

Forcing moving wrist cameras to predict future pixels actively degrades policy performance—decoupling static scene reconstruction from dynamic wrist latent prediction yields up to a 12.8% boost in contact-rich manipulation with zero inference cost.

Diffusion teachers and online score networks are no longer bottlenecks for distilling autoregressive video models: directly minimizing sample MMD against reference videos improves VBench scores while unlocking 14B post-training on just eight H200 GPUs.

Today's video world models excel at hallucinating plausible-looking frames, but their ability to actually preserve initial object identities drops by over 40% as scene complexity scales beyond 40 instances.