簡易檢索 / 詳目顯示
| 研究生： |
陳彥彤 Chen, Yan-Tong |
|---|---|
| 論文名稱： |
結合自我主權身分與大型語言模型之隱私保護保險理賠驗證系統 A Privacy-Preserving Insurance Claim Verification System Integrating Self-Sovereign Identity and Large Language Models |
| 指導教授： | 莊豐源 |
| 口試委員: |
莊豐源
Chuang, Frank 向倩儀 Hsiang, Chien-Yi 陳柏安 Chen, Po-An |
| 學位類別： |
碩士
Master |
| 系所名稱： |
商學院 - 資訊管理學系 Department of Management Information System |
| 論文出版年： | 2026 |
| 畢業學年度： | 114 |
| 語文別： | 中文 |
| 論文頁數： | 97 |
| 中文關鍵詞： | 自我主權身分 、可驗證憑證 、大型語言模型 、最小化揭露 、證明閘控驗證 、證據準備度 、受控多輪補件 |
| 外文關鍵詞： | Self-Sovereign Identity, Verifiable Credentials, Large Language Models, Minimal Disclosure, Proof-Gated Verification, Evidence Readiness, Controlled Multi-roundEvidence Supplementation |
| 相關次數： | 點閱：7 下載：0 |
| 分享至: |
| 查詢本校圖書館目錄 查詢臺灣博碩士論文知識加值系統 勘誤回報 |
保險前置流程，尤其是理賠前證據處理需交換醫療、財務、保單資格與文件等敏感證據。現行流程若只確認文件是否存在，仍可能忽略內容條件、跨來源關係及過度揭露問題。自我主權身分（Self-Sovereign Identity, SSI）與可驗證憑證可支援可信且由持有者控制的證據交換，但不會自動將自然語言保單條款轉換為特定任務所需的證明需求，也未直接處理證據不足後的局部補件與安全停止。
本研究提出一套基於 SSI 的保險證據準備度驗證框架，涵蓋三個相互連接的問題：保單條款至結構化證明需求與最小化揭露證明請求的轉換、證明閘控的證據準備度判定，以及受控多輪證據補件。大型語言模型負責產生候選證明需求與合成測試藍圖；需求是否可執行、測試案例的預期結果、證據是否就緒，以及流程是否補件、停止或轉交人工，均由確定性元件控制。原型整合 Python 代理程式、ACA-Py、AnonCreds、DIDComm 與 SQLite，並以受控功能測試、端到端 SSI 情境及可重播確認性實驗進行評估。
實驗涵蓋三張研究用途保單。RQ1 顯示，資料結構引導可改善必要憑證、一致性檢查與揭露政策，但請求屬性、條件證明及保單特定文件仍需語意驗證；14 案消融中，完整驗證正確率為 1.0000，停用後降為 0.5000。25 個揭露案例皆維持必要聲明完整性與任務成功，平均揭露項目減少 13.76% 至 17.33%，且未暴露敏感原始值。RQ2 的 30 案 SSI 整合測試及三組共 135 案確認性資料中，證明閘控驗證器均未產生錯誤就緒；存在性基準則分別產生 14 個及 90 個錯誤就緒。語意突變測試中，LLM 組與兩個受限制隨機組分別殺死 10、12與 11 個突變體，三組聯集達 13/13，顯示兩種測試生成策略具有互補性。RQ3 顯示，局部補件將後續請求項目由 72 降至 28，減少 61.11%；八個端到端情境各重複十次，共 80 次執行，皆符合預期終止與安全升級行為。
研究結果顯示，保險證據處理應同時驗證證據存在、內容條件與跨來源一致性，並由完整診斷驅動局部補件與安全停止。本研究整合大型語言模型的條款理解與測試探索能力、SSI 證明交換、最小化揭露、確定性驗證及受控多輪流程，形成可執行、可診斷、可審計且可重播的證據準備度框架。相關結論限於三張保單、九項凍結證明需求、十三個語意突變體及受控測試環境；證據準備度不等同於正式承保、拒保、理賠核定或給付決策。
Insurance underwriting and pre-claim workflows require the exchange of sensitive medical, financial, policy-eligibility, and documentary evidence. A process that checks only whether credentials or documents are present may still overlook failed conditions, inconsistent relationships across evidence sources, and excessive disclosure. Self-Sovereign Identity (SSI) and Verifiable Credentials enable trustworthy, holder-controlled evidence exchange, but they do not automatically translate naturallanguage insurance policies into task-specific proof requirements or determine how incomplete evidence should be supplemented and safely terminated.
