arXiv：cs.CL· Huiqiang Rong, Haoran Luo, Hui Feng, Zhonghong Ou, Kaiwen Xue, Guoxin Zhang, Yifan Zhu·· 4 小时前AI 评分42
OmniConfess：用 token 级“自白”缓解全模态模型幻觉
OmniConfess: Eliciting Token Confessions to Mitigate Omni-Modal Hallucination
AI 导读
OmniConfess 是一种免训练方法，通过固定候选回答并在受控的单通道证据干预下按 token 重新打分，生成结构化的“token×通道自白”，揭示回答依赖哪些证据，从而保留有据内容、纠正由无关或矛盾证据驱动的表述。
正文
Abstract:Omni-modal large language models (OmniLLMs) unify text, images, audio, and video, yet hallucinate when generation relies on the wrong evidence. Existing inference-time methods can reduce hallucinations, but rarely reveal which evidence sustains a generated commitment. We introduce OmniConfess, a training-free method for mitigating omni-modal hallucinations. It fixes a candidate response and re-scores it at token resolution under controlled channel-wise evidence interventions, producing a structured token-by-channel confession that reveals the response's evidential dependence. OmniConfess uses this confession to preserve grounded content and correct commitments driven by irrelevant or contradictory evidence. To evaluate OmniConfess, we construct OmniHalluBench, a 3,540-example benchmark built from six datasets spanning text, image, audio, and video settings and both judgment and free-form generation. Experiments show that OmniConfess mitigates hallucinations across heterogeneous modality and task settings. Our code and benchmark are publicly available at this https URL.
| Subjects: | Computation and Language (cs.CL) |
| Cite as: | arXiv:2610.02999 [cs.CL] |
| (or arXiv:2610.02999v1 [cs.CL] for this version) | |
| https://doi.org/10.48550/arXiv.2610.02999 arXiv-issued DOI via DataCite (pending registration) |
Submission history
From: Huiqiang Rong [view email]
[v1]
Fri, 2 Oct 2026 08:31:10 UTC (6,595 KB)
来源：arXiv：cs.CL · arxiv.org