110 citations · 420 across the 55 of their papers we have counts for
7 papers · 1 filter
TGuide: Faster Autoregressive Video Diffusion via Historical Trajectory Guidance
Zhuo Chen, Bin Zou, Fanyue Wei +3
Can autoregressive video diffusion be accelerated further when only four denoising steps remain? Existing feature caching approaches struggle in this regime, as large feature chang…
Ego2World: Compiling Egocentric Cooking Videos into Executable Worlds for Belief-State Planning
Qinchuan Cheng, Zhantao Gong, Pengzhan Sun +2
Egocentric videos capture how people carry out everyday activities, yet testing an agent requires evaluating the consequences of actions it chooses itself. We introduce Ego2World,…
From Reasoning Failures to Composable Video Spatial Intelligence
Pengzhan Sun, Junbin Xiao, Ramanathan Rajaraman +2
Spatial reasoning benchmarks evaluate vision-language models across diverse tasks, but task-level scores do not reveal which underlying capabilities account for success or failure.…
Rethinking Probability-Based Reinforcement Learning From Posterior Concentration
Shiu-Hong Kao, Yubo Zhao, Zhenyu Tian +3
Verifier-free reinforcement learning with probability-based rewards offers a promising way to train LLMs on general reasoning tasks where external verifiers are unavailable. Yet th…
EgoVine: Post-Training with Branching Experience Compiled from Real-World Videos
Qinchuan Cheng, Bowen Lyu, Zhantao Gong +3
Real-world egocentric videos capture multi-step household procedures, but their recorded trajectories leave alternative actions and recovery from the learner's own mistakes largely…
Aligning Thoughts with Answers: Probability Rewards to Tame Thinking Drift
Pengzhan Sun, Shiu-hong Kao, Shijie Li +4
This paper studies \textbf{thinking--answer consistency} in vision-language models. We focus on Visual Intention Grounding, where a model infers a target object based on a human in…
ReSTI: A Source-Grounded Audit and Repair of STI-Bench
Pengzhan Sun, Ramanathan Rajaraman, Shiu-Hong Kao +2
Spatial--temporal benchmarks are valid only when their questions, source annotations, and answer options identify the same physical quantity. We audit STI-Bench against the officia…