揭示大视觉语言模型对视觉缺失令牌的响应
Unveiling the Response of Large Vision-Language Models to Visually Absent Tokens
AI总结：
该研究发现大视觉语言模型易错误感知无视觉依据的文本，识别出视觉缺失感知神经元，开发检测模块并提出输出优化方法，有效缓解相关错误且具备跨模型通用性。
摘要：
大视觉语言模型（LVLMs）通过联合解读视觉与文本输入来生成与上下文相关的响应。然而，我们的研究发现，这类模型常会错误地将缺乏视觉证据的文本输入感知为图像的一部分，进而产生错误响应。基于这一发现，我们探究了LVLMs是否具备判断文本概念是否在图像中存在对应依据的内在能力，并发现了前馈网络（FFN）中的一类特定神经元子集，我们将其命名为视觉缺失感知（VA）神经元，它们会通过独特的激活模式持续传递视觉缺失的信号。利用这些激活模式，我们开发了一个检测模块，可系统地对输入令牌是否具备视觉依据进行分类。在其预测结果的引导下，我们提出了一种优化输出的方法：在生成过程中重新解读问题提示，或替换检测到的缺失令牌。大量实验表明，我们的方法能有效缓解模型错误假定文本输入具备视觉存在性的倾向，且该方法在多种LVLMs上均具备通用性。
英文摘要：
Large Vision-Language Models (LVLMs) generate contextually relevant responses by jointly interpreting visual and textual inputs. However, our finding reveals they often mistakenly perceive text inputs lacking visual evidence as being part of the image, leading to erroneous responses. In light of this finding, we probe whether LVLMs possess an internal capability to determine if textual concepts are grounded in the image, and discover a specific subset of Feed-Forward Network (FFN) neurons, termed Visual Absence-aware (VA) neurons, that consistently signal the visual absence through a distinctive activation pattern. Leveraging these patterns, we develop a detection module that systematically classifies whether an input token is visually grounded. Guided by its prediction, we propose a method to refine the outputs by reinterpreting question prompts or replacing the detected absent tokens during generation. Extensive experiments show that our method effectively mitigates the models' tendency to falsely presume the visual presence of text input and its generality across various LVLMs.
机构、总结、中文摘要均由AI分析而来，可能存在错误，请以论文正文为准。
