arXiv AI· Enhao Cui, Junzhi Yang, Yaoshun Lin, Yuhan Feng, Zijian Chen·· 1 天前评分53
Where Do Vision-Language Models Get Left and Right? Position-Code Counterfactuals in Qwen2.5-VL, Qwen3-VL and InternVL3
摘要
原文摘录（自动中文摘要暂不可用）：Where Do Vision-Language Models Get Left and Right? Position-Code Counterfactuals in Qwen2.5-VL, Qwen3-VL and InternVL3 A vision-language model may take left-right direction from the orientation of image content within patches, its vision transformer's (ViT) position code, or token positions and order in its language model (LLM). Ablations show the ViT code
推荐理由
自动中文摘要尚未通过校验，请核对原文与完整来源。
本站只提供摘要与原文入口。完整内容请阅读原文。
来源：arXiv AI · arxiv.org