VL-ICL Bench: The Devil in the Details of Multimodal In-Context Learning

2024-03-19

AI Summary

Existing evaluation frameworks lack comprehensive assessment of Vision-Language Large Models (VLLMs) in image-text joint few-shot learning. Method: We introduce VL-ICL Bench—the first multimodal in-context learning (ICL) benchmark spanning perception, reasoning, and long-context dimensions—and systematically define and evaluate the scope and limits of multimodal ICL. Using zero-weight-update few-shot prompting, we assess state-of-the-art VLLMs—including GPT-4V, Qwen-VL, and LLaVA—on novel tasks involving cross-modal alignment, structured visual reasoning, and context-sensitive generation. Results: Experiments reveal substantial performance degradation in multi-step visual reasoning, fine-grained localization, and long visual context retention (average accuracy <40%), confirming the benchmark’s diagnostic rigor and challenge. VL-ICL Bench establishes a standardized evaluation protocol for multimodal ICL and identifies critical avenues for model improvement.

Technology Category

• Machine Learning: Large Multimodal Models (LMMs)
• Computer Vision: Large Vision Models
• Natural Language Processing: Language Grounding & Multi-modal NLP

Application Category

• User Modeling, Personalization and Recommendation: Large Language Models (LLM) for user modeling and recommendation
• Search and Retrieval-Augmented AI: Search Tool Learning with LLM: Teaching LLMs to invoke search and make use of retrieved information
• Economics, Online Markets and Human Computation: Cost models of using LLMs in production systems

Abstract

Large language models (LLMs) famously exhibit emergent in-context learning (ICL) -- the ability to rapidly adapt to new tasks using few-shot examples provided as a prompt, without updating the model's weights. Built on top of LLMs, vision large language models (VLLMs) have advanced significantly in areas such as recognition, reasoning, and grounding. However, investigations into emph{multimodal ICL} have predominantly focused on few-shot visual question answering (VQA), and image captioning, which we will show neither exploit the strengths of ICL, nor test its limitations. The broader capabilities and limitations of multimodal ICL remain under-explored. In this study, we introduce a comprehensive benchmark VL-ICL Bench for multimodal in-context learning, encompassing a broad spectrum of tasks that involve both images and text as inputs and outputs, and different types of challenges, from {perception to reasoning and long context length}. We evaluate the abilities of state-of-the-art VLLMs against this benchmark suite, revealing their diverse strengths and weaknesses, and showing that even the most advanced models, such as GPT-4, find the tasks challenging. By highlighting a range of new ICL tasks, and the associated strengths and limitations of existing models, we hope that our dataset will inspire future work on enhancing the in-context learning capabilities of VLLMs, as well as inspire new applications that leverage VLLM ICL. The code and dataset are available at https://github.com/ys-zong/VL-ICL.

Problem

• Visual Language Models
• Multimodal Learning
• Performance Improvement

Innovation

• VL-ICL Bench
• Visual Large Language Models (VLLMs)
• Integrated Learning