Research

Paper

AI LLM February 23, 2026

Test-Time Computing for Referring Multimodal Large Language Models

Authors

Mingrui Wu, Hao Chen, Jiayi Ji, Xiaoshuai Sun, Zhiyuan Liu, Liujuan Cao, Ming-Ming Cheng, Rongrong Ji

Abstract

We propose ControlMLLM++, a novel test-time adaptation framework that injects learnable visual prompts into frozen multimodal large language models (MLLMs) to enable fine-grained region-based visual reasoning without any model retraining or fine-tuning. Leveraging the insight that cross-modal attention maps intrinsically encode semantic correspondences between textual tokens and visual regions, ControlMLLM++ optimizes a latent visual token modifier during inference via a task-specific energy function to steer model attention towards user-specified areas. To enhance optimization stability and mitigate language prompt biases, ControlMLLM++ incorporates an improved optimization strategy (Optim++) and a prompt debiasing mechanism (PromptDebias). Supporting diverse visual prompt types including bounding boxes, masks, scribbles, and points, our method demonstrates strong out-of-domain generalization and interpretability. The code is available at https://github.com/mrwu-mac/ControlMLLM.

Metadata

arXiv ID: 2602.19505
Provider: ARXIV
Primary Category: cs.CV
Published: 2026-02-23
Fetched: 2026-02-24 04:38

Related papers

Raw Data (Debug)
{
  "raw_xml": "<entry>\n    <id>http://arxiv.org/abs/2602.19505v1</id>\n    <title>Test-Time Computing for Referring Multimodal Large Language Models</title>\n    <updated>2026-02-23T04:42:10Z</updated>\n    <link href='https://arxiv.org/abs/2602.19505v1' rel='alternate' type='text/html'/>\n    <link href='https://arxiv.org/pdf/2602.19505v1' rel='related' title='pdf' type='application/pdf'/>\n    <summary>We propose ControlMLLM++, a novel test-time adaptation framework that injects learnable visual prompts into frozen multimodal large language models (MLLMs) to enable fine-grained region-based visual reasoning without any model retraining or fine-tuning. Leveraging the insight that cross-modal attention maps intrinsically encode semantic correspondences between textual tokens and visual regions, ControlMLLM++ optimizes a latent visual token modifier during inference via a task-specific energy function to steer model attention towards user-specified areas. To enhance optimization stability and mitigate language prompt biases, ControlMLLM++ incorporates an improved optimization strategy (Optim++) and a prompt debiasing mechanism (PromptDebias). Supporting diverse visual prompt types including bounding boxes, masks, scribbles, and points, our method demonstrates strong out-of-domain generalization and interpretability. The code is available at https://github.com/mrwu-mac/ControlMLLM.</summary>\n    <category scheme='http://arxiv.org/schemas/atom' term='cs.CV'/>\n    <published>2026-02-23T04:42:10Z</published>\n    <arxiv:comment>arXiv admin note: substantial text overlap with arXiv:2407.21534</arxiv:comment>\n    <arxiv:primary_category term='cs.CV'/>\n    <author>\n      <name>Mingrui Wu</name>\n    </author>\n    <author>\n      <name>Hao Chen</name>\n    </author>\n    <author>\n      <name>Jiayi Ji</name>\n    </author>\n    <author>\n      <name>Xiaoshuai Sun</name>\n    </author>\n    <author>\n      <name>Zhiyuan Liu</name>\n    </author>\n    <author>\n      <name>Liujuan Cao</name>\n    </author>\n    <author>\n      <name>Ming-Ming Cheng</name>\n    </author>\n    <author>\n      <name>Rongrong Ji</name>\n    </author>\n  </entry>"
}