This thesis presents an SSI-based insurance evidence-readiness verification framework that addresses three connected problems: transforming policy clauses into structured proof requirements and minimally disclosing proof requests, performing proof-gated evidence-readiness verification, and controlling multi-round evidence supplementation. A large language model generates candidate proof requirements and synthetic test blueprints. Deterministic components decide whether requirements are executable, derive expected test outcomes, evaluate evidence readiness, and control supplementation, termination, and escalation. The prototype integrates a Python verification service, ACA-Py, AnonCreds, DIDComm, and SQLite, and is evaluated through controlled functional tests, end-to-end SSI scenarios, and replayable confirmatory experiments.
The evaluation covers three research insurance policies. For RQ1, schema-guided generation improved required credentials, consistency checks, and disclosure policies, while requested attributes, predicates, and policy-specific documents still required semantic validation. In a 14-case ablation, the complete validation procedure achieved 1.0000 accuracy, compared with 0.5000 when semantic validation was disabled. Across 25 disclosure cases, required-claim completeness and task success were preserved, while the average number of disclosed items decreased by 13.76% to 17.33%, with no raw sensitive-value exposure. For RQ2, the proof-gated verifier produced no false-ready decisions in a 30-case SSI integration test or in three confirmatory datasets totaling 135 cases. The presenceonly baseline produced 14 and 90 false-ready decisions, respectively. In semantic mutation testing, the LLM-generated suite and two constrained-random suites killed 10, 12, and 11 mutants; their union killed all 13 mutants, indicating complementary fault-detection patterns. For RQ3, localized supplementation reduced follow-up request items from 72 to 28, a 61.11% reduction. Eight end-toend scenarios were each repeated ten times, and all 80 executions reached the expected termination and escalation outcomes.
The results show that insurance evidence processing should jointly verify evidence presence, content conditions, and cross-source consistency, and should use complete diagnostics to drive localized supplementation and safe termination. By combining LLM-assisted policy interpretation and test exploration, SSI proof exchange, minimal disclosure, deterministic verification, and controlled multi-round workflows, this thesis provides an executable, diagnosable, auditable, and replayable evidence-readiness framework. The findings are limited to three policies, nine frozen proof requirements, thirteen semantic mutants, and controlled test environments. Evidence readiness does not constitute a formal underwriting, rejection, claim-adjudication, or payment decision.
