近年來，代理式方法逐漸被應用於病理影像分析中，透過逐步選取感興趣區域、觀察局部影像內容，並整合多步驟資訊產生最終診斷結果。然而，現有評估方式多半仍以最終答案正確率為主，難以反映代理式方法在全切片影像中的探索行為、區域選擇品質與證據收集能力。尤其當不同代理式病理分析方法採用不同倍率、不同區域大小與不同探索策略時，僅以準確率進行比較，無法充分說明模型是否真正有效地定位病理相關區域，或是在未實際觀察關鍵區域的情況下，仍產生看似正確的答案。
為了解決上述問題，本研究提出一套用於病理影像中代理式方法之統一評估框架。此框架將不同方法的推論過程轉換為一致的路徑表示，使各方法在區域選擇、觀察內容與最終回答上能被置於相同基準下進行比較。基於此表示方式，本研究進一步設計多個輔助評估指標，以補足僅依賴最終答案正確率所無法呈現的資訊，包括衡量早期定位能力、評估病理區域覆蓋與選擇品質，以及衡量探索成本。這些指標可用於分析模型是否能快速找到診斷相關區域、是否能充分覆蓋重要病理證據，以及是否在過程中產生過度探索的成本。
實驗部分，本研究在多個病理任務上比較不同代理式病理分析方法，包含腫瘤偵測、腫瘤分類與腫瘤亞型分類。實驗結果顯示，不同方法在最終答案表現、早期定位能力、感興趣區域覆蓋程度與探索成本之間呈現不同取捨；單一準確率無法完整描述這些差異。部分方法能較早抵達病理相關區域，但未必能維持較高的區域選取精確度；另一些方法雖投入較高的探索成本，且在探索過程中具有較高的區域選取精確度，但這些優勢未必能轉化為更好的回答準確率。這些結果說明，代理式病理方法的能力應從多個面向共同評估，而非僅依賴最終預測結果。
整體而言，本研究建立了一個相對標準化且可比較的基準評估框架，使具備多倍率探索與序列式區域觀察能力的代理式病理分析方法，能在一致基準下進行分析。此框架有助於揭示準確率以外的模型行為差異，並提供更完整的依據來理解不同方法在全切片探索、感興趣區域定位能力與探索效率之間的優劣與限制。

In recent years, agentic approaches have gradually been introduced into computational pathology, where models iteratively select regions of interest (ROI), examine local image content, and integrate multi-step observations to generate final diagnostic predictions. However, existing evaluation protocols still primarily rely on final-answer accuracy, making it difficult to reflect the exploration behavior, region selection quality, and evidence collection capability of agentic methods in whole slide images (WSIs). In particular, when different methods employ different magnifications, region sizes, and exploration strategies, accuracy alone cannot adequately determine whether a model truly localizes pathology-relevant regions effectively, or instead produces seemingly correct answers without actually observing critical diagnostic areas.
To address these limitations, this study proposes a unified navigation and evidence evaluation framework for agentic pathology methods. The proposed framework transforms the inference process of different methods into a standardized trajectory representation, allowing region selection, intermediate observations, and final predictions to be compared under a consistent evaluation setting. Based on this representation, we further introduce several complementary evaluation metrics to provide additional insights beyond final-answer accuracy, including Mean Reciprocal First Hit for measuring early localization ability, ROI Precision for evaluating pathological region coverage and selection quality, as well as Navigation Cost for quantifying exploration overhead. These metrics enable analysis of whether a model can rapidly identify diagnostically relevant regions, sufficiently cover important pathological evidence, and avoid excessive exploration during navigation.
In the experimental section, this study compares different agentic pathology analysis methods across multiple pathology tasks, including tumor detection, tumor classification, and tumor subtype classification. The experimental results show that different methods exhibit distinct trade-offs among final-answer performance, early localization ability, region-of-interest coverage, and navigation cost; therefore, accuracy alone cannot fully characterize these differences. Some methods are able to reach pathology relevant regions earlier, but they do not necessarily maintain higher region selection precision. Other methods require higher navigation costs and show higher region selection precision during exploration, yet these advantages do not necessarily translate into better final answer accuracy. These findings indicate that the capability of agentic pathology methods should be evaluated from multiple perspectives rather than relying solely on final prediction results.
