Simple o3: Enhancing Visual Grounding in Multimodal Reasoning via Adaptive Tool Composition

ACL ARR 2026 January Submission3931 Authors

Keywords: Multimodal Large Language Models; Visual Grounding Enhancement; Multimodal Reasoning

Abstract: Multimodal large language models (MLLMs) have achieved strong performance on vision--language tasks, yet often fail to preserve and effectively leverage visual evidence throughout generation. We identify three fundamental types of visual grounding failures: Long-Context Grounding Error, where visual embeddings gradually decay over long sequences; Fine-Grained Grounding Error, where low-resolution or degraded inputs hinder the recovery of detailed visual information; and Regional Grounding Error, where spatially diffuse attention weakens region-level vision--language alignment. To address these issues, we propose a tool-augmented reasoning framework with three targeted compensation strategies: reuse, which re-injects the original image to mitigate visual forgetting; focus_area, which constrains attention to task-relevant regions; and zoom_in, which enhances visual resolution for fine-grained perception. We further construct the TWI-Tools-146K dataset and develop Simple-o3, a tool-augmented MLLM that interleaves visual and textual tokens. Extensive experiments show that each tool yields targeted improvements for its corresponding grounding error, while their combination produces synergistic gains in visual reasoning. Beyond performance, our analysis provides mechanistic insights into how tool-based interventions improve visual grounding, pointing toward more reliable multimodal reasoning.

Paper Type: Long

Research Area: Multimodality and Language Grounding to Vision, Robotics and Beyond

Research Area Keywords: cross-modal content generation; vision question answering; multimodality; image text matching

Languages Studied: English

Submission Number: 3931