請用此 Handle URI 來引用此文件：
http://tdr.lib.ntu.edu.tw/jspui/handle/123456789/103764完整後設資料紀錄
| DC 欄位 | 值 | 語言 |
|---|---|---|
| dc.contributor.advisor | 陳倩瑜 | zh_TW |
| dc.contributor.advisor | Chien-Yu Chen | en |
| dc.contributor.author | 林嘉安 | zh_TW |
| dc.contributor.author | Chia-An Lin | en |
| dc.date.accessioned | 2026-08-19T16:32:41Z | - |
| dc.date.available | 2026-08-20 | - |
| dc.date.copyright | 2026-08-19 | - |
| dc.date.issued | 2025 | - |
| dc.date.submitted | 2026-08-10 21:25:36 | - |
| dc.identifier.citation | Alaggio, R., Amador, C., Anagnostopoulos, I., et al. (2022). The 5th edition of the World Health Organization classification of haematolymphoid tumours: Lymphoid neoplasms. Leukemia, 36(7), 1720–1748. https://doi.org/10.1038/s41375-022-01620-2
Asai, A., Wu, Z., Wang, Y., Sil, A., & Hajishirzi, H. (2024). Self-RAG: Learning to retrieve, generate, and critique through self-reflection. International Conference on Learning Representations (ICLR). https://openreview.net/forum?id=hSyW5go0v8 Bolton, E., Venigalla, A., Yasunaga, M., et al. (2024). BioMedLM: A 2.7B parameter language model trained on biomedical text. arXiv preprint arXiv:2403.18421. https://doi.org/10.48550/arXiv.2403.18421 Chen, Z., Cano, A. H., Romanou, A., et al. (2023). MEDITRON-70B: Scaling medical pretraining for large language models. arXiv preprint arXiv:2311.16079. https://doi.org/10.48550/arXiv.2311.16079 Cormack, G. V., Clarke, C. L. A., & Buettcher, S. (2009). Reciprocal rank fusion outperforms Condorcet and individual rank learning methods. Proceedings of the 32nd International ACM SIGIR Conference on Research and Development in Information Retrieval (SIGIR), 758–759. https://doi.org/10.1145/1571941.1572114 Dettmers, T., Pagnoni, A., Holtzman, A., & Zettlemoyer, L. (2023). QLoRA: Efficient finetuning of quantized LLMs. Advances in Neural Information Processing Systems (NeurIPS), 36. https://doi.org/10.48550/arXiv.2305.14314 Devlin, J., Chang, M.-W., Lee, K., & Toutanova, K. (2019). BERT: Pre-training of deep bidirectional transformers for language understanding. Proceedings of the 2019 Conference of the North American Chapter of the Association for Computational Linguistics (NAACL), 4171–4186. https://doi.org/10.18653/v1/N19-1423 Gao, Y., Xiong, Y., Gao, X., et al. (2023). Retrieval-augmented generation for large language models: A survey. arXiv preprint arXiv:2312.10997. https://doi.org/10.48550/arXiv.2312.10997 Gu, Y., Tinn, R., Cheng, H., et al. (2021). Domain-specific language model pretraining for biomedical natural language processing. ACM Transactions on Computing for Healthcare (HEALTH), 3(1), 1–23. https://doi.org/10.1145/3458754 Gu, Z., Churchman, M., Roberts, K., et al. (2016). Genomic analyses identify recurrent MEF2D fusions in acute lymphoblastic leukaemia. Nature Communications, 7, 13331. https://doi.org/10.1038/ncomms13331 Gu, Z., Churchman, M. L., Roberts, K. G., et al. (2019). PAX5-driven subtypes of B-progenitor acute lymphoblastic leukemia. Nature Genetics, 51(2), 296–307. https://doi.org/10.1038/s41588-018-0315-5 Han, T., Adams, L. C., Papaioannou, J.