TiViBench：面向视频生成的视频内推理评测基准
TiViBench: Benchmarking Think-in-Video Reasoning for Video Generation
为图像到视频生成模型提供了一套分层推理评测协议（4 个维度、24 个任务、3 个难度），并给出无需额外训练或奖励模型的测试时偏好优化策略 VideoTPO，可直接迁移到其他生成模型的推理增强与选优环节。
原文依据 we propose TiViBench, a hierarchical benchmark specifically designed to evaluate the reasoning capabilities of image-to-video (I2V) generation models
摘要译文
由模型翻译 · 术语统一 · 请对照原文核对关键结论视频生成模型的快速演进，已使其关注点从生成视觉上合理的输出，转向处理需要物理合理性与逻辑一致性的任务。然而，尽管近期出现了 Veo 3 的帧链推理等突破，这些模型是否能展现出类似大语言模型（LLM）的推理能力仍不清楚。现有基准主要评估视觉保真度与时序一致性，无法刻画更高阶的推理能力。为弥补这一空白，我们提出 TiViBench，一个专门用于评估图像到视频（I2V）生成模型推理能力的层次化基准。TiViBench 系统地从四个维度评估推理能力：i) 结构推理与搜索，ii) 空间与视觉模式推理，iii) 符号与逻辑推理，iv) 动作规划与任务执行，覆盖 3 个难度等级下的 24 个多样化任务场景。通过大量评测，我们发现商业模型（如 Sora 2、Veo 3.1）展现出更强的推理潜力，而开源模型则显露出尚未被挖掘的潜力，其发挥仍受限于训练规模与数据多样性。为进一步释放这一潜力，我们提出 VideoTPO，一种受偏好优化启发、简单而有效的测试时策略。通过对生成的候选结果进行 LLM 自我分析以识别其优势与不足，VideoTPO 在无需额外训练、数据或奖励模型的情况下显著提升了推理性能。TiViBench 与 VideoTPO 共同为评估和推进视频生成模型的推理能力铺平了道路，并为这一新兴领域的未来研究奠定了基础。
英文原文摘要 1771 字符
The rapid evolution of video generative models has shifted their focus from producing visually plausible outputs to tackling tasks requiring physical plausibility and logical consistency. However, despite recent breakthroughs such as Veo 3's chain-of-frames reasoning, it remains unclear whether these models can exhibit reasoning capabilities similar to large language models (LLMs). Existing benchmarks predominantly evaluate visual fidelity and temporal coherence, failing to capture higher-order reasoning abilities. To bridge this gap, we propose **TiViBench**, a hierarchical benchmark specifically designed to evaluate the reasoning capabilities of image-to-video (I2V) generation models. TiViBench systematically assesses reasoning across four dimensions: i) **Structural Reasoning & Search**, ii) **Spatial & Visual Pattern Reasoning**, iii) **Symbolic & Logical Reasoning**, and iv) **Action Planning & Task Execution**, spanning 24 diverse task scenarios across 3 difficulty levels. Through extensive evaluations, we show that commercial models (*e.g.*, Sora 2, Veo 3.1) demonstrate stronger reasoning potential, while open-source models reveal untapped potential that remains hindered by limited training scale and data diversity. To further unlock this potential, we introduce **VideoTPO**, a simple yet effective test-time strategy inspired by preference optimization. By performing LLM self-analysis on generated candidates to identify strengths and weaknesses, VideoTPO significantly enhances reasoning performance without requiring additional training, data, or reward models. Together, TiViBench and VideoTPO pave the way for evaluating and advancing reasoning in video generation models, setting a foundation for future research in this emerging field.