AI RESEARCH
When Choices Become Priors: Contrastive Decoding for Scientific Figure Multiple-Choice QA
arXiv CS.AI
•
ArXi:2603.28026v1 Announce Type: new Scientific figure multiple-choice question answering (MCQA) requires models to reason over diverse visual evidence, ranging from charts and multipanel figures to microscopy and biomedical images. However, this setting suffers from a distinctive bias: answer choices themselves can act as priors, steering multimodal models toward scientifically plausible options even when the figure s a different answer.