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

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-ofthe-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. We make our code and data available at https: //github.com/Akomand/CausalVLBench . * Light Position: left * Shadow Length: short * Shadow Position: right 2. Desired Manipulation: We want to change the light position to 'right'. 3. Causal Reasoning: * The rule (2) says that if the light position changes, it causes both the shadow length and shadow position to change. It does NOT cause the pendulum angle to change. * Given that the light moves from left to right, the shadow position will move from right to left. * Given that the light moves from left to right, the shadow length will become long. 4. Predicted Final State: * Pendulum Angle: left (unchanged) * Light Position: right * Shadow Length: long * Shadow Position: left