-M., et al. (2023). MedAlpaca – an open-source collection of medical conversational AI models and training data. arXiv preprint arXiv:2304.08247. https://doi.org/10.48550/arXiv.2304.08247 Hirabayashi, S., Ohki, K., Nakabayashi, K., et al. (2017). ZNF384-related fusion genes define a subgroup of childhood B-cell precursor acute lymphoblastic leukemia with a characteristic immunotype. Haematologica, 102(1), 118–129. https://doi.org/10.3324/haematol.2016.151035 Holmfeldt, L., Wei, L., Diaz-Flores, E., et al. (2013). The genomic landscape of hypodiploid acute lymphoblastic leukemia. Nature Genetics, 45(3), 242–252. https://doi.org/10.1038/ng.2532 Holtzman, A., Buys, J., Du, L., Forbes, M., & Choi, Y. (2020). The curious case of neural text degeneration. International Conference on Learning Representations (ICLR). https://openreview.net/forum?id=rygGQyrFvH Huang, L., Yu, W., Ma, W., et al. (2025). A survey on hallucination in large language models: Principles, taxonomy, challenges, and open questions. ACM Transactions on Information Systems, 43(2), 1–55. https://doi.org/10.1145/3703155 Ji, Z., Lee, N., Frieske, R., et al. (2023). Survey of hallucination in natural language generation. ACM Computing Surveys, 55(12), 1–38. https://doi.org/10.1145/3571730 Jiang, A. Q., Sablayrolles, A., Mensch, A., et al. (2023). Mistral 7B. arXiv preprint arXiv:2310.06825. https://doi.org/10.48550/arXiv.2310.06825 Jin, Q., Kim, W., Chen, Q., et al. (2023). MedCPT: Contrastive pre-trained transformers with large-scale PubMed search logs for zero-shot biomedical information retrieval. Bioinformatics, 39(11), btad651. https://doi.org/10.1093/bioinformatics/btad651 Karpukhin, V., Oguz, B., Min, S., et al. (2020). Dense passage retrieval for open-domain question answering. Proceedings of the 2020 Conference on Empirical Methods in Natural Language Processing (EMNLP), 6769–6781. https://doi.org/10.18653/v1/2020.emnlp-main.550 Khattab, O., & Zaharia, M. (2020). ColBERT: Efficient and effective passage search via contextualized late interaction over BERT. Proceedings of the 43rd International ACM SIGIR Conference on Research and Development in Information Retrieval (SIGIR), 39–48. https://doi.org/10.1145/3397271.3401075 Lafferty, J., McCallum, A., & Pereira, F. C. N. (2001). Conditional random fields: Probabilistic models for segmenting and labeling sequence data. Proceedings of the 18th International Conference on Machine Learning (ICML), 282–289. Leaman, R., & Gonzalez, G. (2008). BANNER: An executable survey of advances in biomedical named entity recognition. Pacific Symposium on Biocomputing, 13, 652–663. Lee, J., Yoon, W., Kim, S., et al. (2020). BioBERT: A pre-trained biomedical language representation model for biomedical text mining. Bioinformatics, 36(4), 1234–1240. https://doi.org/10.1093/bioinformatics/btz682 