致謝 i
摘要 iii
Abstract v
第一章 緒論 1
1.1 研究背景與動機 1
1.2 研究目的 3
1.3 研究問題 3
1.4 研究範圍與限制 4
1.5 研究貢獻 5
1.6 論文架構 6
第二章 文獻回顧 9
2.1 保險證據流程與資料互通 9
2.1.1 文件數位化與保險證據處理 9
2.1.2 FHIR 與跨機構資料互通 10
2.1.3 既有數位化流程的能力邊界 10
2.2 SSI、VC 與選擇性揭露技術 10
2.2.1 可驗證憑證資料模型與交換協定 10
2.2.2 選擇性揭露、Predicate 與撤銷 11
2.2.3 密碼學能力與保險需求規劃之落差 11
2.3 LLM 輔助需求抽取與正式化 12
2.3.1 文件理解、檢索與結構化生成 12
2.3.2 受約束解碼與語意解析 12
2.3.3 LLM-to-formal 與確定性執行 13
2.4 證據驗證、工具型代理人與可靠工作流程 13
2.4.1 從密碼學有效性到任務層充分性 13
2.4.2 測試生成、獨立 Oracle 與語意突變測試 14
2.4.3 多輪工具使用與狀態依賴 15
2.4.4 故障復原、動作風險與受約束控制 15
2.5 文獻綜整與研究缺口 15
2.5.1 現有技術能力比較 15
2.5.2 三項研究缺口 17
2.5.3 本研究定位與用語邊界 17
2.6 本章小結 18
第三章 研究設計與方法架構 19
3.1 研究設計與任務範圍 20
3.1.1 研究問題與方法對應 20
3.1.2 保險證據處理任務 21
3.1.3 研究方法的責任邊界 21
3.2 系統整體架構與角色分工 21
3.2.1 SSI 參與角色 22
3.2.2 邏輯元件與資料流 22
3.2.3 三層互補評估環境 23
3.3 可驗證憑證與證據資料模型 24
3.3.1 憑證類型與代表性欄位 24
3.3.2 跨來源一致性關係 24
3.3.3 結構化 SSI 證明需求 25
3.4 結構化證明需求生成與驗證 26
3.4.1 候選需求生成 26
3.4.2 條件抽取與資料結構映射 26
3.4.3 結構與語意驗證程序 27
3.5 最小化揭露與證明規劃 27
3.5.1 揭露策略 27
3.5.2 證明請求與呈現 28
3.5.3 最小化揭露的驗證條件 28
3.6 證明閘控的證據準備度驗證 28
3.6.1 證據集合與檢查層次 28
3.6.2 狀態判定、完整診斷與失敗關閉原則 28
3.6.3 證明閘控的證據準備度驗證方法 29
3.6.4 存在性基準方法 30
3.7 確認性合成測試與語意突變評估方法 31
3.7.1 凍結需求、案例配額與生成範圍 31
3.7.2 LLM 與受限制隨機測試生成 31
3.7.3 確定性具體化與獨立 Oracle 32
3.7.4 驗證器語意突變測試 32
3.7.5 最小測試案例組合 32
3.7.6 不可變實驗凍結與重播 33
3.8 受控多輪證據補件與安全控制 34
3.8.1 缺漏診斷與局部補件計畫 34
3.8.2 跨回合狀態模型 34
3.8.3 狀態持久化與中斷後復原 35
3.8.4 重試、退避與資源預算 35
3.8.5 衝突防護與安全升級 36
3.8.6 受控多輪補件演算法 36
3.8.7 流程不變量與安全性約束 38
3.9 保險驗證任務與比較基準 38
3.9.1 核保前證據驗證 38
3.9.2 理賠前證據準備度 38
3.9.3 三個研究問題的比較與消融基準 39
3.10 LLM、自動化控制與可審計性 39
3.10.1 LLM、確定性元件與人工審查 39
3.10.2 可審計紀錄 39
3.11 本章小結 40
第四章 原型系統實作 43
4.1 原型範圍與整體架構 43
4.1.1 實作範圍與責任邊界 44
4.1.2 系統分層與主要元件 45
4.1.3 端到端處理流程 45
4.1.4 部署範圍與技術限制 46
4.2 核心資料物件與憑證資料模型 46
4.2.1 結構化 SSI 證明需求 46
4.2.2 憑證與證據物件 48
4.2.3 驗證報告與補件狀態 49
4.2.4 合成測試案例與實驗凍結物件 49
4.3 結構化證明需求生成與驗證 50
4.3.1 任務導向與資料結構引導抽取 50
4.3.2 結構與語意驗證 50
4.3.3 需求轉換與執行邊界 50
4.4 最小化揭露編譯 51
4.4.1 揭露政策與最小化呈現規格 51
4.4.2 呈現資料與原始值抑制 51
4.4.3 揭露成本與選擇性揭露評估 52
4.5 證明閘控的證據準備度驗證 52
4.5.1 檢查層次與診斷集合 52
4.5.2 失敗關閉的狀態判定 53
4.5.3 憑證／文件存在性基準 54
4.6 確認性合成評估與實驗重播工具鏈 55
4.6.1 測試藍圖驗證與受控替換 55
4.6.2 確定性具體化與 Oracle 實作 56
4.6.3 基準測試、生成器比較與突變執行器 56
4.6.4 最小測試案例組合、實驗凍結與結果產生 56
4.7 受控多輪證據補件與安全控制 57
4.7.1 缺漏診斷與局部補件計畫 57
4.7.2 多輪補件狀態機 58
4.7.3 狀態持久化、冪等性與中斷後恢復 58
4.7.4 重試、退避與資源預算 58
4.7.5 衝突防護與安全停止 59
4.7.6 ACA-Py 與 DIDComm 控制器協調 60
4.7.7 實作不變量與安全性約束 62
4.8 應用程式介面、實驗工具與自動化測試 62
4.8.1 Python 驗證服務介面與輸出契約 62
4.8.2 控制器事件與跨服務介面 63
4.8.3 確認性實驗命令列介面 64
4.8.4 自動化測試分類 64
4.9 本章小結 64
第五章 實驗結果與分析 67
5.1 評估設計 67
5.1.1 研究問題與實驗對應 67
5.1.2 資料與測試設計 67
5.1.3 實驗環境 68
5.1.4 評估指標 68
5.2 RQ1：LLM 輔助的 SSI 證明需求與最小化揭露 69
5.2.1 資料結構引導提示比較 69
5.2.2 跨保單元件層級結果 69
5.2.3 語意驗證消融 70
5.2.4 最小化揭露編譯 70
5.2.5 編譯式最小化揭露結果 70
5.2.6 選擇性揭露消融 71
5.2.7 RQ1 結論 71
5.3 RQ2：證明閘控的證據準備度驗證 72
5.3.1 主要基準比較 72
5.3.2 條件證明驗證消融實驗 72
5.3.3 跨來源一致性驗證消融實驗 73
5.3.4 條件證明邊界與失敗關閉測試 73
