Deeper is Not Always Better：Mitigating the Alignment Tax via Confident Layer Decoding
(arxiv 2026) Qwen 团队、清华大学和南洋理工大学
Traceable Inference with Graph-Based Evidence Routing for Mitigating Hallucinations in Multimodal Generation
(arxiv 2026) EMNLP 2026在投
HulluEdit： Single-Pass Evidence-Consistent Subspace Editing for Mitigating Hallucinations in Large Vision-Language Models
（CVPR 2026） 增强版Nullu
Causally-Grounded Dual-Path Attention Intervention for Object Hallucination Mitigation in LVLMs
（AAAI 2026）
Cross-Modal Attention Calibration for LVLM Hallucination Mitigation
（CVPR 2026） 该论文提出了一种无需训练的模态间相关性校准解码框架，通过跨模态价值增强解码和内容驱动的注意力修正机制，有效缓解了大型视觉语言模型中由单模态过度依赖和虚假跨模态相关性引发的生成幻觉问题。
First Logit Boosting： Visual Grounding Method to Mitigate Object Hallucination in Large Vision-Language Models
（CVPR 2026）