---
title: "RLShield — paper by Tong et al. (2026) — summary and ideas"
description: "Jailbreak detectors for LLMs usually rely on fixed decision thresholds or fixed internal components. 45 references."
authors: ["Zhao Tong", "Pengfei Yang", "Yimeng Gu", "Haichao Shi", "Qiang Liu", "Xingcheng Xu", "Shu Wu", "Xiao-Yu Zhang"]
year: 2026
venue: "Findings of the Association for Computational Linguistics: ACL 2026"
source: "https://aclanthology.org/2026.findings-acl.1182/"
doi: "10.18653/v1/2026.findings-acl.1182"
licence: "CC BY 4.0 (ACL Anthology)"
licence_url: "https://creativecommons.org/licenses/by/4.0/"
summary_licence: "Elamur Cosmic License"
canonical: "https://research.elamur.ai/free/en/papers/tong-2026-rlshield-dynamic-jailbreak-detection-llms/"
---
[Elamur Research](https://md.research.elamur.ai/free/en/) · [this page in HTML](https://research.elamur.ai/free/en/papers/tong-2026-rlshield-dynamic-jailbreak-detection-llms/)

[home](https://md.research.elamur.ai/free/en/) / [papers](https://md.research.elamur.ai/free/en/papers/) / [2026](https://md.research.elamur.ai/free/en/papers/2026/page/41/) / RLShield: Dynamic Jailbreak Detection for LLMs via Reinforced Adaptive Learning

# RLShield: Dynamic Jailbreak Detection for LLMs via Reinforced Adaptive Learning

**Authors:** Zhao Tong, Pengfei Yang, Yimeng Gu, Haichao Shi, [Qiang Liu](https://md.research.elamur.ai/free/en/authors/qiang-liu/), Xingcheng Xu, [Shu Wu](https://md.research.elamur.ai/free/en/authors/shu-wu/) and Xiao-Yu Zhang · **Year:** 2026 · **Venue:** [Findings of the Association for Computational Linguistics: ACL 2026](https://md.research.elamur.ai/free/en/venues/findings/) ([Findings of the Association for Computational Linguistics 2026](https://md.research.elamur.ai/free/en/venues/findings/2026/)) · **Publisher:** [Association for Computational Linguistics](https://md.research.elamur.ai/free/en/publishers/association-for-computational-linguistics/) · **Licence:** [CC BY 4.0 (ACL Anthology)](https://creativecommons.org/licenses/by/4.0/)

**Links:** [Publisher's page](https://aclanthology.org/2026.findings-acl.1182/) · [PDF](https://aclanthology.org/2026.findings-acl.1182.pdf) · [DOI 10.18653/v1/2026.findings-acl.1182](https://doi.org/10.18653/v1/2026.findings-acl.1182) · [OpenAlex](https://openalex.org/W7166879563)

**Citations:** 0 citations (OpenAlex, 2026-10-14)

## Summary
Jailbreak detectors for LLMs usually rely on fixed decision thresholds or fixed internal components. This makes them brittle when attack styles change or the underlying model is updated. RLShield treats threshold choice as a continuous control problem and learns a separate cut-off for each prompt. The framework has three parts:
- **Reference selection:** for each input, it retrieves similar safe and unsafe reference prompts from a knowledge base and diversifies them through LLM rewriting.
- **Parameter localization:** it compares cross-layer representations to find the layers and parameters most relevant to safety, so only a small subset of the model is analysed.
- **Adaptive threshold:** a Soft Actor-Critic (SAC) agent sets the threshold against which the harmfulness score is compared.

Tested on ToxicChat, XSTest and WildGuardTest with several LLaMA and Qwen backbones, RLShield beats strong baselines such as GradSafe in both AUPRC and F1 and runs considerably faster. This matters because it allows detection that adapts to new attacks without manually re-tuning thresholds whenever the backbone changes.

## Key points
- RLShield improves F1 by up to 7.3% over state-of-the-art baselines and averages roughly 3× faster inference across backbones.
- With LLaMA3.1-8B-Instruct, it uses about 84.9% fewer selected parameters and about 66.6% less computation time than GradSafe.
- Retrieval uses cosine similarity to pick top-k candidates, then randomly samples two safe and two unsafe references to avoid near-duplicates. An LLM rewriting step keeps each reference's safety label while varying its style.
- In ablations (LLaMA3.1-8B-Instruct), removing reference selection, parameter localization or the RL threshold each lowers performance. Four reward hyperparameters (α, β, r_max, r_pen) matter most on ToxicChat.
- Stated limitations:
  - The knowledge base can become outdated as new attacks appear.
  - Evaluation covers text-only prompts.
  - Only LLaMA- and Qwen-series backbones were tested.