5.3.5 確認性資料集建構與凍結 74
5.3.6 135 案確認性基準比較 74
5.3.7 生成器涵蓋率與案例特性 75
5.3.8 語意突變測試與生成器互補性 75
5.3.9 最小測試案例組合 76
5.3.10 RQ2 結論 77
5.4 RQ3：受控多輪 SSI 證據補件與安全控制 77
5.4.1 完整重送與局部補件比較 77
5.4.2 端到端情境與重複試驗 78
5.4.3 RQ3 結論 78
5.5 綜合討論 79
5.6 研究限制與效度威脅 79
5.7 本章小結 79
第六章 結論與未來研究 81
6.1 研究總結 81
6.2 主要研究發現 81
6.2.1 RQ1：LLM 輔助 SSI 證明需求與最小化揭露 81
6.2.2 RQ2：證明閘控的證據準備度驗證 82
6.2.3 RQ3：受控多輪 SSI 證據補件與安全控制 82
6.3 研究貢獻 83
6.3.1 LLM 輔助且由確定性驗證把關的證明需求轉換 83
6.3.2 將最小化揭露納入證明請求編譯 83
6.3.3 可診斷的證明閘控驗證與可重播穩健性評估 83
6.3.4 受控多輪 SSI 補件與安全控制 83
6.4 研究與實務意涵 84
6.5 研究限制 84
6.6 未來研究方向 84
6.7 結語 85
參考文獻 87
附錄 A 研究資料與正式證明需求 91
A.1 正式證明需求資料結構 91
A.2 六項標準答案與九項凍結需求 91
附錄 B 確認性合成測試與語意突變規格 93
B.1 十三個驗證器語意突變體 93
附錄 C 存在性基準的代表性錯誤就緒案例 95
附錄 D 受控多輪 SSI 證據補件情境 97
D.1 端到端情境 97
Allen, C. (2016). The path to self-sovereign identity [Accessed 2026-06-27]. https://www.lifewithalacrity.com/article/the-path-to-self-soverereign-identity/
AnonCreds Working Group. (2026). Anoncreds specification [Version 1.0 Draft; Accessed July 2026]. https://hyperledger.github.io/anoncreds-spec/
Barr, E. T., Harman, M., McMinn, P., Shahbaz, M., & Yoo, S. (2015). The oracle problem in software testing: A survey. IEEE Transactions on Software Engineering, 41(5), 507–525. https://doi.org/10.1109/TSE.2014.2372785
Bünz, B., Bootle, J., Boneh, D., Poelstra, A., Wuille, P., & Maxwell, G. (2018). Bulletproofs: Short proofs for confidential transactions and more. 2018 IEEE Symposium on Security and Privacy, 315–334. https://doi.org/10.1109/SP.2018.00020
Camenisch, J., & Lysyanskaya, A. (2001). An efficient system for non-transferable anonymous credentials with optional anonymity revocation. Advances in Cryptology — EUROCRYPT 2001, 2045, 93–118. https://doi.org/10.1007/3-540-44987-6_7
Debenedetti, E., Zhang, J., Balunovic, M., Beurer-Kellner, L., Fischer, M., & Tramèr, F. (2024). Agentdojo: A dynamic environment to evaluate prompt injection attacks and defenses for LLM agents. Advances in Neural Information Processing Systems, Datasets and Benchmarks Track, 37. https://doi.org/10.52202/079017-2636
Fett, D., Yasuda, K., & Campbell, B. (2025, November). Selective disclosure for json web tokens (RFC No. 9901). Internet Engineering Task Force. https://doi.org/10.17487/RFC9901
Fraser, G., Staats, M., McMinn, P., Arcuri, A., & Padberg, F. (2015). Does automated unit test generation really help software testers? a controlled empirical study. ACM Transactions on Software Engineering and Methodology, 24(4), Article 23, 1–49. https://doi.org/10.1145/2699688
Gao, L., Madaan, A., Zhou, S., Alon, U., Liu, P., Yang, Y., Callan, J., & Neubig, G. (2023). PAL: Program-aided language models. Proceedings of the 40th International Conference on Machine Learning, 202, 10764–10799.
Garcia-Molina, H., & Salem, K. (1987). Sagas. Proceedings of the 1987 ACM SIGMOD International Conference on Management of Data, 249–259. https://doi.org/10.1145/38713.38742
