Question Causal Drive

New framework evaluates vision-language models by tracing input influence

Researchers have introduced a new framework to evaluate vision-language models (VLMs) by analyzing how different inputs influence their output generation process. This temporal causal drive evaluation framework uses int…