Overall, this study establishes a relatively standardized and comparable benchmark framework for agentic pathology methods with multi-scale exploration and sequential region-observation capabilities. The proposed framework helps reveal behavioral differences beyond accuracy and provides a more comprehensive basis for analyzing the strengths and limitations of different methods in terms of WSI navigation, ROI localization, and exploration efficiency.

[1] Maximilian Ilse, Jakub Tomczak, and MaxWelling. Attention-based deep multipleinstance learning. In International conference on machine learning, pages 2127–2136. PMLR, 2018.
[2] Ming Y Lu, Drew FK Williamson, Tiffany Y Chen, Richard J Chen, Matteo Barbieri,and Faisal Mahmood. Data-efficient and weakly supervised computationalpathology on whole-slide images. Nature biomedical engineering, 5(6):555–570,2021.
[3] Bin Li, Yin Li, and Kevin W Eliceiri. Dual-stream multiple instance learning network for whole slide image classification with self-supervised contrastive learning. In Proceedings of the IEEE/CVF conference on computer vision and pattern recognition, pages 14318–14328, 2021.
[4] Richard J Chen, Tong Ding, Ming Y Lu, Drew FK Williamson, Guillaume Jaume, Andrew H Song, Bowen Chen, Andrew Zhang, Daniel Shao, Muhammad Shaban, et al. Towards a general-purpose foundation model for computational pathology. Nature medicine, 30(3):850–862, 2024.
[5] Hanwen Xu, Naoto Usuyama, Jaspreet Bagga, Sheng Zhang, Rajesh Rao, Tristan Naumann, Cliff Wong, Zelalem Gero, Javier Gonz´alez, Yu Gu, et al. A whole-slide foundation model for digital pathology from real-world data. Nature, 630(8015):181–188, 2024.
[6] Eric Zimmermann, Eugene Vorontsov, Julian Viret, Adam Casson, Michal Zelechowski, George Shaikovski, Neil Tenenholtz, James Hall, David Klimstra, Razik Yousfi, et al. Virchow2: Scaling self-supervised mixed magnification models in pathology. arXiv preprint arXiv:2408.00738, 2024.
[7] Xiyue Wang, Junhan Zhao, Eliana Marostica, Wei Yuan, Jietian Jin, Jiayu Zhang, Ruijiang Li, Hongping Tang, Kanran Wang, Yu Li, et al. A pathology foundation model for cancer diagnosis and prognosis prediction. Nature, 634(8035):970–978, 2024.
[8] Mehmet Saygin Seyfioglu, Wisdom O Ikezogwo, Fatemeh Ghezloo, Ranjay Krishna, and Linda Shapiro. Quilt-llava: Visual instruction tuning by extracting localized narratives from open-source histopathology videos. In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, pages 13183–13192, 2024.
[9] Yuci Liang, Xinheng Lyu, Wenting Chen, Meidan Ding, Jipeng Zhang, Xiangjian He, Song Wu, Xiaohan Xing, Sen Yang, Xiyue Wang, et al. Wsi-llava: A multimodal large language model for whole slide image. In Proceedings of the IEEE/CVF International Conference on Computer Vision, pages 22718–22727, 2025.
[10] Wenchuan Zhang, Penghao Zhang, Jingru Guo, Tao Cheng, Jie Chen, Shuwan Zhang, Zhang Zhang, Yuhao Yi, and Hong Bu. Patho-r1: A multimodal reinforcement learning-based pathology expert reasoner. In Proceedings of the AAAI Conference on Artificial Intelligence, volume 40, pages 28418–28426, 2026.