Harrold, M. J., Gupta, R., & Soffa, M. L. (1993). A methodology for controlling the size of a test suite. ACM Transactions on Software Engineering and Methodology, 2(3), 270–285. https://doi.org/10.1145/152388.152391
HL7 International. (2023). FHIR release 5 [Version 5.0.0]. https://hl7.org/fhir/R5/
Ji, Z., Lee, N., Frieske, R., Yu, T., Su, D., Xu, Y., Ishii, E., Bang, Y. J., Madotto, A., & Fung, P. (2023). Survey of hallucination in natural language generation. ACM Computing Surveys, 55(12), 1–38. https://doi.org/10.1145/3571730
Jia, Y., & Harman, M. (2011). An analysis and survey of the development of mutation testing. IEEE Transactions on Software Engineering, 37(5), 649–678. https://doi.org/10.1109/TSE.2010.62
Jiang, F., Jiang, Y., Zhi, H., Dong, Y., Li, H., Ma, S., Wang, Y., Dong, Q., Shen, H., & Wang, Y. (2017). Artificial intelligence in healthcare: Past, present and future. Stroke and Vascular Neurology, 2(4), 230–243. https://doi.org/10.1136/svn-2017-000101
Laranjo, L., Dunn, A. G., Tong, H. L., Kocaballi, A. B., Chen, J., Bashir, R., Surian, D., Gallego, B., Magrabi, F., Lau, A. Y. S., & Coiera, E. (2018). Conversational agents in healthcare: A systematic review. Journal of the American Medical Informatics Association, 25(9), 1248–1258. https://doi.org/10.1093/jamia/ocy072
Lavin, R., Liu, X., Mohanty, H., Norman, L., Zaarour, G., & Krishnamachari, B. (2024). A survey on the applications of zero-knowledge proofs. arXiv preprint arXiv:2408.00243. https://arxiv.org/abs/2408.00243
Lemieux, C., Inala, J. P., Lahiri, S. K., & Sen, S. (2023). CodaMOSA: Escaping coverage plateaus in test generation with pre-trained large language models. Proceedings of the 45th IEEE/ACM International Conference on Software Engineering, 919–931. https://doi.org/10.1109/ICSE48619.2023.00085
Lewis, P., Perez, E., Piktus, A., Petroni, F., Karpukhin, V., Goyal, N., Küttler, H., Lewis, M., Yih, W.-t., Rocktäschel, T., Riedel, S., & Kiela, D. (2020). Retrieval-augmented generation for knowledge-intensive NLP tasks. Advances in Neural Information Processing Systems, 33, 9459–9474. https://arxiv.org/abs/2005.11401
Liu, X., Yu, H., Zhang, H., Xu, Y., Lei, X., Lai, H., Gu, Y., Ding, H., Men, K., Yang, K., Zhang, S., Deng, X., Zeng, A., Du, Z., Zhang, C., Shen, S., Zhang, T., Su, Y., Sun, H., … Tang, J. (2024). Agentbench: Evaluating LLMs as agents. International Conference on Learning Representations. https://proceedings.iclr.cc/paper_files/paper/2024/hash/e9df36b21ff4ee211a8b71ee8b7e9f57-Abstract-Conference.html
Lodderstedt, T., Yasuda, K., Looker, T., & Bastian, P. (2025). Openid for verifiable credential issuance 1.0. https://openid.net/specs/openid-4-verifiable-credential-issuance-1_0.html