Lewis, P., Perez, E., Piktus, A., et al. (2020). Retrieval-augmented generation for knowledge-intensive NLP tasks. Advances in Neural Information Processing Systems (NeurIPS), 33, 9459–9474. https://doi.org/10.48550/arXiv.2005.11401 Lilljebjörn, H., Henningsson, R., Hyrenius-Wittsten, A., et al. (2016). Identification of ETV6-RUNX1-like and DUX4-rearranged subtypes in paediatric B-cell precursor acute lymphoblastic leukaemia. Nature Communications, 7, 11790. https://doi.org/10.1038/ncomms11790 Liu, N. F., Lin, K., Hewitt, J., et al. (2024). Lost in the middle: How language models use long contexts. Transactions of the Association for Computational Linguistics, 12, 157–173. https://doi.org/10.1162/tacl_a_00638 Liu, Y., Easton, J., Shao, Y., et al. (2017). The genomic landscape of pediatric and young adult T-lineage acute lymphoblastic leukemia. Nature Genetics, 49(8), 1211–1218. https://doi.org/10.1038/ng.3909 Luan, Y., Eisenstein, J., Toutanova, K., & Collins, M. (2021). Sparse, dense, and attentional representations for text retrieval. Transactions of the Association for Computational Linguistics, 9, 329–345. https://doi.org/10.1162/tacl_a_00369 Malkov, Y. A., & Yashunin, D. A. (2020). Efficient and robust approximate nearest neighbor search using hierarchical navigable small world graphs. IEEE Transactions on Pattern Analysis and Machine Intelligence, 42(4), 824–836. https://doi.org/10.1109/TPAMI.2018.2889473 Maynez, J., Narayan, S., Bohnet, B., & McDonald, R. (2020). On faithfulness and factuality in abstractive summarization. Proceedings of the 58th Annual Meeting of the Association for Computational Linguistics (ACL), 1906–1919. https://doi.org/10.18653/v1/2020.acl-main.173 Nogueira, R., Jiang, Z., & Lin, J. (2020). Document ranking with a pretrained sequence-to-sequence model. Findings of the Association for Computational Linguistics: EMNLP 2020, 708–718. https://doi.org/10.18653/v1/2020.findings-emnlp.63 OpenAI. (2023). GPT-4 technical report. arXiv preprint arXiv:2303.08774. https://doi.org/10.48550/arXiv.2303.08774 Roberts, K. G., Li, Y., Payne-Turner, D., et al. (2014). Targetable kinase-activating lesions in Ph-like acute lymphoblastic leukemia. New England Journal of Medicine, 371(11), 1005–1015. https://doi.org/10.1056/NEJMoa1403088 Robertson, S., & Zaragoza, H. (2009). The probabilistic relevance framework: BM25 and beyond. Foundations and Trends in Information Retrieval, 3(4), 333–389. https://doi.org/10.1561/1500000019 Sahoo, P., Singh, A. K., Saha, S., et al. (2024). A systematic survey of prompt engineering in large language models: Techniques and applications. arXiv preprint arXiv:2402.07927. https://doi.org/10.48550/arXiv.2402.07927 Santhanam, K., Khattab, O., Saad-Falcon, J., Potts, C., & Zaharia, M. (2022). ColBERTv2: Effective and efficient retrieval via lightweight late interaction. Proceedings of the 2022 Conference of the North American Chapter of the Association for Computational Linguistics: