Share your thoughts, 1 month free Claude Pro on usSee more
WorkDL logo mark

Temporal Backtracking Search for Test-time Generative Video Reasoning

About

While test-time scaling has revolutionized reasoning in large language models, generative video reasoning remains bottlenecked by a single-shot paradigm. We demonstrate that searching over denoising steps cannot rescue logically flawed rollouts because spatial trajectories commit early in the diffusion process. Root-level Best-of-N (BoN) sampling is similarly inefficient: reasoning errors cluster early in the temporal axis, and resampling blindly discards verified upstream progress. To unlock effective test-time scaling for video models, we introduce Temporal Backtracking Search (TBS), which shifts the search space to the temporal axis. TBS transforms video generation into an iterative generate-verify-restart loop via three core mechanisms: (1) variable-K conditioning to resume generation from arbitrary clean prefixes; (2) temporal process verification to localize failures and extract valid restart anchors; and (3) prefix-based search to reallocate compute toward extending correct trajectories rather than root resampling. Across algorithmic, navigation, and robotics domains, TBS Pareto-dominates matched-budget BoN. In a strict out-of-distribution setting where one-shot generation collapses (0.7% for BoN), TBS achieves 22.7%, with every solved episode stemming from a restarted branch. Ultimately, TBS reveals that the local reasoning competence of video models far exceeds what single-shot rollouts indicate, providing a scalable test-time framework to unlock it.

Sejoon Jun, Zheng Ding, Huangyuan Su, Weirui Ye, Yilun Du• 2026

Related benchmarks

TaskDatasetResultRank
Structured video reasoningMaze Overall
Exact Match Percentage90.3
6
Structured video reasoningMaze Med
Exact Match Rate99.5
6
Structured video reasoningMaze Hard
Exact Match Rate81
6
Structured video reasoningMaze3D Overall
Exact Match (%)97.7
6
Structured video reasoningMaze3D Med.
Exact Match Rate100
6
Structured video reasoningMaze3D Hard
Exact Match95.3
6
Structured video reasoningSokoban Overall
Exact Match Rate82.7
6
Structured video reasoningSokoban Easy
Exact Match (%)99
6
Structured video reasoningSokoban Hard
Exact Match79.7
6
Structured video reasoningSliding Puzzle Overall
Exact Match (%)86.5
6
Showing 10 of 24 rows

Other info

Follow for update