arXiv NLP
techCenter
FPCO-Dialog: A Multi-Turn False-Premise Benchmark for Correction and Cooperation in Vision-Language Modelstranslating…
arXiv:2609.03331v1 Announce Type: new
Abstract: Vision-language models (VLMs) are increasingly deployed in multi-turn settings where users may describe visual content with incorrect assumptions. Yet existing evaluations rarely isolate how models respond when the same visually…
Keywords#Models#Vision-Language Models#Multi-Turn#A Multi-Turn#A Multi-Turn False-Premise
Comments
Sign in to join the discussion.
Loading comments…