CoT4Det: A Chain-of-Thought Framework for Perception-Oriented Vision-Language Tasks

TL;DR

This paper introduces CoT4Det, a framework that reformulates perception tasks into interpretable steps to enhance vision-language models' performance on detection, segmentation, and related tasks, bridging the gap with task-specific models.

Contribution

It proposes a novel Chain-of-Thought approach for perception tasks, significantly improving LVLMs' accuracy on detection and other perception benchmarks.

Findings

Boosts COCO2017 mAP from 19% to 33%.

Outperforms baselines by +2% on RefCOCO.

Achieves 19% improvement on Flickr30k entities.

Abstract

Large Vision-Language Models (LVLMs) have demonstrated remarkable success in a broad range of vision-language tasks, such as general visual question answering and optical character recognition (OCR). However, their performance on perception-centric tasks -- such as object detection, semantic segmentation, and depth estimation -- remains significantly inferior to that of task-specific expert models. For example, Qwen2.5-VL-7B-Instruct achieves only 19% mAP on COCO2017 val, particularly struggling with dense scenes and small object recall. In this work, we introduce Chain-of-Thought for Detection (CoT4Det), a simple but efficient strategy that reformulates perception tasks into three interpretable steps: classification, counting, and grounding -- each more naturally aligned with the reasoning capabilities of LVLMs. Extensive experiments demonstrate that our method significantly improves perception performance without compromising general vision language capabilities. With a standard Qwen2.5-VL-7B-Instruct, CoT4Det boosts mAP from 19.0% to 33.0% on COCO2017 val and achieves competitive results across a variety of perception benchmarks, outperforming baselines by +2% on RefCOCO series and 19% on Flickr30k entities.