*Our summary of the paper, in our own words — not its text. The work belongs to its authors; read it at the links above.*

## Institutions

- [Harbin Institute of Technology](https://md.research.elamur.ai/free/en/universities/harbin-institute-of-technology/) — University, Harbin, China · 158 works · 28,841 citations
- [Chinese Academy of Sciences](https://md.research.elamur.ai/free/en/government/chinese-academy-of-sciences/) — Government body, Beijing, China · 194 works · 23,906 citations
- [Shanghai Artificial Intelligence Laboratory](https://md.research.elamur.ai/free/en/labs/shanghai-artificial-intelligence-laboratory/) — Research institute, Shanghai, China · 117 works · 11,281 citations
- [Institute of Automation](https://md.research.elamur.ai/free/en/labs/institute-of-automation/) — Research institute, Beijing, China · 70 works · 14,251 citations
- [University of Chinese Academy of Sciences](https://md.research.elamur.ai/free/en/universities/university-of-chinese-academy-of-sciences/) — University, Beijing, China · 147 works · 8,480 citations
- [Queen Mary University of London](https://md.research.elamur.ai/free/en/universities/queen-mary-university-of-london/) — University, London, United Kingdom · 54 works · 3,542 citations
- [Institute of Information Engineering](https://md.research.elamur.ai/free/en/labs/institute-of-information-engineering/) — Research institute, Beijing, China · 30 works · 1,411 citations

📚 Shelves (4): [🖥️ Computer Science](https://md.research.elamur.ai/free/en/shelves/computer-science/) · 16,667 works · [🤖 Artificial Intelligence](https://md.research.elamur.ai/free/en/shelves/computer-science/artificial-intelligence/) · 16,194 works · [🎭 Arts and Humanities](https://md.research.elamur.ai/free/en/shelves/arts-and-humanities/) · 16,039 works · [🗣️ Language and Linguistics](https://md.research.elamur.ai/free/en/shelves/arts-and-humanities/language-and-linguistics/) · 15,891 works

🧭 Topics (1): [🛡️ Bias, fairness & safety](https://md.research.elamur.ai/free/en/topics/bias-fairness-safety/) · 517 works

## Cite

Cite as: Zhao Tong, Pengfei Yang, Yimeng Gu, Haichao Shi, Qiang Liu, Xingcheng Xu, Shu Wu and Xiao-Yu Zhang. 2026. RLShield: Dynamic Jailbreak Detection for LLMs via Reinforced Adaptive Learning. Findings of the Association for Computational Linguistics: ACL 2026.

The paper itself, not this page. BibTeX from the library's record of the work.

```bibtex
@inproceedings{tong2026rlshield,
  title = {{RLShield: Dynamic Jailbreak Detection for LLMs via Reinforced Adaptive Learning}},
  author = {Zhao Tong, Pengfei Yang, Yimeng Gu, Haichao Shi, Qiang Liu, Xingcheng Xu, Shu Wu and Xiao-Yu Zhang},
  booktitle = {Findings of the Association for Computational Linguistics: ACL 2026},
  year = {2026},
  doi = {10.18653/v1/2026.findings-acl.1182},
  url = {https://doi.org/10.18653/v1/2026.findings-acl.1182}
}
```

- [← previous River-LLM: Large Language Model Seamless Exit Based on KV Share](https://md.research.elamur.ai/free/en/papers/shen-2026-river-llm-large-language-model/)
- [next → RoadMapper: A Multi-Agent System for Roadmap Generation of Solving Complex Research Problems](https://md.research.elamur.ai/free/en/papers/liu-2026-roadmapper-multi-agent-system-roadmap/)
- [↑ back to papers of 2026](https://md.research.elamur.ai/free/en/papers/2026/page/41/)

## References

45 references, 11 of them with a card of ours. ↗ leads to the work, «find» searches for one we have no id for yet.

