Video-DeepResearch: Towards the Next-Generation Multimodal Deepresearch Agent

Hugging Face Daily Papers Papers
multimodal video deep-research agent benchmark grpo

Summary

Video-DeepResearch (Video-DR) extends multimodal agents from static images to continuous video streams, introducing a decoupled perception-exploration pipeline and a new benchmark Video-DR-Bench. Their Video-DeepResearch-35B-A3B model achieves 64.0% accuracy, surpassing Claude-4.5-Sonnet, GPT-5, and Gemini 2.5 Pro.

We introduce Video-DeepResearch (Video-DR), extending multimodal agents from static images to continuous video streams, a setting that demands dense spatiotemporal grounding coupled with open-web exploration. Preliminary evaluations reveal two critical bottlenecks in current models: (1) modality bias, where agents bypass visual tools in favor of textual search, and (2) parametric knowledge leakage, where models rely on internal memory rather than genuine tool-augmented execution. To address these challenges, we propose Video-DR, featuring a decoupled perception-exploration pipeline with stage-wise tool unlocking that compels exhaustive cross-frame visual grounding prior to web retrieval. Our framework adopts a two-stage training recipe: supervised fine-tuning followed by Group Relative Policy Optimization (GRPO), enabling autonomous exploration that breaks the imitation-learning ceiling. Furthermore, we curate Video-DR-Bench, a human-AI collaborative benchmark comprising 200 complex, multi-hop VQA instances. Empirical results demonstrate that our Video-DeepResearch-35B-A3B establishes a new state-of-the-art of 64.0% average accuracy, surpassing proprietary Claude-4.5-Sonnet (59.0%) by 5.0 points and significantly outperforming GPT-5 (52.5%) and Gemini 2.5 Pro (57.5%). The 30B-A3B variant achieves 59.3%, competitive with Claude-4.5-Sonnet and demonstrating the effectiveness of our training paradigm even at compact scale. Code: https://github.com/Osilly/Vision-DeepResearch.
Original Article
View Cached Full Text

Cached at: 08/05/26, 05:43 AM

Paper page - Video-DeepResearch: Towards the Next-Generation Multimodal Deepresearch Agent

Source: https://huggingface.co/papers/2608.03979 Authors:

,

,

,

,

,

,

,

,

,

,

,

,

,

,

,

,

,

,

Abstract

WeintroduceVideo-DeepResearch(Video-DR),extendingmultimodalagentsfromstaticimagestocontinuousvideostreams,asettingthatdemandsdensespatiotemporalgroundingcoupledwithopen-webexploration.Preliminaryevaluationsrevealtwocriticalbottlenecksincurrentmodels:(1)modalitybias,whereagentsbypassvisualtoolsinfavoroftextualsearch,and(2)parametricknowledgeleakage,wheremodelsrelyoninternalmemoryratherthangenuinetool-augmentedexecution.Toaddressthesechallenges,weproposeVideo-DR,featuringadecoupledperception-explorationpipelinewithstage-wisetoolunlockingthatcompelsexhaustivecross-framevisualgroundingpriortowebretrieval.Ourframeworkadoptsatwo-stagetrainingrecipe:supervisedfine-tuningfollowedbyGroupRelativePolicyOptimization(GRPO),enablingautonomousexplorationthatbreakstheimitation-learningceiling.Furthermore,wecurateVideo-DR-Bench,ahuman-AIcollaborativebenchmarkcomprising200complex,multi-hopVQAinstances.EmpiricalresultsdemonstratethatourVideo-DeepResearch-35B-A3Bestablishesanewstate-of-the-artof64.0%averageaccuracy,surpassingproprietaryClaude-4.5-Sonnet(59.0%)by5.0pointsandsignificantlyoutperformingGPT-5(52.5%)andGemini2.5Pro(57.5%).The30B-A3Bvariantachieves59.3%,competitivewithClaude-4.5-Sonnetanddemonstratingtheeffectivenessofourtrainingparadigmevenatcompactscale.Code:https://github.com/Osilly/Vision-DeepResearch.

View arXiv pageView PDFProject pageAdd to collection

Get this paper in your agent:

hf papers read 2608\.03979

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/2608.03979 in a model README.md to link it from this page.

Datasets citing this paper0

No dataset linking this paper

Cite arxiv.org/abs/2608.03979 in a dataset README.md to link it from this page.

Spaces citing this paper0

No Space linking this paper

Cite arxiv.org/abs/2608.03979 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

Mind DeepResearch Technical Report

Hugging Face Daily Papers

MindDR is a multi-agent deep research framework using a three-agent architecture (Planning, DeepSearch, Report) and a four-stage training pipeline, achieving competitive performance with ~30B-parameter models on multiple benchmarks. Developed by Li Auto and deployed as an online product, it also introduces MindDR Bench, a 500-query Chinese benchmark for evaluating deep research capabilities.

VideoGen-Agent: Reinforcing Video Generation Agents

Hugging Face Daily Papers

The paper presents VideoGen-Agent, a reinforcement learning-based multimodal agent that coordinates tools for video generation, significantly improving performance on the new VABench benchmark.