数据来源：HuggingFace Papers

Latest Papers

1. Omni-Woodpecker: Hallucination Correction via Evidence-Prior Rebalancing for Omni LLMs

Abstract:Recent Omni LLMs (OLLMs) have demonstrated remarkable capabilities in joint audio-visual reasoning, yet cross-modal hallucinations remain a major challenge. Existing mitigation methods mainly operate at the output level, overlooking the entanglement of grounding evidence and learned priors within fused representations. To address this issue, we propose Omni-Woodpecker (OWP), a training-free framework that shifts hallucination mitigation from output-level manipulation to conflict-conditioned representation rebalancing. Specifically, OWP first identifies cross-modal conflicts from complementary modality-specific views. It then employs orthogonal decomposition to separate grounding evidence from competing priors within the hidden states. Finally, OWP adaptively rebalances fused representations by suppressing misleading priors and restoring ignored evidence before decoding. Extensive experiments demonstrate the effectiveness and strong cross-architecture generalization of OWP. Using Qwen2.5-Omni-7B as the backbone, OWP improves performance by 12.9 and 8.6 points on CMM and AVHBench, respectively, while consistently achieving gains of 0.67-2.25 points across four general OLLM reasoning benchmarks. Furthermore, OWP generalizes effectively to VideoLLaMA2-AV, improving performance by 4.6 and 3.9 points on the same benchmarks. Code is publicly available at this https URL

中文摘要

摘要：近期的全能大语言模型（Omni LLMs，OLLMs）在联合音视频推理方面表现出显著能力，但跨模态幻觉仍然是一个主要挑战。现有的缓解方法主要在输出层进行操作，忽略了融合表征中基础证据与学习先验的纠缠。为了解决这一问题，我们提出了 Omni-Woodpecker（OWP），这是一个无需训练的框架，将幻觉缓解从输出层操作转移到冲突条件下的表征再平衡。具体而言，OWP 首先从互补的模态特定视角识别跨模态冲突。然后，它利用正交分解在隐藏状态中将基础证据与竞争先验分离。最后，OWP 通过抑制误导先验和恢复被忽略的证据，在解码前自适应地再平衡融合表征。大量实验表明 OWP 的有效性及强大的跨架构通用性。以 Qwen2.5-Omni-7B 为骨干，OWP 在 CMM 和 AVHBench 上分别提升性能 12.9 和 8.6 点，同时在四个通用 OLLM 推理基准上持续实现 0.67-2.25 点的增益。此外，OWP 在 VideoLLaMA2-AV 上也能有效泛化，在相同基准上的性能分别提升 4.6 和 3.9 点。代码公开可在此 https URL 获取。

LLM Analysis

LLM Analysis Failed: Error: Kimi内容未就绪或抓取失败: 2610.27103 (HTTP 429)

Authors: Zijian Liu,Chaoyou Fu,Yan Yang,Caifeng Shan,Ran He

Categories:

PDF URL: https://arxiv.org/pdf/2610.27103.pdf

Arxiv URL: https://arxiv.org/abs/2610.27103

Arxiv ID: 2610.27103

CoolPaper URL: https://papers.cool/arxiv/2610.27103

Published: 2026-10-26T01:41:08.362Z

Updated: 2026-10-26T01:41:08.362Z