1. Elisa Bassignana, Valerio Basile, Viviana Patti, and 1 others. 2018. [Hurtlex: A multilingual lexicon of words to hurt](https://md.research.elamur.ai/free/en/papers/bassignana-2018-hurtlex-multilingual-lexicon-words-hurt/). In CEUR Workshop proceedings. CEUR-WS.
2. Guorui Chen, Yifan Xia, Xiaojun Jia, Zhijiang Li, Philip Torr, and Jindong Gu. 2025. Llm jailbreak detection for (almost) free! In Findings of the Association for Computational Linguistics: EMNLP 2025, page 5777–5807. Association for Computational Linguistics. [find](https://scholar.google.com/scholar?q=Guorui+Chen%2C+Yifan+Xia%2C+Xiaojun+Jia%2C+Zhijiang+Li%2C+Philip+Torr%2C+and+Jindong+Gu%2E+2025%2E+Llm+jailbreak+detection+for+%28almost%29+free%21+In+Findings+of+the+Association+for+Computational+Linguistics%3A+EMNLP+2025)
3. Jianlv Chen, Shitao Xiao, Peitian Zhang, Kun Luo, Defu Lian, and Zheng Liu. 2024. Bge m3-embedding: Multi-lingual, multi-functionality, multi-granularity text embeddings through self-knowledge distillation. ArXiv preprint. [find](https://scholar.google.com/scholar?q=Jianlv+Chen%2C+Shitao+Xiao%2C+Peitian+Zhang%2C+Kun+Luo%2C+Defu+Lian%2C+and+Zheng+Liu%2E+2024%2E+Bge+m3%2Dembedding%3A+Multi%2Dlingual%2C+multi%2Dfunctionality%2C+multi%2Dgranularity+text+embeddings+through+self%2Dknowledge+distill)
4. Jessica Maria Echterhoff, Fartash Faghri, Raviteja Vemulapalli, Ting-Yao Hu, Chun-Liang Li, Oncel Tuzel, and Hadi Pouransari. 2024. Muscle: A model update strategy for compatible llm evolution. In Findings of the Association for Computational Linguistics: EMNLP 2024. [find](https://scholar.google.com/scholar?q=Jessica+Maria+Echterhoff%2C+Fartash+Faghri%2C+Raviteja+Vemulapalli%2C+Ting%2DYao+Hu%2C+Chun%2DLiang+Li%2C+Oncel+Tuzel%2C+and+Hadi+Pouransari%2E+2024%2E+Muscle%3A+A+model+update+strategy+for+compatible+llm+evolution%2E+In+Fin)
5. Yanlin Feng, Simone Papicchio, and Sajjadur Rahman. 2025. Cypherbench: Towards precise retrieval over full-scale modern knowledge graphs in the llm era. In Proceedings of the 63rd Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers). [find](https://scholar.google.com/scholar?q=Yanlin+Feng%2C+Simone+Papicchio%2C+and+Sajjadur+Rahman%2E+2025%2E+Cypherbench%3A+Towards+precise+retrieval+over+full%2Dscale+modern+knowledge+graphs+in+the+llm+era%2E+In+Proceedings+of+the+63rd+Annual+Meeting+of+th)
6. Aaron Grattafiori, Abhimanyu Dubey, Abhinav Jauhri, Abhinav Pandey, Abhishek Kadian, Ahmad Al-Dahle, Aiesha Letman, Akhil Mathur, Alan Schelten, Alex Vaughan, and 1 others. 2024. [The llama 3 herd of models](https://md.research.elamur.ai/free/en/papers/grattafiori-2024-llama-3-herd-models/). arXiv preprint arXiv:2407.21783.
7. Peijian Gu, Quan Wang, and Zhendong Mao. 2025. Improve safety training of large language models with safety-critical singular vectors localization. In Proceedings of the 63rd Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers). [find](https://scholar.google.com/scholar?q=Peijian+Gu%2C+Quan+Wang%2C+and+Zhendong+Mao%2E+2025%2E+Improve+safety+training+of+large+language+models+with+safety%2Dcritical+singular+vectors+localization%2E+In+Proceedings+of+the+63rd+Annual+Meeting+of+the+Ass)
