发表机构
National Taiwan University; Appier Inc.; Cornell University（国立台湾大学; Appier 公司; 康奈尔大学）
机构由 AI 辅助整理，请以论文原文为准。AI 中文总结
针对视频生成模型多样且成本各异的问题，提出个性化路由方法TasteRoute，根据请求、用户偏好和预算选择模型，在保持偏好路由竞争力的同时降低成本，并发布数据集TasteRoute-3k。
AI 中文摘要
视频生成的快速发展催生了大量在能力和生成成本上存在显著差异的模型。这引出了一个自然的问题：能否将每个请求高效地路由到合适的模型？我们发现，即使将其他标注者的共识作为预言机，它也只有34-55%的时间与每位标注者自己的偏好一致。受此观察启发，我们引入了TasteRoute，一种个性化的视频生成路由器，它基于输入请求、用户偏好和可用生成预算联合选择生成器。在文本到视频和图像到视频的设置中，TasteRoute在偏好路由方面与强大的简单基线相当，同时降低了平均生成成本。成本节省在更高的预算上限下增加。最后，我们发布了TasteRoute-3k，一个包含多模型视频比较、质量判断、偏好排名和用户画像信号的人工标注数据集，以促进个性化且成本感知的视频路由的未来研究。
英文摘要
Rapid progress in video generation has led to a plethora of models that differ substantially in capability and generation cost. This raises a natural question: can each request be efficiently routed to an appropriate model? We find that even when the consensus of the other annotators is used as an oracle, it agrees with each annotator's own favorite only 34-55% of the time. Motivated by this observation, we introduce TasteRoute, a personalized video-generation router that selects a generator jointly based on the input request, user preferences, and available generation budget. Across text-to-video and image-to-video settings, TasteRoute is competitive with strong simple baselines on preference routing while reducing average generation cost. The cost saving increases under higher budget caps. Finally, we release TasteRoute-3k, a human-annotated dataset containing multi-model video comparisons, quality judgments, preference rankings, and user-profile signals to facilitate future research on personalized and cost-aware video routing.
