4 papers
WorldGuide: Learning Success-Failure Boundaries in Latent World Models for Vision-Language-Action Policies
Lin Liu, Lu Zhang, Ziying Song +6
Latent world models offer a promising way to improve Vision-Language-Action policies by capturing the consequences of actions. However, models trained primarily on expert demonstra…
Cache-Aware Conv3D Lowering Across Embedded World-Model Decoders
Jiaming Zhang, Wu Yang, Shuai Tao +1
Generative world models can provide visual rollouts for embodied planning, yet their feasibility on edge devices depends not only on the learned model but also on how the execution…
STORM-Bench: Evaluating Online Video QA under Evolving and Incomplete Evidence
Siru Zhong, Shenghan Tan, Rihong Yan +6
Reliable online video question answering requires tracking state transitions while selectively abstaining when visual evidence is insufficient. Existing benchmarks focus on static…
PREM: Prefix-Steered Recurrent Memory for Long-Video Understanding
Siru Zhong, Qiongyan Wang, Xiaohui Lv +5
Long-video understanding must capture transient visual evidence under strict token budgets, yet existing methods compress frames, append memory tokens, or alter internal key-value…