Human Language Technologies (NAACL), 3715–3734. https://doi.org/10.18653/v1/2022.naacl-main.272 Shinn, N., Cassano, F., Berman, E., et al. (2023). Reflexion: Language agents with verbal reinforcement learning. Advances in Neural Information Processing Systems (NeurIPS), 36. https://doi.org/10.48550/arXiv.2303.11366 Smith, L., Tanabe, L. K., et al. (2008). Overview of BioCreative II gene mention recognition. Genome Biology, 9(Suppl 2), S2. https://doi.org/10.1186/gb-2008-9-s2-s2 Summers, R. J., Teachey, D. T., & Hunger, S. P. (2025). How I treat ETP-ALL in children. Blood, 145(1), 43–52. https://doi.org/10.1182/blood.2023023155 Thirunavukarasu, A. J., Ting, D. S. J., Elangovan, K., et al. (2023). Large language models in medicine. Nature Medicine, 29(8), 1930–1940. https://doi.org/10.1038/s41591-023-02448-8 Touvron, H., Lavril, T., Izacard, G., et al. (2023). LLaMA: Open and efficient foundation language models. arXiv preprint arXiv:2302.13971. https://doi.org/10.48550/arXiv.2302.13971 Vaswani, A., Shazeer, N., Parmar, N., et al. (2017). Attention is all you need. Advances in Neural Information Processing Systems (NeurIPS), 30. https://doi.org/10.48550/arXiv.1706.03762 Yao, S., Zhao, J., Yu, D., et al. (2023). ReAct: Synergizing reasoning and acting in language models. International Conference on Learning Representations (ICLR). https://openreview.net/forum?id=WE_vluYUL-X | - |
| dc.identifier.uri | http://tdr.lib.ntu.edu.tw/jspui/handle/123456789/103764 | - |
| dc.description.abstract | 急性淋巴性白血病（Acute Lymphoblastic Leukemia, ALL）具有高度分子異質性，不同亞型的預後與治療策略存在根本性差異——精準的亞型診斷直接決定是否需要加入靶向藥物，誤判可能使患者錯失關鍵治療或承受不必要的毒性。然而，ALL 的診斷特徵廣泛分散於大規模生醫文獻中，且歷史命名異質性嚴重阻礙自動化知識整合。現有的基因資訊擷取技術仍高度仰賴傳統文獻檢索與人工整理，缺乏具擴展性與自動語意理解能力的工具；而直接依賴大型語言模型（LLM）參數化記憶的生成方式，則在臨床場景中存在不可接受的幻覺風險。
針對上述挑戰，本研究設計一套多代理檢索增強生成（Multi-Agent Retrieval-Augmented Generation, Multi-Agent RAG）系統，以「文本可溯源擷取」作為核心設計約束，要求每項擷取結果均須追溯至原始文獻的明確文字證據。系統整合倒數排名融合混合檢索與交叉編碼器重排序，實現雙階段精煉檢索；多代理人協作框架包含文件評判器、生成代理、幻覺偵測器與深度推理代理，並搭配依亞型診斷邏輯差異設計的動態提示詞路由。 實驗結果顯示，Agentic RAG 策略的整體命中率由無檢索增強基準線的 17.84% 大幅提升至 95.69%。相較於前沿商用大型語言模型 Claude Opus 4.8 與 GPT-5.5 之純參數知識（No-RAG）——兩者雖具高召回率（78.8%、73.6%），卻因過度生成而精確率偏低（29.9%、27.4%）——本系統在文本可溯源擷取的約束下兼具高召回與高精確（F1 91.4%），顯示其效能源自可溯源設計而非過度生成；切塊消融實驗揭示切塊粒度對擷取效能呈倒 U 型影響，此模式可能同時反映證據粒度、檢索排序與上下文利用率的綜合效應。 本研究的主要貢獻在於提出一套面向 ALL 多亞型診斷特徵擷取、以文本可溯源性為核心約束的 Multi-Agent RAG 框架，為生醫知識的自動化擷取提供一套可溯源且可透過語料庫更新持續擴充的系統性方法。 | zh_TW |
| dc.description.abstract | Acute Lymphoblastic Leukemia (ALL) exhibits high molecular heterogeneity, with fundamentally different prognoses and treatment strategies across subtypes — accurate subtype diagnosis directly determines whether targeted therapies are indicated, and misclassification may deprive patients of critical treatment or expose them to unnecessary toxicity. Yet diagnostic features are dispersed across large-scale biomedical literature and further complicated by substantial historical naming heterogeneity. Existing geneinformation extraction techniques remain heavily reliant on traditional literature search and manual curation, lacking scalable tools with automatic semantic understanding; meanwhile, generation that relies solely on the parametric memory of large language models (LLMs) poses unacceptable hallucination risks in clinical contexts.