8. Tuomas Haarnoja, Aurick Zhou, Pieter Abbeel, and Sergey Levine. 2018. Soft actor-critic: Off-policy maximum entropy deep reinforcement learning with a stochastic actor. In Proceedings of the 35th International Conference on Machine Learning, ICML 2018, Stockholmsmässan, Stockholm, Sweden, July 10-15… [find](https://scholar.google.com/scholar?q=Tuomas+Haarnoja%2C+Aurick+Zhou%2C+Pieter+Abbeel%2C+and+Sergey+Levine%2E+2018%2E+Soft+actor%2Dcritic%3A+Off%2Dpolicy+maximum+entropy+deep+reinforcement+learning+with+a+stochastic+actor%2E+In+Proceedings+of+the+35th+Inte)
9. Seungju Han, Kavel Rao, Allyson Ettinger, Liwei Jiang, Bill Yuchen Lin, Nathan Lambert, Yejin Choi, and Nouha Dziri. 2024. [Wildguard: Open one-stop moderation tools for safety risks, jailbreaks, and refusals of llms](https://md.research.elamur.ai/free/en/papers/han-2024-wildguard-open-one-stop-moderation/). In Advances in Neural Information Processing Systems 38: Annual Conference on Neura…
10. Xiaomeng Hu, Pin-Yu Chen, and Tsung-Yi Ho. 2024. Gradient cuff: Detecting jailbreak attacks on large language models by exploring refusal loss landscapes. In Advances in Neural Information Processing Systems 38: Annual Conference on Neural Information Processing Systems 2024, NeurIPS 2024, Vancouver… [find](https://scholar.google.com/scholar?q=Xiaomeng+Hu%2C+Pin%2DYu+Chen%2C+and+Tsung%2DYi+Ho%2E+2024%2E+Gradient+cuff%3A+Detecting+jailbreak+attacks+on+large+language+models+by+exploring+refusal+loss+landscapes%2E+In+Advances+in+Neural+Information+Processing+)
11. Tiansheng Huang, Sihao Hu, Fatih Ilhan, Selim Furkan Tekin, and Ling Liu. 2024. Harmful fine-tuning attacks and defenses for [large language models: A survey](https://md.research.elamur.ai/free/en/papers/minaee-2024-large-language-models-survey/). ArXiv preprint.
12. Neel Jain, Avi Schwarzschild, Yuxin Wen, Gowthami Somepalli, John Kirchenbauer, Ping-yeh Chiang, Micah Goldblum, Aniruddha Saha, Jonas Geiping, and Tom Goldstein. 2023. [Baseline defenses for adversarial attacks against aligned language models](https://md.research.elamur.ai/free/en/papers/jain-2023-baseline-defenses-adversarial-attacks-against/). ArXiv preprint.
13. Alyssa Lees, Vinh Q Tran, Yi Tay, Jeffrey Sorensen, Jai Gupta, Donald Metzler, and Lucy Vasserman. 2022. A new generation of perspective api: Efficient multilingual character-level transformers. In Proceedings of the 28th ACM SIGKDD conference on knowledge discovery and data mining, pages 3197–3207. [find](https://scholar.google.com/scholar?q=Alyssa+Lees%2C+Vinh+Q+Tran%2C+Yi+Tay%2C+Jeffrey+Sorensen%2C+Jai+Gupta%2C+Donald+Metzler%2C+and+Lucy+Vasserman%2E+2022%2E+A+new+generation+of+perspective+api%3A+Efficient+multilingual+character%2Dlevel+transformers%2E+In+Pr)
14. Shen Li, Liuyi Yao, Lan Zhang, and Yaliang Li. 2025. Safety layers in aligned large language models: The key to LLM security. In The Thirteenth International Conference on Learning Representations. [find](https://scholar.google.com/scholar?q=Shen+Li%2C+Liuyi+Yao%2C+Lan+Zhang%2C+and+Yaliang+Li%2E+2025%2E+Safety+layers+in+aligned+large+language+models%3A+The+key+to+LLM+security%2E+In+The+Thirteenth+International+Conference+on+Learning+Representations%2E)
