Tag
该论文提出一个两参数贝叶斯线索组合模型,解释视觉-语言模型在文本与图像证据冲突时如何整合信息,并能准确预测模型在未见过的输入退化组合下的偏好,统一了此前关于文本主导和视觉主导的矛盾报告。
该论文系统研究了混合模态检索器的模态偏好问题,发现无关文本比无关图像对检索性能的干扰更严重(称为 "Chaos in the Text"),并提出 Trident 方法,通过 Multi-Positive View InfoNCE 联合优化文本、图像和融合视图来缓解该偏差。
The paper introduces Tri-PvP, a benchmark that exposes visual bias and asymmetric evidence-form preferences in omni-modal large language models, revealing deep-seated modality biases that are linearly decodable from early layers and resistant to surface mitigation.
Introduces C3PO, a benchmark of 3,404 samples for evaluating cross-modal composition and counterfactual reasoning in multimodal LLMs. It finds modality dominance causes most failures, with even the best model (Gemini-3.1-Pro) far below human accuracy.