[11] Jingyun Chen, Linghan Cai, Zhikang Wang, Yi Huang, Songhan Jiang, Shenjin Huang, Hongpeng Wang, and Yongbing Zhang. Pathagent: Toward interpretable analysis of whole-slide pathology images via large language model-based agentic reasoning. arXiv preprint arXiv:2511.17052, 2025.
[12] Sheng Wang, Ruiming Wu, Charles Herndon, Yihang Liu, Shunsuke Koga, Jeanne Shen, and Zhi Huang. Pathology-cot: Learning visual chain-of-thought agent from expert whole slide image diagnosis behavior. arXiv preprint arXiv:2510.04587, 2025.
[13] Luca L Weishaupt, Chengkuan Chen, Drew FK Williamson, Richard J Chen, Guillaume Jaume, Tong Ding, Bowen Chen, Anurag Vaidya, Long Phi Le, Ming Y Lu, et al. Evidence-based diagnostic reasoning with multi-agent copilot for human pathology. arXiv preprint arXiv:2506.20964, 2025.
[14] Yuxuan Sun, Yixuan Si, Chenglu Zhu, Kai Zhang, Zhongyi Shui, Bowen Ding, Tao Lin, and Lin Yang. Cpathagent: An agent-based foundation model for interpretable high-resolution pathology image analysis mimicking pathologists’ diagnostic logic. Advances in Neural Information Processing Systems, 38:101673–101731, 2026.
[15] Pingyi Chen, Chenglu Zhu, Sunyi Zheng, Honglin Li, and Lin Yang. Wsi-vqa: Interpreting whole slide images by generative visual question answering. In European Conference on Computer Vision, pages 401–417. Springer, 2024.
[16] Ying Chen, Guoan Wang, Yuanfeng Ji, Yanjun Li, Jin Ye, Tianbin Li, Ming Hu, Rongshan Yu, Yu Qiao, and Junjun He. Slidechat: A large vision-language assistant for whole-slide pathology image understanding. In Proceedings of the Computer Vision and Pattern Recognition Conference, pages 5134–5143, 2025.
[17] Nick Craswell. Mean reciprocal rank. In Encyclopedia of Database Systems, page 1703. Springer, 2009.
[18] Patrick Lewis, Ethan Perez, Aleksandra Piktus, Fabio Petroni, Vladimir Karpukhin, Naman Goyal, Heinrich K¨uttler, Mike Lewis, Wen-tau Yih, Tim Rockt¨aschel, Sebastian Riedel, and Douwe Kiela. Retrieval-augmented generation for knowledge-intensive nlp tasks. In Advances in Neural Information Processing Systems, volume 33, pages 9459–9474, 2020.
[19] Sewon Min, Kalpesh Krishna, Xinxi Lyu, Mike Lewis, Wen-tau Yih, Pang Koh, Mohit Iyyer, Luke Zettlemoyer, and Hannaneh Hajishirzi. Factscore: Fine-grained atomic evaluation of factual precision in long form text generation. In roceedings of the 2023 Conference on Empirical Methods in Natural Language Processing, pages 12076–12100, 2023.
[20] Wojciech Kryscinski, Bryan McCann, Caiming Xiong, and Richard Socher. Evaluating the factual consistency of abstractive text summarization. In Proceedings of the 2020 Conference on Empirical Methods in Natural Language Processing, pages 9332–9346, 2020.
[21] Alex Wang, Kyunghyun Cho, and Mike Lewis. Asking and answering questions to evaluate the factual consistency of summaries. In Proceedings of the 58th Annual Meeting of the Association for Computational Linguistics, pages 5008–5020, 2020.
[22] James Thorne, Andreas Vlachos, Christos Christodoulopoulos, and Arpit Mittal. Fever: a large-scale dataset for fact extraction and verification. In Proceedings of the 2018 Conference of the North American Chapter of the Association for Com-putational Linguistics: Human Language Technologies, Volume 1 (Long Papers), pages 809–819, 2018.