Temporal Backtracking Search for Test-time Generative Video Reasoning

• URL: http://arxiv.org/abs/2606.13861v1
• Date: Thu, 11 Jun 2026 19:41:18 GMT
• Title: Temporal Backtracking Search for Test-time Generative Video Reasoning
• Abstract summary: generative video reasoning remains bottlenecked by a single-shot paradigm.<n>We introduce Temporal Backtracking Search (TBS), which shifts the search space to the temporal axis.<n>In a strict out-of-distribution setting, TBS achieves 22.7%, with every solved episode stemming from a restarted branch.
• Score: 51.4821857044108
• License: http://arxiv.org/licenses/nonexclusive-distrib/1.0/
• Abstract: 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.

Related papers

• Flash-BoN: Instant Drafts for Inference-Time Scaling in Diffusion Models [91.19049123903137]
  Inference-time scaling for text-to-image generation has progressed from simple Best-of-$N$ sampling to guided search methods.<n>These approaches focus on when and how often to verify during denoising but largely treat the cost of generation itself as fixed.<n>We show that under wall-clock evaluation, simple BoN already matches or outperforms several guided search techniques.
  arXiv  Detail & Related papers  (2026-07-05T19:02:32Z)
• Efficient and Trainable Language Model Test-Time Scaling via Local Branch Routing [65.59464371673623]
  Local Branch Routing (LBR) is a token-level test-time scaling framework.<n>It expands a small local lookahead tree, forwards all sampled branches through the language model, and uses a lightweight router to select the depth-1 subtree to commit.<n>LBR allows each token decision to use evidence beyond the root next-token distribution while avoiding full solution-level search.
  arXiv  Detail & Related papers  (2026-06-24T03:42:44Z)
• ScalingAttention: Discovering Intrinsic Sparse Attention Topology for Video Diffusion Transformers [7.879917783645986]
  Diffusion Transformers (DiTs) have revolutionized high-fidelity video generation, their reliance on 3D full attention creates a quadratic computational bottleneck.<n>Existing sparse methods face a dilemma: dynamic pruning suffers from prohibitive runtime overhead and memory fragmentation, while statics fail to capture fine-grained dependencies.<n>In this work, we propose ScalingAttention, a training-free framework grounded in a key inductive bias.<n> Experiments on Wan2.1 show up to 1.90X end-to-end speedup with superior fidelity, establishing a new frontier over state-of-the-art baselines.
  arXiv  Detail & Related papers  (2026-06-22T08:32:07Z)
• Detecting AI-Generated Videos with Spiking Neural Networks [26.67301552503132]
  We present MAST, a detector that processes multi-channel temporal residuals with a spike-driven temporal branch alongside a frozen semantic encoder for cross-generator evaluation.<n>On the GenVideo benchmark, MAST achieves 93.14% mean accuracy across 10 unseen generators under strict cross-generator evaluation, matching or surpassing the strongest ANN-based detectors.
  arXiv  Detail & Related papers  (2026-05-07T09:08:32Z)
• Stream-T1: Test-Time Scaling for Streaming Video Generation [30.575612184373025]
  Stream-T1 is a pioneering comprehensive Test-Time Scaling framework exclusively tailored for streaming video generation.<n> evaluated on both 5s and 30s comprehensive video benchmarks, Stream-T1 demonstrates profound superiority.
  arXiv  Detail & Related papers  (2026-05-06T03:40:05Z)
• LatSearch: Latent Reward-Guided Search for Faster Inference-Time Scaling in Video Diffusion [87.42285185305813]
  We introduce a novel inference-time search mechanism that performs Reward-Guided Resampling and Pruning.<n>LatSearch consistently improves video generation across multiple evaluation dimensions compared to the baseline Wan2.1 model.
  arXiv  Detail & Related papers  (2026-03-15T18:07:29Z)
• Test-Time Scaling with Diffusion Language Models via Reward-Guided Stitching [66.39914384073145]
  We propose a self-consistency framework that turns cheap diffusion-sampled reasoning into a reusable pool of step-level candidates.<n>We find that step-level recombination is most beneficial on harder problems.<n>Our training-free framework improves average accuracy by up to 2 across six math and coding tasks.
  arXiv  Detail & Related papers  (2026-02-26T11:08:39Z)
• A Unified Reasoning Framework for Holistic Zero-Shot Video Anomaly Analysis [64.42659342276117]
  Most video-anomaly research stops at frame-wise detection, offering little insight into why an event is abnormal.<n>Recent video anomaly localization and video anomaly understanding methods improve explainability but remain data-dependent and task-specific.<n>We propose a unified reasoning framework that bridges the gap between temporal detection, spatial localization, and textual explanation.
  arXiv  Detail & Related papers  (2025-11-02T14:49:08Z)
• TAR-TVG: Enhancing VLMs with Timestamp Anchor-Constrained Reasoning for Temporal Video Grounding [28.79516973256083]
  Temporal Video Grounding aims to precisely localize video segments corresponding to natural language queries.<n>We propose Timestamp Anchor-constrained Reasoning for Temporal Video Grounding (TAR-TVG)<n>TAR-TVG introduces timestamp anchors within the reasoning process to enforce explicit supervision to the thought content.
  arXiv  Detail & Related papers  (2025-08-11T06:59:32Z)
• RoHM: Robust Human Motion Reconstruction via Diffusion [58.63706638272891]
  RoHM is an approach for robust 3D human motion reconstruction from monocular RGB(-D) videos. It conditioned on noisy and occluded input data, reconstructs complete, plausible motions in consistent global coordinates. Our method outperforms state-of-the-art approaches qualitatively and quantitatively, while being faster at test time.
  arXiv  Detail & Related papers  (2024-01-16T18:57:50Z)

This list is automatically generated from the titles and abstracts of the papers in this site.

This site does not guarantee the quality of this site (including all information) and is not responsible for any consequences.