{"status":"ok","message-type":"work","message-version":"1.0.0","message":{"indexed":{"date-parts":[[2026,7,9]],"date-time":"2026-07-09T15:41:18Z","timestamp":1783611678104,"version":"3.55.0"},"reference-count":32,"publisher":"Institute of Electrical and Electronics Engineers (IEEE)","license":[{"start":{"date-parts":[[2025,1,1]],"date-time":"2025-01-01T00:00:00Z","timestamp":1735689600000},"content-version":"vor","delay-in-days":0,"URL":"https:\/\/creativecommons.org\/licenses\/by\/4.0\/legalcode"}],"content-domain":{"domain":[],"crossmark-restriction":false},"short-container-title":["IEEE Access"],"published-print":{"date-parts":[[2025]]},"DOI":"10.1109\/access.2025.3583759","type":"journal-article","created":{"date-parts":[[2025,6,27]],"date-time":"2025-06-27T17:46:40Z","timestamp":1751046400000},"page":"113194-113207","source":"Crossref","is-referenced-by-count":6,"title":["Meticulous Thought Defender: Fine-Grained Chain-of-Thought (CoT) for Detecting Prompt Injection Attacks of Large Language Models"],"prefix":"10.1109","volume":"13","author":[{"ORCID":"https:\/\/orcid.org\/0009-0002-7013-4541","authenticated-orcid":false,"given":"Lijuan","family":"Shi","sequence":"first","affiliation":[{"name":"China Telecom Research Institute, Beijing, China"}],"role":[{"vocabulary":"crossref","role":"author"}]},{"ORCID":"https:\/\/orcid.org\/0009-0000-4071-7880","authenticated-orcid":false,"given":"Yajing","family":"Kang","sequence":"additional","affiliation":[{"name":"China Telecom Research Institute, Beijing, China"}],"role":[{"vocabulary":"crossref","role":"author"}]},{"given":"Jie","family":"Hu","sequence":"additional","affiliation":[{"name":"China Telecom Research Institute, Beijing, China"}],"role":[{"vocabulary":"crossref","role":"author"}]},{"given":"Xinchi","family":"Li","sequence":"additional","affiliation":[{"name":"China Telecom Research Institute, Beijing, China"}],"role":[{"vocabulary":"crossref","role":"author"}]},{"given":"Mingchuan","family":"Yang","sequence":"additional","affiliation":[{"name":"China Telecom Research Institute, Beijing, China"}],"role":[{"vocabulary":"crossref","role":"author"}]}],"member":"263","reference":[{"key":"ref1","doi-asserted-by":"publisher","DOI":"10.1109\/ICHMS59971.2024.10555871"},{"key":"ref2","article-title":"Prompt injection attack against LLM-integrated applications","author":"Liu","year":"2023","journal-title":"arXiv:2306.05499"},{"key":"ref3","article-title":"Baseline defenses for adversarial attacks against aligned language models","author":"Jain","year":"2023","journal-title":"arXiv:2309.00614"},{"key":"ref4","first-page":"24824","article-title":"Chain-of-thought prompting elicits reasoning in large language models","volume-title":"Proc. 36th Int. Conf. Neural Inf. Process. Syst.","author":"Lee"},{"key":"ref5","volume-title":"GPT-4 is OpenAI\u2019s Most Advanced System, Producing Safer and More Useful Responses","year":"2023"},{"key":"ref6","volume-title":"The Official Repo of Qwen Chat & Pretrained Large Language Model Proposed by Alibaba Cloud","year":"2023"},{"key":"ref7","volume-title":"Qwen2.5: A Party of Foundation Models","year":"2024"},{"key":"ref8","doi-asserted-by":"publisher","DOI":"10.48550\/ARXIV.1706.03762"},{"key":"ref9","article-title":"Token-level adversarial prompt detection based on perplexity measures and contextual information","author":"Hu","year":"2023","journal-title":"arXiv:2311.11509"},{"key":"ref10","first-page":"257","article-title":"Embedding-based classifiers can detect prompt injection attacks","volume-title":"Proc. Conf. Appl. Mach. Learn. Inf. Secur.","author":"Ayub"},{"key":"ref11","first-page":"48","article-title":"Defending against indirect prompt injection attacks with spotlighting","volume-title":"Proc. Conf. Appl. Mach. Learn. Inf. Secur.","author":"Hines"},{"key":"ref12","first-page":"159","article-title":"Robust safety classifier against jailbreaking attacks: Adversarial