15. Zi Lin, Zihan Wang, Yongqi Tong, Yangkun Wang, Yuxin Guo, Yujia Wang, and Jingbo Shang. 2023. [ToxicChat: Unveiling hidden challenges of toxicity detection in real-world user-AI conversation](https://md.research.elamur.ai/free/en/papers/lin-2023-toxicchat-unveiling-hidden-challenges-toxicity/). In Findings of the Association for Computational Linguistics: EMNLP 2023, Singapore.
16. Michael L Littman. 2015. Reinforcement learning improves behaviour from evaluative feedback. Nature, (7553). [find](https://scholar.google.com/scholar?q=Michael+L+Littman%2E+2015%2E+Reinforcement+learning+improves+behaviour+from+evaluative+feedback%2E+Nature%2C+%287553%29%2E)
17. Yi Liu, Junzhe Yu, Huijia Sun, Ling Shi, Gelei Deng, Yuqi Chen, and Yang Liu. 2024. Efficient detection of toxic prompts in large language models. In Proceedings of the 39th IEEE/ACM International Conference on Automated Software Engineering. [find](https://scholar.google.com/scholar?q=Yi+Liu%2C+Junzhe+Yu%2C+Huijia+Sun%2C+Ling+Shi%2C+Gelei+Deng%2C+Yuqi+Chen%2C+and+Yang+Liu%2E+2024%2E+Efficient+detection+of+toxic+prompts+in+large+language+models%2E+In+Proceedings+of+the+39th+IEEE%2FACM+International+Con)
18. Edward Loper and Steven Bird. 2002. NLTK: The natural language toolkit. In Proceedings of the ACL-02 Workshop on Effective Tools and Methodologies for Teaching Natural Language Processing and Computational Linguistics, Philadelphia, Pennsylvania, USA. [find](https://scholar.google.com/scholar?q=Edward+Loper+and+Steven+Bird%2E+2002%2E+NLTK%3A+The+natural+language+toolkit%2E+In+Proceedings+of+the+ACL%2D02+Workshop+on+Effective+Tools+and+Methodologies+for+Teaching+Natural+Language+Processing+and+Computat)
19. Todor Markov, Chong Zhang, Sandhini Agarwal, Florentine Eloundou Nekoul, Theodore Lee, Steven Adler, Angela Jiang, and Lilian Weng. 2023. [A holistic approach to undesired content detection in the real world](https://md.research.elamur.ai/free/en/papers/markov-2022-holistic-approach-undesired-content-detection/). In Proceedings of the AAAI conference on artificial intelligence, volume 37, pages 15009–150…
20. Sarah Masud, Sahajpreet Singh, Viktor Hangya, Alexander Fraser, and Tanmoy Chakraborty. 2024. Hate personified: Investigating the role of llms in content moderation. ArXiv preprint. [find](https://scholar.google.com/scholar?q=Sarah+Masud%2C+Sahajpreet+Singh%2C+Viktor+Hangya%2C+Alexander+Fraser%2C+and+Tanmoy+Chakraborty%2E+2024%2E+Hate+personified%3A+Investigating+the+role+of+llms+in+content+moderation%2E+ArXiv+preprint%2E)
21. Meta. 2024. meta-llama/llama-3.2-3b-instruct. Hugging Face model card. [find](https://scholar.google.com/scholar?q=Meta%2E+2024%2E+meta%2Dllama%2Fllama%2D3%2E2%2D3b%2Dinstruct%2E+Hugging+Face+model+card%2E)
22. Meta. 2025. meta-llama/llama-guard-4-12b. Hugging Face model card. [find](https://scholar.google.com/scholar?q=Meta%2E+2025%2E+meta%2Dllama%2Fllama%2Dguard%2D4%2D12b%2E+Hugging+Face+model+card%2E)
23. Microsoft. 2024. Azure ai content safety. [https://azure.microsoft.com/en-us/](https://azure.microsoft.com/en-us/) products/ai-services/ai-content-safety. [↗](https://azure.microsoft.com/en-us/)
24. OpenAI. 2025. Gpt-5 system card. [https://cdn.openai.com/gpt-5-system-card.pdf](https://cdn.openai.com/gpt-5-system-card.pdf). [↗](https://cdn.openai.com/gpt-5-system-card.pdf)
