作者位次 #5/9ACL 2026Findings· Findings of ACL
Token-level Inference-Time Alignment for Vision-Language Models
Kejia Chen · Junjun Zheng · Jiawen Zhang · Manxi Lin · Xiao Pan · Jiacong Hu · Jian Lou · Zunlei Feng · Mingli Song
摘要
Vision-Language Models (VLMs) often prioritize linguistic fluency over visual fidelity, leading to hallucinations where generated text contradicts the image. Countering this bias typically requires resource-heavy fine-tuning or high-latency verification methods that provide feedback only after th...