CBX-Bench: A Human-Aligned MLLM Council for Benchmarking Concept Bottleneck Model Explanations
A multimodal large language model (MLLM) council is developed that, given an image and its CBM explanation, produces an explanation quality score, and CBX-Bench, a public benchmark and leaderboard, provides a human-aligned, scalable evaluation of CBM explanations beyond accuracy and isolated qualitative examples.