Episode
WorldReasonBench: Human-Aligned Stress Testing of Video Generators as Future World-State Predictors
- Podcast
- Daily Paper Cast
- Published
- May 13, 2026
- Duration seconds
- 1346
- Processing state
not_requested- Canonical source
- https://share.transistor.fm/s/60f7d5d5
Actions
POST https://stenobird.com/v1/public/podcasts/daily-paper-cast-7079649/episodes/worldreasonbench-human-aligned-stress-testing-of-video-generators-as-future-world-state-predictors/transcription-requests
Idempotently request low-priority transcript generation for this episode.GET https://stenobird.com/podcast/daily-paper-cast-7079649/worldreasonbench-human-aligned-stress-testing-of-video-generators-as-future-world-state-predictors.md
Read the agent-friendly Markdown representation of this episode resource.
Summary
🤗 Upvotes: 24 | cs.CV Authors: Keming Wu, Yijing Cui, Wenhan Xue, Qijie Wang, Xuan Luo, Zhiyuan Feng, Zuhao Yang, Sudong Wang, Sicong Jiang, Haowei Zhu, Zihan Wang, Ping Nie, Wenhu Chen, Bin Wang Title: WorldReasonBench: Human-Aligned Stress Testing of Video Generators as Future World-State Predictors Arxiv: http://arxiv.org/abs/2605.10434v1 Abstract: Commercial video generation systems such as Seedance2.0 and Veo3.1 have rapidly improved, strengthening the view that video generators may be evolving into "world simulators." Yet the community still lacks a benchmark that directly tests whether a model can reason about how an observed world should evolve over time. We introduce WorldReasonBench, which reframes video generation evaluation as world-state prediction: given an initial state and an action, can a model generate a future video whose state evolution remains physically, socially, logically, and informationally consistent? WorldReasonBench contains 436 curated test cases with structured ground-truth QA annotations spanning four reasoning dimensions and 22 subcategories. We evaluate generated videos with a human-aligned two-part methodology: Process-aware Reasoning Verification uses structured QA and reasoning-phase diagnostics to detect temporal and causal failures, while Multi-dimensional Quality Assessment scores reasoning quality, temporal consistency, and visual aesthetics for ranking and reward modeling. We further introduce WorldRewardBench, a preference benchmark with approximately 6K expert-annotated pairs over 1.4K videos, supporting pair-wise and point-wise reward-model evaluation. Across modern video generators, our results expose a persistent gap between visual plausibility and world reasoning: videos can look convincing while failing dynamics, causali…