PathGLS: 通过多维一致性评估病理视觉-语言模型无需真实数据
PathGLS: Evaluating Pathology Vision-Language Models without Ground Truth through Multi-Dimensional Consistency
AI总结：
本文提出PathGLS，一种无需真实数据的病理VLM评估框架，从 grounding、logic 和 stability 三个维度评估模型性能，通过实验验证其在多个数据集上的优越性，显著优于其他方法。
摘要：
视觉-语言模型（VLMs）在计算病理学中具有巨大潜力，能够实现可解释的图像分析、自动化报告和可扩展的决策支持。然而，其广泛临床应用受限于缺乏可靠、自动化的评估指标，无法识别细微的失败如幻觉。为此，我们提出PathGLS，一种新颖的参考自由评估框架，通过三个维度评估病理VLMs：Grounding（细粒度视觉-文本对齐）、Logic（使用自然语言推理的蕴含图一致性）和 Stability（对抗性视觉-语义扰动下的输出方差）。PathGLS支持片段级和整张滑动图像（WSI）级分析，产生综合信任分数。在Quilt-1M、TCGA、REG2025、PathMMU和TCGA-Sarcoma数据集上的实验表明PathGLS的优越性。具体而言，在Quilt-1M数据集上，PathGLS揭示幻觉报告的敏感度下降40.2%，而BERTScore仅为2.1%。验证专家定义的临床错误层次结构，PathGLS达到强的Spearman's rank相关性（ρ=0.71（p < 0.0001）），显著优于基于大语言模型（LLM）的方法（Gemini 3.0 Pro: ρ=0.39，p < 0.0001）。这些结果确立PathGLS为稳健的参考自由度量。通过直接量化幻觉率和领域转移鲁棒性，它为在私人临床数据集上基准测试VLMs和安全部署提供可靠标准。代码可在：https://github.com/My13ad/PathGLS找到。
英文摘要：
Vision-Language Models (VLMs) offer significant potential in computational pathology by enabling interpretable image analysis, automated reporting, and scalable decision support. However, their widespread clinical adoption remains limited due to the absence of reliable, automated evaluation metrics capable of identifying subtle failures such as hallucinations. To address this gap, we propose PathGLS, a novel reference-free evaluation framework that assesses pathology VLMs across three dimensions: Grounding (fine-grained visual-text alignment), Logic (entailment graph consistency using Natural Language Inference), and Stability (output variance under adversarial visual-semantic perturbations). PathGLS supports both patch-level and whole-slide image (WSI)-level analysis, yielding a comprehensive trust score. Experiments on Quilt-1M, TCGA, REG2025, PathMMU and TCGA-Sarcoma datasets demonstrate the superiority of PathGLS. Specifically, on the Quilt-1M dataset, PathGLS reveals a steep sensitivity drop of 40.2% for hallucinated reports compared to only 2.1% for BERTScore. Moreover, validation against expert-defined clinical error hierarchies reveals that PathGLS achieves a strong Spearman's rank correlation of $ρ=0.71$ ($p < 0.0001$), significantly outperforming Large Language Model (LLM)-based approaches (Gemini 3.0 Pro: $ρ=0.39$, $p < 0.0001$). These results establish PathGLS as a robust reference-free metric. By directly quantifying hallucination rates and domain shift robustness, it serves as a reliable criterion for benchmarking VLMs on private clinical datasets and informing safe deployment. Code can be found at: https://github.com/My13ad/PathGLS
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。
