BrowseComp-$V^3$: A Visual, Vertical, and Verifiable Benchmark for Multimodal Browsing Agents

TL;DR

BrowseComp-$V^3$ tackles the inadequacy of existing multimodal deep-search benchmarks by introducing a 300-question benchmark with multi-hop, cross-modal reasoning and publicly searchable evidence. It pairs this with OmniSeeker, a unified framework that integrates diverse web-search and visual-perception tools to enhance open-source models. Across human and multiple model configurations, results show a large performance gap, with humans at 68–82% in process and final-success metrics, while state-of-the-art models remain below 40% SR, underscoring bottlenecks in multimodal integration and long-horizon planning. The dataset construction pipeline, process-oriented evaluation, and open-tool framework establish a rigorous, reproducible platform to advance robust multimodal deep search in open-world settings.

Abstract

Multimodal large language models (MLLMs), equipped with increasingly advanced planning and tool-use capabilities, are evolving into autonomous agents capable of performing multimodal web browsing and deep search in open-world environments. However, existing benchmarks for multimodal browsing remain limited in task complexity, evidence accessibility, and evaluation granularity, hindering comprehensive and reproducible assessments of deep search capabilities. To address these limitations, we introduce BrowseComp-$V^3$, a novel benchmark consisting of 300 carefully curated and challenging questions spanning diverse domains. The benchmark emphasizes deep, multi-level, and cross-modal multi-hop reasoning, where critical evidence is interleaved across textual and visual modalities within and across web pages. All supporting evidence is strictly required to be publicly searchable, ensuring fairness and reproducibility. Beyond final-answer accuracy, we incorporate an expert-validated, subgoal-driven process evaluation mechanism that enables fine-grained analysis of intermediate reasoning behaviors and systematic characterization of capability boundaries. In addition, we propose OmniSeeker, a unified multimodal browsing agent framework integrating diverse web search and visual perception tools. Comprehensive experiments demonstrate that even state-of-the-art models achieve only 36% accuracy on our benchmark, revealing critical bottlenecks in multimodal information integration and fine-grained perception. Our results highlight a fundamental gap between current model capabilities and robust multimodal deep search in real-world settings.

Figures (5)

• Figure 1: An overview of the data construction process of BrowseComp-V$^3$.
• Figure 2: Statistics of BrowseComp-V$^3$. (Left) Category distribution across primary domains. (Right) Summary of statistics.
• Figure 3: Difficulty and Ability Analysis
• Figure 4: Test Time Scaling
• Figure 5: Failure Mode Analysis