25. Sindhu Padakandla, Sadbhavana Babar, Manohar Kaul, and 1 others. 2025. Safequant: Llm safety analysis via quantized gradient inspection. In Proceedings of the 2025 Conference of the Nations of the Americas Chapter of the Association for Computational Linguistics: Human Language Technologies (Volume… [find](https://scholar.google.com/scholar?q=Sindhu+Padakandla%2C+Sadbhavana+Babar%2C+Manohar+Kaul%2C+and+1+others%2E+2025%2E+Safequant%3A+Llm+safety+analysis+via+quantized+gradient+inspection%2E+In+Proceedings+of+the+2025+Conference+of+the+Nations+of+the+Ame)
26. Yikang Pan, Liangming Pan, Wenhu Chen, Preslav Nakov, Min-Yen Kan, and William Wang. 2023. [On the risk of misinformation pollution with large language models](https://md.research.elamur.ai/free/en/papers/pan-2023-risk-misinformation-pollution-large-language/). In Findings of the Association for Computational Linguistics: EMNLP 2023, Singapore.
27. Yu Peng, Zewen Long, Fangming Dong, Congyi Li, Shu Wu, and Kai Chen. 2024. Playing language game with llms leads to jailbreaking. arXiv preprint arXiv:2411.12762. — [arXiv:2411.12762](https://arxiv.org/abs/2411.12762) [find](https://scholar.google.com/scholar?q=Yu+Peng%2C+Zewen+Long%2C+Fangming+Dong%2C+Congyi+Li%2C+Shu+Wu%2C+and+Kai+Chen%2E+2024%2E+Playing+language+game+with+llms+leads+to+jailbreaking%2E+arXiv+preprint+arXiv%3A2411%2E12762%2E+%E2%80%94+%5BarXiv%3A2411%2E12762%5D%28https%3A%2F%2Farxiv%2Eor)
28. Paul Röttger, Hannah Kirk, Bertie Vidgen, Giuseppe Attanasio, Federico Bianchi, and Dirk Hovy. 2024. [XSTest: A test suite for identifying exaggerated safety behaviours in large language models](https://md.research.elamur.ai/free/en/papers/rottger-2023-xstest-test-suite-identifying-exaggerated/). In Proceedings of the 2024 Conference of the North American Chapter of the Association for Computational L…
29. Zhao Tong, Chunlin Gong, Yimeng Gu, Haichao Shi, Qiang Liu, Shu Wu, and Xiao-Yu Zhang. 2025a. Group-adaptive adversarial learning for robust fake news detection against malicious comments. arXiv preprint arXiv:2510.09712. — [arXiv:2510.09712](https://arxiv.org/abs/2510.09712) [find](https://scholar.google.com/scholar?q=Zhao+Tong%2C+Chunlin+Gong%2C+Yimeng+Gu%2C+Haichao+Shi%2C+Qiang+Liu%2C+Shu+Wu%2C+and+Xiao%2DYu+Zhang%2E+2025a%2E+Group%2Dadaptive+adversarial+learning+for+robust+fake+news+detection+against+malicious+comments%2E+arXiv+prepr)
30. Zhao Tong, Yimeng Gu, Huidong Liu, Qiang Liu, Shu Wu, Haichao Shi, and Xiao-Yu Zhang. 2025b. Generate first, then sample: Enhancing fake news detection with llm-augmented reinforced sampling. In Proceedings of the 63rd Annual Meeting of the Association for Computational Linguistics (Volume 1: Long P… [find](https://scholar.google.com/scholar?q=Zhao+Tong%2C+Yimeng+Gu%2C+Huidong+Liu%2C+Qiang+Liu%2C+Shu+Wu%2C+Haichao+Shi%2C+and+Xiao%2DYu+Zhang%2E+2025b%2E+Generate+first%2C+then+sample%3A+Enhancing+fake+news+detection+with+llm%2Daugmented+reinforced+sampling%2E+In+Proce)
31. Zhao Tong, Yimeng Gu, Haichao Shi, Qiang Liu, Shu Wu, and Xiao-Yu Zhang. 2025c. R 2 fnd: Reinforced rationale learning for fake news detection with llms. In 2025 International Joint Conference on Neural Networks (IJCNN), pages 1–8. IEEE. [find](https://scholar.google.com/scholar?q=Zhao+Tong%2C+Yimeng+Gu%2C+Haichao+Shi%2C+Qiang+Liu%2C+Shu+Wu%2C+and+Xiao%2DYu+Zhang%2E+2025c%2E+R+2+fnd%3A+Reinforced+rationale+learning+for+fake+news+detection+with+llms%2E+In+2025+International+Joint+Conference+on+Neur)
