小模型自知错误却说不出来，内部概率比口头自信更可信。
Small Vision-Language Models Know When They Are Wrong But Cannot Say So: A Two-Model Study of Stated versus Internal Confidence Under Realistic Image Degradation
- 用内部词元概率替代口头自信，能更好识别模型错误
- 在严重过暗场景下，两种信心信号均失效，准确率暴跌至0.22
- 适合部署在低算力设备的模型，需依赖内部概率做决策
视觉语言模型（VLMs）日益部署于消费级硬件，输入图像常因压缩、抖动和弱光而退化。此时，可靠的不确定性信号比原始准确率更重要，因为它决定系统是否应放弃回答。我们评估了两个小型开源VLM（Qwen2-VL-2B-Instruct 和 SmolVLM-Instruct），在六种真实摄影退化下，三种严重程度，对比两种信心信号：模型自然语言中表达的信心，以及其生成答案的平均词元概率。3800次预测结果显示显著差距：Qwen2-VL的口头信心几乎恒定（均值0.87-0.90），对自身错误的检测仅达随机水平（AUROC 0.39-0.75，通常约0.50），而内部概率可实现0.92-0.99的准确区分。SmolVLM中，口头信心基本无法获取（五次尝试仅一次成功解析），内部概率仍能有效检测错误（AUROC 0.54-0.92）。两模型在严重欠曝时表现一致崩溃：准确率从0.99降至0.22（Qwen2-VL），0.97降至0.42（SmolVLM），但信心信号几乎不变，内部检测能力退至随机。结论是，小型VLM虽具备内部自知能力，但未在输出中体现；内部概率是受限部署下的更优拒答信号；极端弱光条件下，任一信号均不可信。
原文摘要 · Abstract (English)
Vision-language models (VLMs) are increasingly deployed on consumer hardware where input images are degraded by compression, camera shake, and poor lighting. In such settings, a reliable uncertainty signal matters more than raw accuracy, because it determines when a system should defer rather than answer. We evaluate two small open-weight VLMs -- Qwen2-VL-2B-Instruct and SmolVLM-Instruct -- across six realistic photographic degradations at three severity levels, comparing two confidence signals: the confidence the model states in natural language, and the model's own mean token probability over its generated answer. Across 3,800 predictions, we find a large and consistent gap. Verbalized confidence in Qwen2-VL is almost constant (mean 0.87-0.90 across all conditions) and detects its own errors at chance level (AUROC 0.39-0.75, typically ~0.50), while internal token probability from the same model separates correct from incorrect answers with AUROC 0.92-0.99. In SmolVLM, verbalized confidence proved largely unobtainable: across three prompt templates, only one of five pilot attempts produced a parseable confidence value, while internal probability again yielded above-chance error detection (AUROC 0.54-0.92). Both models fail in the same place: under severe underexposure, accuracy collapses (0.99->0.22 for Qwen2-VL, 0.97->0.42 for SmolVLM) while both confidence signals barely move, and internal error-detection falls to chance. We conclude that small VLMs encode usable self-knowledge that their verbalized output does not express, that internal probability is therefore the better deferral signal in constrained deployment, and that neither signal should be trusted under severe low-light conditions.
Thank you to arXiv for use of its open access interoperability. PaperDance 不是 arXiv 官方产品；中文卡片由大模型生成，请以原文为准。