Tag
This paper introduces FPCO-Dialog, a benchmark for evaluating correction and cooperation behavior in vision-language models under repeated false premises in multi-turn dialogues.