SPIRAL: Self-Evolving Action-Conditioned Video Generation via Reflective Planning Agents
Fuente:
arXiv
Guardado en:
| Autores principales: | , , , , , , , , , , , , , |
|---|---|
| Formato: | Preprint |
| Publicado: |
2026
|
| Materias: | |
| Acceso en línea: | |
| Etiquetas: |
Agregar Etiqueta
Sin Etiquetas, Sea el primero en etiquetar este registro!
|
| _version_ | 1866911703991582720 |
|---|---|
| author | Yang, Yu Liao, Yue Mei, Jianbiao Wang, Baisen Yang, Xuemeng Wen, Licheng Zhang, Jiangning Li, Xiangtai Lv, Liang Chen, Hanlin Shi, Botian Liu, Yong Yan, Shuicheng Lee, Gim Hee |
| author_facet | Yang, Yu Liao, Yue Mei, Jianbiao Wang, Baisen Yang, Xuemeng Wen, Licheng Zhang, Jiangning Li, Xiangtai Lv, Liang Chen, Hanlin Shi, Botian Liu, Yong Yan, Shuicheng Lee, Gim Hee |
| contents | Long-horizon action-conditioned video generation aims to synthesize temporally coherent videos that follow complex action instructions over extended horizons, requiring procedural ordering, persistent action execution, and scene consistency beyond conventional TI2V's short-term fidelity. Existing single-shot video generation models typically operate in an open-loop manner, leading to incomplete action execution, hallucinated motions, and temporal drift. To address this, we propose SPIRAL, a closed-loop framework that performs sequential planning and iterative reflection for action-conditioned long-horizon video generation. Specifically, SPIRAL instantiates a think-act-reflect process: a PlanAgent decomposes high-level goals into sub-actions, which condition a VideoGenerator to synthesize each segment alongside a memory context, while a CriticAgent evaluates intermediate video segments to provide corrective feedback for iterative refinement. This closed-loop design further supports self-evolution by utilizing PlanAgent-proposed actions and CriticAgent-derived rewards for GRPO-based post-training to enhance the video generator's long-horizon consistency. Moreover, we introduce ActVideoGen-Dataset for task-specific training, and establish ActVideoGen-Bench as a dedicated evaluation suite for measuring action quality and temporal coherence. Experiments across multiple TI2V backbones alongside the self-evolving strategy show consistent gains on ActVideoGen-Bench and VBench, demonstrating the effectiveness of SPIRAL. |
| format | Preprint |
| id |
arxiv_https___arxiv_org_abs_2603_08403 |
| institution | arXiv |
| publishDate | 2026 |
| record_format | arxiv |
| spellingShingle | SPIRAL: Self-Evolving Action-Conditioned Video Generation via Reflective Planning Agents Yang, Yu Liao, Yue Mei, Jianbiao Wang, Baisen Yang, Xuemeng Wen, Licheng Zhang, Jiangning Li, Xiangtai Lv, Liang Chen, Hanlin Shi, Botian Liu, Yong Yan, Shuicheng Lee, Gim Hee Computer Vision and Pattern Recognition Long-horizon action-conditioned video generation aims to synthesize temporally coherent videos that follow complex action instructions over extended horizons, requiring procedural ordering, persistent action execution, and scene consistency beyond conventional TI2V's short-term fidelity. Existing single-shot video generation models typically operate in an open-loop manner, leading to incomplete action execution, hallucinated motions, and temporal drift. To address this, we propose SPIRAL, a closed-loop framework that performs sequential planning and iterative reflection for action-conditioned long-horizon video generation. Specifically, SPIRAL instantiates a think-act-reflect process: a PlanAgent decomposes high-level goals into sub-actions, which condition a VideoGenerator to synthesize each segment alongside a memory context, while a CriticAgent evaluates intermediate video segments to provide corrective feedback for iterative refinement. This closed-loop design further supports self-evolution by utilizing PlanAgent-proposed actions and CriticAgent-derived rewards for GRPO-based post-training to enhance the video generator's long-horizon consistency. Moreover, we introduce ActVideoGen-Dataset for task-specific training, and establish ActVideoGen-Bench as a dedicated evaluation suite for measuring action quality and temporal coherence. Experiments across multiple TI2V backbones alongside the self-evolving strategy show consistent gains on ActVideoGen-Bench and VBench, demonstrating the effectiveness of SPIRAL. |
| title | SPIRAL: Self-Evolving Action-Conditioned Video Generation via Reflective Planning Agents |
| topic | Computer Vision and Pattern Recognition |
| url | https://arxiv.org/abs/2603.08403 |