Rui Liu

Even state-of-the-art streaming VLMs miss crucial fast-paced events in real-world video, scoring below 51% on a new high-dynamic benchmark.

Input-blind controls can account for substantial oracle headroom in adaptive computation, challenging the assumption that observed gains directly reflect the efficacy of selected layer programs.

Accelerating VLA models without sacrificing task success is now certifiable, guaranteeing over $85\%$ of reference performance with $10\times$ speedups.

Achieve 10x lower quantization errors on MCUs with 16-bit precision, matching or beating 8-bit inference speed and energy costs.

Task-specific LoRA-Experts and prototype-ensemble matching unlock rehearsal-free class-incremental learning with reduced interference and improved discrimination.

Treating automated feature transformations as permutation-invariant hierarchies rather than ordered sequences eliminates a fundamental representation bias, enabling policy-guided RL to efficiently navigate non-convex search spaces.

Learning from real-world environments follows a precise log-sigmoid scaling law, with agent performance and learning speed improving dramatically over time.