prompt shield","volume-title":"Proc. 8th Workshop Online Abuse Harms (WOAH)","author":"Kim"},{"key":"ref13","first-page":"557","article-title":"Evaluating the instruction-following robustness of large language models to prompt injection","volume-title":"Proc. Conf. Empirical Methods Natural Lang. Process.","author":"Li"},{"key":"ref14","article-title":"Adversarial contrastive decoding: Boosting safety alignment of large language models via opposite prompt optimization","author":"Zhao","year":"2024","journal-title":"arXiv:2406.16743"},{"key":"ref15","volume-title":"Incorrect Count of \u2018r\u2019 Characters in the Word \u2018strawberry\u2019","year":"2024"},{"key":"ref16","first-page":"357","article-title":"Decomposed prompting: A modular approach for solving complex tasks","volume-title":"Proc. 11th Int. Conf. Learn. Represent.","author":"Khot"},{"key":"ref17","volume-title":"Learning to Reason With LLMs","year":"2024"},{"key":"ref18","article-title":"DeepSeek-r1: Incentivizing reasoning capability in LLMs via reinforcement learning","author":"DeepSeek-AI","year":"2025","journal-title":"arXiv:2501.12948"},{"key":"ref19","first-page":"261","article-title":"NCL_NLP at SemEval-2024 task 7: CoT-NumHG: A CoT-based SFT training strategy with large language models for number-focused headline generation","volume-title":"Proc. 18th Int. Workshop Semantic Eval. (SemEval)","author":"Zhao"},{"key":"ref20","first-page":"11809","article-title":"Tree of thoughts: Deliberate problem solving with large language models","volume-title":"Proc. 37th Int. Conf. Neural Inf. Process. Syst.","author":"Yao"},{"key":"ref21","first-page":"7101","article-title":"Prompt-based Monte\u2013Carlo tree search for goal-oriented dialogue policy planning","volume-title":"Proc. Conf. Empirical Methods Natural Lang. Process.","author":"Yu"},{"key":"ref22","article-title":"Marco-o1: Towards open reasoning models for open-ended solutions","author":"Zhao","year":"2024","journal-title":"arXiv:2411.14405"},{"key":"ref23","volume-title":"Deepset\/Prompt-Injections","year":"2024"},{"key":"ref24","volume-title":"XTRam1\/Safe-Guard-Prompt-Injection","year":"2024"},{"key":"ref25","doi-asserted-by":"publisher","DOI":"10.1609\/aaai.v38i21.30566"},{"key":"ref26","article-title":"DeBERTa: Decoding-enhanced BERT with disentangled attention","author":"He","year":"2020","journal-title":"arXiv:2006.03654"},{"key":"ref27","article-title":"Improving language understanding by generative pre-training","author":"Radford","year":"2018"},{"key":"ref28","article-title":"Rethinking embedding coupling in pre-trained language models","author":"Won Chung","year":"2020","journal-title":"arXiv:2010.12821"},{"key":"ref29","article-title":"BGE M3-embedding: Multi-lingual, multi-functionality, multi-granularity text embeddings through self-knowledge distillation","author":"Chen","year":"2024","journal-title":"arXiv:2402.03216"},{"key":"ref30","volume-title":"O1-Open\/Openo1-SFT","year":"2024"},{"key":"ref31","doi-asserted-by":"publisher","DOI":"10.18653\/v1\/2024.acl-demos.38"},{"key":"ref32","volume-title":"Deliberative Alignment: Reasoning Enables Safer Language Models","author":"Guan","year":"2024"}],"container-title":["IEEE Access"],"original-title":[],"link":[{"URL":"http:\/\/xplorestaging.ieee.org\/ielx8\/6287639\/10820123\/11053836.pdf?arnumber=11053836","content-type":"unspecified","content-version":"vor","intended-application":"similarity-checking"}],"deposited":{"date-parts":[[2025,7,5]],"date-time":"2025-07-05T04:35:01Z","timestamp":1751690101000},"score":1,"resource":{"primary":{"URL":"https:\/\/ieeexplore.ieee.org\/document\/11053836\/"}},"subtitle":[],"short-title":[],"issued":{"date-parts":[[2025]]},"references-count":32,"URL":"https:\/\/doi.org\/10.1109\/access.2025.3583759","relation":{},"ISSN":["2169-3536"],"issn-type":[{"value":"2169-3536","type":"electronic"}],"subject":[],"published":{"date-parts":[[2025]]}}}