CoVLM: Composing Visual Entities and Relationships in Large Language Models Via Communicative Decoding

TL;DR

CoVLM introduces a novel communicative decoding framework that enables large language models to explicitly compose visual entities and relationships through dynamic communication tokens, significantly improving compositional reasoning in vision-language tasks.

Contribution

The paper proposes a new communication-based approach for LLMs to explicitly model visual entities and relations, enhancing compositional reasoning capabilities.

Findings

Outperforms previous VLMs by ~20% in HICO-DET mAP

Achieves ~14% improvement in Cola top-1 accuracy

Sets new state-of-the-art in referring expression comprehension and visual question answering

Abstract

A remarkable ability of human beings resides in compositional reasoning, i.e., the capacity to make "infinite use of finite means". However, current large vision-language foundation models (VLMs) fall short of such compositional abilities due to their "bag-of-words" behaviors and inability to construct words that correctly represent visual entities and the relations among the entities. To this end, we propose CoVLM, which can guide the LLM to explicitly compose visual entities and relationships among the text and dynamically communicate with the vision encoder and detection network to achieve vision-language communicative decoding. Specifically, we first devise a set of novel communication tokens for the LLM, for dynamic communication between the visual detection system and the language system. A communication token is generated by the LLM following a visual entity or a relation, to inform the detection network to propose regions that are relevant to the sentence generated so far. The proposed regions-of-interests (ROIs) are then fed back into the LLM for better language generation contingent on the relevant regions. The LLM is thus able to compose the visual entities and relationships through the communication tokens. The vision-to-language and language-to-vision communication are iteratively performed until the entire sentence is generated. Our framework seamlessly bridges the gap between visual perception and LLMs and outperforms previous VLMs by a large margin on compositional reasoning benchmarks (e.g., ~20% in HICO-DET mAP, ~14% in Cola top-1 accuracy, and ~3% on ARO top-1 accuracy). We also achieve state-of-the-art performances on traditional vision-language tasks such as referring expression comprehension and visual question answering.