VTC-Bench: Evaluating Agentic Multimodal Models via Compositional Visual Tool Chaining

2026-03-16

AI Summary

Existing benchmarks struggle to evaluate the capability of multimodal large language models (MLLMs) to perform long-horizon, multi-step tool invocation in complex visual tasks. To address this gap, this work introduces a novel evaluation benchmark grounded in real-world computer vision workflows, integrating 32 OpenCV tools and 680 hierarchically structured tasks. It presents the first fine-grained, multi-level evaluation framework that enables systematic assessment of both multi-tool composition and generalization to unseen operations. Comprehensive experiments across 19 state-of-the-art models reveal significant bottlenecks in tool adaptability and compositional planning, with even the best-performing model, Gemini-3.0-Pro, achieving only 51% accuracy—highlighting the current limitations of vision agents in complex tool-based reasoning.

Technology Category

• Computer Vision: Multi-modal Vision
• Machine Learning: Large Multimodal Models (LMMs)
• Natural Language Processing: Language Grounding & Multi-modal NLP

Application Category

• Search and Retrieval-Augmented AI: Search Tool Learning with LLM: Teaching LLMs to invoke search and make use of retrieved information
• Economics, Online Markets and Human Computation: Cost models of using LLMs in production systems
• Graph Algorithms and Modeling for the Web: Foundation models and LLMs for Web-related graphs

Abstract

Recent advancements extend Multimodal Large Language Models (MLLMs) beyond standard visual question answering to utilizing external tools for advanced visual tasks. Despite this progress, precisely executing and effectively composing diverse tools for complex tasks remain persistent bottleneck. Constrained by sparse tool-sets and simple tool-use trajectories, existing benchmarks fail to capture complex and diverse tool interactions, falling short in evaluating model performance under practical, real-world conditions. To bridge this gap, we introduce VisualToolChain-Bench~(VTC-Bench), a comprehensive benchmark designed to evaluate tool-use proficiency in MLLMs. To align with realistic computer vision pipelines, our framework features 32 diverse OpenCV-based visual operations. This rich tool-set enables extensive combinations, allowing VTC-Bench to rigorously assess multi-tool composition and long-horizon, multi-step plan execution. For precise evaluation, we provide 680 curated problems structured across a nine-category cognitive hierarchy, each with ground-truth execution trajectories. Extensive experiments on 19 leading MLLMs reveal critical limitations in current models' visual agentic capabilities. Specifically, models struggle to adapt to diverse tool-sets and generalize to unseen operations, with the leading model Gemini-3.0-Pro only achieving 51\% on our benchmark. Furthermore, multi-tool composition remains a persistent challenge. When facing complex tasks, models struggle to formulate efficient execution plans, relying heavily on a narrow, suboptimal subset of familiar functions rather than selecting the optimal tools. By identifying these fundamental challenges, VTC-Bench establishes a rigorous baseline to guide the development of more generalized visual agentic models.

Problem

• Multimodal Large Language Models
• tool composition
• visual reasoning
• benchmark evaluation
• agentic capabilities

Innovation

• tool chaining
• multimodal agentic models
• visual reasoning benchmark
• compositional tool use
• OpenCV-based operations