Link-Context Learning for Multimodal LLMs

TL;DR

This paper introduces link-context learning (LCL), a causal reasoning approach for multimodal large language models, enhancing their ability to recognize unseen images and understand novel concepts without additional training.

Contribution

The paper proposes link-context learning (LCL), a novel causal reasoning method that improves MLLMs' zero-shot recognition of unseen images and concepts, and introduces the ISEKAI dataset for evaluation.

Findings

LCL significantly outperforms vanilla MLLMs on unseen image recognition.

LCL enhances the understanding of causal relationships in data.

The ISEKAI dataset effectively evaluates link-context learning capabilities.

Abstract

The ability to learn from context with novel concepts, and deliver appropriate responses are essential in human conversations. Despite current Multimodal Large Language Models (MLLMs) and Large Language Models (LLMs) being trained on mega-scale datasets, recognizing unseen images or understanding novel concepts in a training-free manner remains a challenge. In-Context Learning (ICL) explores training-free few-shot learning, where models are encouraged to ``learn to learn" from limited tasks and generalize to unseen tasks. In this work, we propose link-context learning (LCL), which emphasizes "reasoning from cause and effect" to augment the learning capabilities of MLLMs. LCL goes beyond traditional ICL by explicitly strengthening the causal relationship between the support set and the query set. By providing demonstrations with causal links, LCL guides the model to discern not only the analogy but also the underlying causal associations between data points, which empowers MLLMs to recognize unseen images and understand novel concepts more effectively. To facilitate the evaluation of this novel approach, we introduce the ISEKAI dataset, comprising exclusively of unseen generated image-label pairs designed for link-context learning. Extensive experiments show that our LCL-MLLM exhibits strong link-context learning capabilities to novel concepts over vanilla MLLMs. Code and data will be released at https://github.com/isekai-portal/Link-Context-Learning.