Lu, J., Holleis, T., Zhang, Y., Aumayer, B., Nan, F., Bai, H., Ma, S., Ma, S., Li, M., Yin, G., Wang, Z., & Pang, R. (2025). ToolSandbox: A stateful, conversational, interactive evaluation benchmark for LLM tool use capabilities. Findings of the Association for Computational Linguistics: NAACL 2025, 1160–1183. https://doi.org/10.18653/v1/2025.findings-naacl.65
Madaan, A., Tandon, N., Gupta, P., Hallinan, S., Gao, L., Wiegreffe, S., Alon, U., Dziri, N., Prabhumoye, S., Yang, Y., Gupta, S., Majumder, B. P., Hermann, K., Welleck, S., Yazdanbakhsh, A., & Clark, P. (2023). Self-refine: Iterative refinement with self-feedback. Advances in Neural Information Processing Systems, 36.
Mandel, J. C., Kreda, D. A., Mandl, K. D., Kohane, I. S., & Ramoni, R. B. (2016). SMART on FHIR: A standards-based, interoperable apps platform for electronic health records. Journal of the American Medical Informatics Association, 23(5), 899–908. https://doi.org/10.1093/jamia/ocv189
Milne-Ives, M., de Cock, C., Lim, E., Shehadeh, M. H., de Pennington, N., Mole, G., Normando, E., & Meinert, E. (2020). The effectiveness of artificial intelligence conversational agents in health care: Systematic review. Journal of Medical Internet Research, 22(10), e20346. https://doi.org/10.2196/20346
Moradi Dakhel, A., Nikanjam, A., Majdinasab, V., Khomh, F., & Desmarais, M. C. (2024). Effective test generation using pre-trained large language models and mutation testing. Information and Software Technology, 171, 107468. https://doi.org/10.1016/j.infsof.2024.107468
Mühle, A., Grüner, A., Gayvoronskaya, T., & Meinel, C. (2018). A survey on essential components of a self-sovereign identity. Computer Science Review, 30, 80–86. https://doi.org/10.1016/j.cosrev.2018.10.002
Nissenbaum, H. (2004). Privacy as contextual integrity. Washington Law Review, 79(1), 119–158. https://digitalcommons.law.uw.edu/wlr/vol79/iss1/10
Oh, S.-R., Seo, Y.-D., Lee, E., & Kim, Y.-G. (2021). A comprehensive survey on security and privacy for electronic health data. International Journal of Environmental Research and Public Health, 18(18), 9668. https://doi.org/10.3390/ijerph18189668
Pan, L., Albalak, A., Wang, X., & Wang, W. Y. (2023). Logic-LM: Empowering large language models with symbolic solvers for faithful logical reasoning. Findings of the Association for Computational Linguistics: EMNLP 2023, 3806–3824. https://doi.org/10.18653/v1/2023.findings-emnlp.248
Papadakis, M., Kintis, M., Zhang, J., Jia, Y., Le Traon, Y., & Harman, M. (2019). Mutation testing advances: An analysis and survey. In A. M. Memon (Ed.), Advances in computers (pp. 275–378, Vol. 112). Elsevier. https://doi.org/10.1016/bs.adcom.2018.03.015
Ruan, Y., Dong, H., Wang, A., Pitis, S., Zhou, Y., Ba, J., Dubois, Y., Maddison, C., & Hashimoto, T. (2024). Identifying the risks of LM agents with an LM-emulated sandbox. International Conference on Learning Representations.
