Test-Time Scaling for Video Diffusion Models via Diagnosis-Guided Candidate Recycling
ACM Transactions on Graphics · 09.11.2026 · 45 (6) · 1-12
DOI: 10.1145/3842526 ↗Kurz gefasst
Fachartikel in der Zeitschrift ACM Transactions on Graphics, online erschienen am 09.11.2026. Verfasst von Hangzhou He und 6 weiteren Autor:innen. Die Arbeit nennt 58 Literaturangaben. Der Volltext ist frei zugänglich (Open Access, CC BY 4.0). Gefördert von: National Key Research and Development Program of China, National Natural Science Foundation of China, Science Foundation of Peking University Cancer Hospital.
Abstract (Originalsprache)
Recent video diffusion models have achieved remarkable generation quality, but high-fidelity results still largely depend on closed-source systems or costly large-scale infrastructure. Test-time scaling (TTS) offers a training-free way to improve lightweight generators by spending additional inference compute, yet existing methods mostly remain within a noise-search paradigm: they sample, select, or perturb denoising trajectories and discard low-scoring candidates after expensive generation. This generate-and-discard process wastes not only computation but also the partial motion, layout, or appearance structure already encoded in recoverable samples. We present GEARS ( G uided E diting for A daptive R ecycling S earch), a training-free framework that introduces diagnosis-guided candidate recycling into video TTS by turning such candidates into editable priors through a generation-evaluation-editing loop. GEARS consists of two collaborative components. The Stage-Aware Scheduler determines what to repair, when to repair it, and which candidates should be preserved, recycled, or discarded. The Candidate Recycler diagnoses recoverable failures from keyframes and multi-dimensional reward feedback, derives candidate-specific repair prompts, and repairs the corresponding candidates through manifold-aware latent SDEdit. The repaired candidates are recycled into the search pool, creating refinement paths beyond standard noise perturbation while preserving useful structure. Under matched NFE budgets, GEARS consistently outperforms existing video TTS methods on VBench, bringing a 1.3B model to a total score comparable to a 14B counterpart, and ablations verify the necessity of adaptive scheduling, diagnosis-conditioned editing, and manifold-aware re-denoising. Code is available at https://github.com/ATH-MaaS/Guided_Editing_for_Adaptive_Recycling_Search.
Förderung
- National Key Research and Development Program of China
- National Natural Science Foundation of China
- Science Foundation of Peking University Cancer Hospital
Zitieren
APA
He, H., Duan, L., Zhao, S., Li, K., Chen, Q., Luo, W., & Lu, Y. (2026). Test-Time Scaling for Video Diffusion Models via Diagnosis-Guided Candidate Recycling. ACM Transactions on Graphics, 45(6), 1-12. https://doi.org/10.1145/3842526
Vancouver
He H, Duan L, Zhao S, Li K, Chen Q, Luo W, et al. Test-Time Scaling for Video Diffusion Models via Diagnosis-Guided Candidate Recycling. ACM Transactions on Graphics. 2026;45(6):1-12. doi:10.1145/3842526
BibTeX
@article{He2026,
author = {He, Hangzhou and Duan, Lunhao and Zhao, Shanshan and Li, Kaiwen and Chen, Qing-Guo and Luo, Weihua and Lu, Yanye},
title = {Test-Time Scaling for Video Diffusion Models via Diagnosis-Guided Candidate Recycling},
journal = {ACM Transactions on Graphics},
year = {2026},
volume = {45},
number = {6},
pages = {1--12},
doi = {10.1145/3842526}
}
EUVIMED