CausalVLBench: Benchmarking Visual Causal Reasoning in Large Vision-Language Models

TL;DR

This paper introduces CausalVLBench, a comprehensive benchmark for evaluating the causal reasoning capabilities of large vision-language models across multiple tasks and datasets, highlighting their strengths and weaknesses.

Contribution

It presents the first dedicated benchmark for visual causal reasoning in LVLMs, covering three key tasks and providing insights into current model performance.

Findings

LVLMs show mixed performance on causal reasoning tasks.

Benchmark reveals specific weaknesses in current LVLMs.

Results motivate future research to improve visual causal reasoning.

Abstract

Large language models (LLMs) have shown remarkable ability in various language tasks, especially with their emergent in-context learning capability. Extending LLMs to incorporate visual inputs, large vision-language models (LVLMs) have shown impressive performance in tasks such as recognition and visual question answering (VQA). Despite increasing interest in the utility of LLMs in causal reasoning tasks such as causal discovery and counterfactual reasoning, there has been relatively little work showcasing the abilities of LVLMs on visual causal reasoning tasks. We take this opportunity to formally introduce a comprehensive causal reasoning benchmark for multi-modal in-context learning from LVLMs. Our CausalVLBench encompasses three representative tasks: causal structure inference, intervention target prediction, and counterfactual prediction. We evaluate the ability of state-of-the-art open-source LVLMs on our causal reasoning tasks across three causal representation learning datasets and demonstrate their fundamental strengths and weaknesses. We hope that our benchmark elucidates the drawbacks of existing vision-language models and motivates new directions and paradigms in improving the visual causal reasoning abilities of LVLMs.