To address these challenges, this study presents a Multi-Agent Retrieval-Augmented Generation (Multi-Agent RAG) system constrained by evidence-grounded extraction as a core design principle, requiring every extracted result to be traceable to explicit textual evidence in source documents. The system integrates Reciprocal Rank Fusion (RRF) hybrid retrieval with cross-encoder reranking for two-stage refinement, and employs a multi-agent framework comprising a Document Grader, Generator, Hallucination Detector, and Deep Reasoner, complemented by dynamic prompt routing tailored to differences in diagnostic logic across subtype groups. Experimental results show that the Agentic RAG strategy improves overall hit rate from a no-RAG LLM baseline of 17.84% to 95.69%. Compared with the parametric knowledge (No-RAG) of frontier commercial LLMs Claude Opus 4.8 and GPT-5.5 — which attain high recall (78.8% and 73.6%) but low precision (29.9% and 27.4%) due to over-generation — the proposed system achieves both high recall and high precision (F1 = 91.4%) under its evidence-grounded design, indicating that its performance stems from traceable retrieval rather than overgeneration; chunking ablation experiments reveal an inverted-U pattern in extraction performance across chunk sizes, likely reflecting joint effects of evidence granularity, retrieval ranking, and context utilization. The key contribution of this study is a Multi-Agent RAG framework for extracting diagnostic features across multiple ALL subtypes under an evidence-grounded constraint. This framework provides a traceable and systematically extensible approach to automated biomedical knowledge extraction through continued corpus updates. | en |
| dc.description.provenance | Submitted by admin ntu (admin@lib.ntu.edu.tw) on 2026-08-19T16:32:41Z No. of bitstreams: 0 | en |
| dc.description.provenance | Made available in DSpace on 2026-08-19T16:32:41Z (GMT). No. of bitstreams: 0 | en |
| dc.description.tableofcontents | 論文口試委員審定書 i
誌謝 ii 摘要 iii ABSTRACT iv 目次 vi 圖次 x 表次 xi 1 前言 1 1.1 研究背景與動機 1 1.1.1 急性淋巴性白血病的診斷複雜性 1 1.1.2 多組學診斷特徵的文獻分散問題 2 1.1.3 人工整理的瓶頸與臨床需求 2 1.2 研究問題與目標 3 1.2.1 自動化擷取的核心挑戰 3 1.2.2 醫療幻覺（Hallucination）的臨床風險 4 1.2.3 本研究目標 5 1.3 論文架構說明 5 2 文獻探討 7 2.1 急性淋巴性白血病的分子診斷現況 7 2.1.1 分子亞型分類的演進 7 2.1.2 術語命名的歷史異質性 8 2.1.3 診斷知識的文獻分散性 9 2.2 生醫文本探勘技術的發展 9 2.2.1 命名實體識別與關係抽取 9 2.2.2 預訓練語言模型的突破 10 2.3 大型語言模型於生醫領域的應用與挑戰 10 2.3.1 生成式大型語言模型的崛起 10 2.3.2 幻覺問題與臨床應用的根本矛盾 10 2.3.3 上下文長度與注意力衰退 11 2.4 檢索增強生成技術 11 2.4.1 RAG 的基本架構與核心優勢 11 2.4.2 稠密向量檢索 12 2.4.3 稀疏關鍵字檢索與混合檢索 12 2.4.4 延遲交互檢索（Late Interaction） 13 2.4.5 交叉編碼器重排序與領域特化 13 2.5 Agentic RAG 與多代理人框架 14 2.5.1 從靜態 RAG 到代理人驅動的動態推理 14 2.5.2 Self-RAG 與生成階段的自我反思 15 2.5.3 