1 paper
Dasol Choi, Guijin Son, Hanwool Lee +7
Current vision-language benchmarks predominantly feature well-structured questions with clear, explicit prompts. However, real user queries are often informal and underspecified. U…