VideoSearch-R1: Iterative Video Retrieval and Reasoning via Soft Query Refinement
Summary
VideoSearch-R1 introduces an agentic framework that iteratively retrieves videos and refines search queries using continuous latent space refinement and policy optimization, achieving state-of-the-art performance on video corpus moment retrieval and temporal grounding tasks.
View Cached Full Text
Cached at: 07/02/26, 03:46 AM
Paper page - VideoSearch-R1: Iterative Video Retrieval and Reasoning via Soft Query Refinement
Source: https://huggingface.co/papers/2607.00446
Abstract
VideoSearch-R1 is an agentic framework that iteratively retrieves videos and refines search queries using continuous latent space refinement and policy optimization for improved video moment retrieval and temporal grounding.
As video corpora continue to expand in both scale and task complexity, there is increasing demand for approaches that retrieve relevant videos from large-scale corpora (inter-video reasoning) and subsequently perform fine-grained, query-conditioned tasks (intra-video reasoning) within the retrieved content, such astemporal grounding. However, existing approaches typically treat retrieval as a preprocessing step, and consequently, when the initial retrieval fails, there is no mechanism to refine the search, leading to the failure of subsequent fine-grainedintra-video reasoning. Moreover, while recentagentic frameworks have advanced video understanding, they typically assume that the query-relevant video is already given, focusing exclusively onintra-video reasoningtasks. To address these limitations, we propose VideoSearch-R1, anagentic frameworkfor iterativevideo retrievaland reasoning through multi-turn interaction with a video search engine. Specifically, we introduceSoft Query Refinement(SQR) to refine search query tokens in acontinuous latent spacerather than rewriting queries in the discrete text space, enabling more efficient and fine-grained adjustments. SQR and its reasoning process are trained usingGroup Relative Policy Optimization(GRPO), guided bytask-level reward signalsderived from retrieval and downstream tasks. Building upon this, VideoSearch-R1 achieves state-of-the-art performance across three datasets onVideo Corpus Moment Retrieval(VCMR), iteratively retrieving videos from large-scale corpora, refining search queries, and performing precise query-conditionedtemporal groundingwithin the retrieved content. Our analyses show that SQR effectively refines the original query, requiring significantly fewer generated tokens than explicit text-level query refinement. Code and model checkpoints are publicly available at mlvlab.github.io/VideoSearch-R1.
View arXiv pageView PDFProject pageGitHub20Add to collection
Get this paper in your agent:
hf papers read 2607\.00446
Don’t have the latest CLI?curl \-LsSf https://hf\.co/cli/install\.sh \| bash
Models citing this paper0
No model linking this paper
Cite arxiv.org/abs/2607.00446 in a model README.md to link it from this page.
Datasets citing this paper0
No dataset linking this paper
Cite arxiv.org/abs/2607.00446 in a dataset README.md to link it from this page.
Spaces citing this paper0
No Space linking this paper
Cite arxiv.org/abs/2607.00446 in a Space README.md to link it from this page.
Collections including this paper0
No Collection including this paper
Add this paper to acollectionto link it from this page.
Similar Articles
VideoSeeker: Incentivizing Instance-level Video Understanding via Native Agentic Tool Invocation
VideoSeeker introduces a paradigm for instance-level video understanding that integrates agentic reasoning with visual prompts, achieving superior performance through automated data synthesis and reinforcement learning, outperforming GPT-4o and Gemini-2.5-Pro.
Scaling Retrieval-Augmented Reasoning with Parallel Search and Explicit Merging
Introduces MultiSearch, an RL-based framework that generates multiple queries at each reasoning step and explicitly merges retrieved information to improve signal-to-noise ratio and reasoning accuracy in question-answering tasks.
OneSearch-VL: Unified Multimodal Deep Research Agent for Image and Video
该论文提出 OneSearch-VL，一个以视觉接地证据图（VGEG）为核心的统一多模态深度研究智能体，用于单图、多图和视频研究任务，并构建了 SFT/RL 数据集及细粒度评测基准，在多个图文与视频基准上显著超越 Qwen3-VL-8B。
Visual-Seeker: Towards Visual-Native Multimodal Agentic Search via Active Visual Reasoning
Visual-Seeker proposes a visual-native multimodal deep search agent that actively reasons over fine-grained visual details and synthesizes multimodal evidence, achieving state-of-the-art performance on five challenging multimodal search benchmarks.
EasyVideoR1: Easier RL for Video Understanding
EasyVideoR1 is an efficient reinforcement learning framework for training large vision-language models on video understanding tasks, featuring offline preprocessing with tensor caching for 1.47x throughput improvement, a task-aware reward system covering 11 problem types, and evaluation across 22 video benchmarks. It also supports joint image-video training and a mixed offline-online data training paradigm.