多代理人協作系統 15 2.5.4 動態提示詞路由 15 2.6 現有方法的研究缺口 16 3 材料與方法 18 3.1 研究設計概覽 18 3.2 核心模型與向量資料庫 19 3.2.1 大型語言模型與生成設定 19 3.2.2 嵌入模型與向量資料庫 20 3.3 生醫語料庫建構 21 3.3.1 文獻蒐集策略 21 3.3.2 文獻預處理與品質過濾 22 3.3.3 文本切塊策略與消融設計 22 3.4 檢索策略設計 23 3.4.1 稠密向量檢索 23 3.4.2 稀疏關鍵字檢索（BM25） 24 3.4.3 倒數排名融合混合檢索（RRF） 25 3.4.4 MedCPT 交叉編碼器重排序 25 3.4.5 延遲交互檢索之對照組設計 26 3.5 多代理人協作框架 26 3.5.1 文件評判器 27 3.5.2 生成代理 28 3.5.3 幻覺偵測器 28 3.5.4 深度推理代理 29 3.6 動態提示詞路由 29 3.6.1 動態路由的設計動機 29 3.6.2 T-ALL 路由設計 30 3.6.3 B-ALL 基因融合型路由設計 30 3.6.4 B-ALL 轉錄組相似型路由設計 31 3.6.5 B-ALL 染色體異常型路由設計 31 3.6.6 自適應 top-N 截斷的設計依據 31 3.7 哨兵標記（Sentinel Token）輸出解析 32 3.8 評估框架設計 33 3.8.1 字串完全比對的侷限性 33 3.8.2 同義詞白名單比對 33 3.8.3 染色體數值比對規則 33 3.8.4 免疫表型 marker 比對 34 3.9 評估基準建構與資料來源 34 3.9.1 建構流程與可溯源性 34 3.9.2 評估目標涵蓋範圍 35 3.9.3 結構性不可達目標的識別與排除 35 3.10 實驗設計 36 3.10.1 消融實驗變因 36 3.10.2 蒙地卡羅重複輪次與統計顯著性 37 3.10.3 評估指標定義 38 3.10.4 硬體與執行環境 39 4 結果與討論 40 4.1 檢索策略整體效能比較 40 4.2 嵌入模型選擇驗證 45 4.3 文本切塊大小消融實驗 49 4.4 亞型層級深度分析 51 4.4.1 T-ALL 亞型命中率分析 51 4.4.2 B-ALL 各亞型群命中率差異 54 4.4.3 PAM 分類亞型的結構性不可達 56 4.5 實驗發現 59 4.5.1 迷失在中間現象 59 4.5.2 哨兵標記對輸出格式穩定性的效益 61 4.5.3 提示詞例子污染效應 62 4.5.4 品質閘門的過度校準悖論 63 4.5.5 命名歧義與較低命中率的關聯現象 64 4.5.6 未達完全命中特徵的相關現象分析 64 4.6 商用大型語言模型對照與多回答效度分析 65 5 結論 69 5.1 研究成果總結 69 5.2 主要貢獻 70 5.2.1 學術貢獻 70 5.2.2 工程貢獻 71 5.3 研究限制 71 5.4 未來研究方向 73 參考文獻 75 A 完整評估目標與各策略命中率 80 B 動態提示詞路由設計 84 B.1 T-ALL 路由 84 B.2 B-ALL 基因融合型路由 84 B.3 B-ALL 轉錄組相似型路由 85 B.4 B-ALL 染色體異常型路由 85 C 評估比對的同義詞白名單 86 | - |
| dc.language.iso | zh_TW | - |
| dc.subject | 急性淋巴性白血病 | - |
| dc.subject | 亞型診斷特徵擷取 | - |
| dc.subject | 檢索增強生成 | - |
| dc.subject | 多代理系統 | - |
| dc.subject | 文本可溯源擷取 | - |
| dc.subject | Acute Lymphoblastic Leukemia | - |
| dc.subject | Subtype Diagnostic Features Extraction | - |
| dc.subject | Retrieval-Augmented Generation | - |
| dc.subject | Multi-Agent System | - |
| dc.subject | Evidence-Grounded Extraction | - |
| dc.title | 應用多代理檢索增強生成自動擷取急性淋巴性白血病亞型診斷特徵 | zh_TW |
| dc.title | Automated Extraction of Acute Lymphoblastic Leukemia Subtype Diagnostic Features Using Multi-Agent Retrieval-Augmented Generation | en |
| dc.type | Thesis | - |
| dc.date.schoolyear | 114-2 | - |
| dc.description.degree | 碩士 | - |
| dc.contributor.oralexamcommittee | 謝秉翰;劉希哲 | zh_TW |
| dc.contributor.oralexamcommittee | Ping-Han Hsieh;Hsi-Che Liu | en |
| dc.subject.keyword | 急性淋巴性白血病; 亞型診斷特徵擷取; 檢索增強生成; 多代理系統; 文本可溯源擷取 | zh_TW |
| dc.subject.keyword | Acute Lymphoblastic Leukemia; Subtype Diagnostic Features Extraction; Retrieval-Augmented Generation; Multi-Agent System; Evidence-Grounded Extraction | en |
| dc.relation.page | 88 | - |
| dc.identifier.doi | 10.6342/NTU202603221 | - |
| dc.rights.note | 同意授權(全球公開) | - |
| dc.date.accepted | 2026-08-13 | - |
| dc.contributor.author-college | 生物資源暨農學院 | - |
| dc.contributor.author-dept | 生物機電工程學系 | - |
| dc.date.embargo-lift | 2026-08-20 | - |
| 顯示於系所單位： | 生物機電工程學系 | |
文件中的檔案：
| 檔案 | 大小 | 格式 | |
|---|---|---|---|
| ntu-114-2.pdf | 2.18 MB | Adobe PDF | 檢視/開啟 |
系統中的文件，除了特別指名其著作權條款之外，均受到著作權保護，並且保留所有的權利。