Scholak, T., Schucher, N., & Bahdanau, D. (2021). PICARD: Parsing incrementally for constrained auto-regressive decoding from language models. Proceedings of the 2021 Conference on Empirical Methods in Natural Language Processing, 9895–9901. https://doi.org/10.18653/v1/2021.emnlp-main.779
Shin, R., Lin, C., Thomson, S., Chen, C., Roy, S., Platanios, E. A., Pauls, A., Klein, D., Eisner, J., & Van Durme, B. (2021). Constrained language models yield few-shot semantic parsers. Proceedings of the 2021 Conference on Empirical Methods in Natural Language Processing, 7699–7715. https://doi.org/10.18653/v1/2021.emnlp-main.608
Stokkink, Q., & Pouwelse, J. (2018). Deployment of a blockchain-based self-sovereign identity. 2018 IEEE International Conference on Blockchain, 1336–1342. https://doi.org/10.1109/Cybermatics_2018.2018.00230
Terbu, O., Lodderstedt, T., Yasuda, K., Fett, D., & Heenan, J. (2025). Openid for verifiable presentations 1.0. https://openid.net/specs/openid-4-verifiable-presentations-1_0.html
Villegas-Ortega, J., Bellido-Boza, L., & Mauricio, D. (2021). Fourteen years of manifestations and factors of health insurance fraud, 2006–2020: A scoping review. Health & Justice, 9, Article 26. https://doi.org/10.1186/s40352-021-00149-3
W3C Decentralized Identifier Working Group. (2022). Decentralized identifiers (DIDs) v1.0 (W3C Recommendation) (W3C Recommendation, 19 July 2022). World Wide Web Consortium. https://www.w3.org/TR/did-core/
W3C Verifiable Credentials Working Group. (2025). Verifiable credentials data model v2.0 [15 May 2025]. https://www.w3.org/TR/vc-data-model-2.0/
W3C Verifiable Credentials Working Group. (2026). Data integrity BBS cryptosuites v1.0 [7 April 2026]. https://www.w3.org/TR/vc-di-bbs/
Wang, X., Wang, Z., Liu, J., Chen, Y., Yuan, L., Peng, H., & Ji, H. (2024). MINT: Evaluating LLMs in multi-turn interaction with tools and language feedback. International Conference on Learning Representations.
World Health Organization. (2021). Ethics and governance of artificial intelligence for health. https://www.who.int/publications/i/item/9789240029200
World Health Organization. (2024). Ethics and governance of artificial intelligence for health: Guidance on large multi-modal models [Accessed 2026-07-11]. https://www.who.int/publications/i/item/9789240084759
Xu, Y., Xu, Y., Lv, T., Cui, L., Wei, F., Wang, G., Lu, Y., Florencio, D., Zhang, C., Che, W., Zhang, M., & Zhou, L. (2021). LayoutLMv2: Multi-modal pre-training for visually-rich document understanding. Proceedings of the 59th Annual Meeting of the Association for Computational Linguistics and the 11th International Joint Conference on Natural Language Processing, 2579–2591. https://doi.org/10.18653/v1/2021.acl-long.201
Yao, S., Shinn, N., Razavi, P., & Narasimhan, K. (2024). τ-bench: A benchmark for tool-agent-user interaction in real-world domains. https://arxiv.org/abs/2406.12045
Yao, S., Zhao, J., Yu, D., Du, N., Shafran, I., Narasimhan, K., & Cao, Y. (2023). ReAct: Synergizing reasoning and acting in language models. International Conference on Learning Representations.
Zheng, H., You, L., & Hu, G. (2022). A novel insurance claim blockchain scheme based on zero-knowledge proof technology. Computer Communications, 195, 207–216. https://doi.org/10.1016/j.comcom.2022.08.007
中華民國法務部. (2025). 保險法（第六十四條）[修正日期：民國 114 年 6 月 18 日；Accessed: 2026-07]. https://law.fsc.gov.tw/LawContent.aspx?id=FL006746
全文公開日期 2030/08/25