ImagerySearch: Adaptive Test-Time Search for Video Generation Beyond Semantic Dependency Constraints

TL;DR

ImagerySearch addresses the difficulty of generating videos from prompts with long-distance semantic relationships by introducing a prompt-guided adaptive test-time search that jointly tunes the inference search space and the reward function. It combines semantic-distance aware dynamic search space (SaDSS) with an adaptive imagery reward (AIR) to adapt generation dynamics to prompt semantics within diffusion-based video models, and introduces LDT-Bench for robust evaluation of imaginative prompts. The method demonstrates state-of-the-art or competitive improvements on LDT-Bench and VBench across multiple metrics and baselines, validating its effectiveness in handling long-distance semantics. The authors also provide a benchmark and code release to foster future research in imaginative video generation.

Abstract

Video generation models have achieved remarkable progress, particularly excelling in realistic scenarios; however, their performance degrades notably in imaginative scenarios. These prompts often involve rarely co-occurring concepts with long-distance semantic relationships, falling outside training distributions. Existing methods typically apply test-time scaling for improving video quality, but their fixed search spaces and static reward designs limit adaptability to imaginative scenarios. To fill this gap, we propose ImagerySearch, a prompt-guided adaptive test-time search strategy that dynamically adjusts both the inference search space and reward function according to semantic relationships in the prompt. This enables more coherent and visually plausible videos in challenging imaginative settings. To evaluate progress in this direction, we introduce LDT-Bench, the first dedicated benchmark for long-distance semantic prompts, consisting of 2,839 diverse concept pairs and an automated protocol for assessing creative generation capabilities. Extensive experiments show that ImagerySearch consistently outperforms strong video generation baselines and existing test-time scaling approaches on LDT-Bench, and achieves competitive improvements on VBench, demonstrating its effectiveness across diverse prompt types. We will release LDT-Bench and code to facilitate future research on imaginative video generation.

Figures (13)

• Figure 1: The motivation of ImagerySearch. The figure illustrates two semantic dependency scenarios related to camels. Left: The distance depicts the corresponding strength of prompt tokens during the denoising process. LDT-Bench consists of imaginative scenarios with long-distance semantics, whose semantic dependencies are typically weak. Right: Wan2.1 performs well on short-distance semantics but fails under long-distance. Test time scaling methods ($e.g.$, Video T1 liu2025videot1, Evosearch he2025evosearch) also struggle. However, ImagerySearch generates coherent, context-aware motions (orange box).
• Figure 2: Overview of our ImagerySearch. The prompt is scored by the Constrained Semantic Scorer (producing $\bar{\mathcal{D}}_{\text{sem}}$) and simultaneously fed to the T2V backbone (Wan2.1). At every step $t$ specified by the imagery scheduler, we sample a set of candidate clips, rank them with a reward function conditioned on $\bar{\mathcal{D}}_{\text{sem}}$, and retain only a $\bar{\mathcal{D}}_{\text{sem}}$-controlled subset. The loop repeats until generation completes.
• Figure 3: Overview of our LDT-Bench. Upper: (a) LDT-Bench is built by first extracting meta-information from existing recognition datasets; (b) GPT-4o is then used to generate candidate prompts, which are filtered jointly by DeepSeek and humans to obtain the final prompt set; (c) Additionally, we design a set of three MLLM-based QA tasks that serve as the creativity metric. Lower: (d) Compared with other benchmarks, LDT-Bench covers a much richer variety of categories; (e) its prompts also exhibit a semantic-distance distribution that is shifted toward substantially longer ranges. Note that "ASD" denotes the average semantic distance of prompts.
• Figure 4: Visualization of examples. Upper: Results from general models. Lower: ImagerySearch versus other test-time scaling methods. Ours produces more vivid actions under long-distance semantic prompts.
• Figure 5: (a) Effect of semantic distance across different models. As semantic distance increases, our method remains the most stable. (b-e) Our AIR consistently delivers superior performance. Scaling behavior of ImagerySeach and baselines as inference-time computation increases. From left to right, the $y$-axes represent the score changes for $\mathrm{MQ}$, $\mathrm{TA}$, $\mathrm{VQ}$, and Overall (VideoAlign liu2025videoalign). (f) Effect of reward weight.