32. Zhao Tong, Xiang Yuan, Qiang Liu, Libin Han, Haichao Shi, Shu Wu, and Xiao-Yu Zhang. 2025d. Multimodal test-time adaptation for fake news detection. In International Conference on Intelligent Computing, pages 122–133. Springer. [find](https://scholar.google.com/scholar?q=Zhao+Tong%2C+Xiang+Yuan%2C+Qiang+Liu%2C+Libin+Han%2C+Haichao+Shi%2C+Shu+Wu%2C+and+Xiao%2DYu+Zhang%2E+2025d%2E+Multimodal+test%2Dtime+adaptation+for+fake+news+detection%2E+In+International+Conference+on+Intelligent+Computin)
33. Hugo Touvron, Louis Martin, Kevin Stone, Peter Albert, Amjad Almahairi, Yasmine Babaei, Nikolay Bashlykov, Soumya Batra, Prajjwal Bhargava, Shruti Bhosale, and 1 others. 2023. [Llama 2: Open foundation and fine-tuned chat models](https://md.research.elamur.ai/free/en/papers/touvron-2023-llama-2-open-foundation-fine/). ArXiv preprint.
34. Shangqing Tu, Zhuoran Pan, Wenxuan Wang, Zhexin Zhang, Yuliang Sun, Jifan Yu, Hongning Wang, Lei Hou, and Juanzi Li. 2025. Knowledge-to-jailbreak: Investigating knowledge-driven jailbreaking attacks for large language models. In Proceedings of the 31st ACM SIGKDD Conference on Knowledge Discovery an… [find](https://scholar.google.com/scholar?q=Shangqing+Tu%2C+Zhuoran+Pan%2C+Wenxuan+Wang%2C+Zhexin+Zhang%2C+Yuliang+Sun%2C+Jifan+Yu%2C+Hongning+Wang%2C+Lei+Hou%2C+and+Juanzi+Li%2E+2025%2E+Knowledge%2Dto%2Djailbreak%3A+Investigating+knowledge%2Ddriven+jailbreaking+attacks+f)
35. Neural Information Processing Systems 38: Annual Conference on Neural Information Processing Systems 2024, NeurIPS 2024, Vancouver, BC, Canada, December 10 - 15, 2024. [find](https://scholar.google.com/scholar?q=Neural+Information+Processing+Systems+38%3A+Annual+Conference+on+Neural+Information+Processing+Systems+2024%2C+NeurIPS+2024%2C+Vancouver%2C+BC%2C+Canada%2C+December+10+%2D+15%2C+2024%2E)
36. Yueqi Xie, Minghong Fang, Renjie Pi, and Neil Gong. 2024. Gradsafe: Detecting jailbreak prompts for llms via safety-critical gradient analysis. ArXiv preprint. [find](https://scholar.google.com/scholar?q=Yueqi+Xie%2C+Minghong+Fang%2C+Renjie+Pi%2C+and+Neil+Gong%2E+2024%2E+Gradsafe%3A+Detecting+jailbreak+prompts+for+llms+via+safety%2Dcritical+gradient+analysis%2E+ArXiv+preprint%2E)
37. Zitao Xuan, Xiaofeng Mao, Da Chen, Xin Zhang, Yuhan Dong, and Jun Zhou. 2025. Shieldhead: Decoding-time safeguard for large language models. In Findings of the Association for Computational Linguistics: ACL 2025. [find](https://scholar.google.com/scholar?q=Zitao+Xuan%2C+Xiaofeng+Mao%2C+Da+Chen%2C+Xin+Zhang%2C+Yuhan+Dong%2C+and+Jun+Zhou%2E+2025%2E+Shieldhead%3A+Decoding%2Dtime+safeguard+for+large+language+models%2E+In+Findings+of+the+Association+for+Computational+Linguistic)
38. Haoming Yang, Ke Ma, Xiaojun Jia, Yingfei Sun, Qianqian Xu, and Qingming Huang. 2025. Cannot see the forest for the trees: Invoking heuristics and biases to elicit irrational choices of llms. ArXiv preprint. [find](https://scholar.google.com/scholar?q=Haoming+Yang%2C+Ke+Ma%2C+Xiaojun+Jia%2C+Yingfei+Sun%2C+Qianqian+Xu%2C+and+Qingming+Huang%2E+2025%2E+Cannot+see+the+forest+for+the+trees%3A+Invoking+heuristics+and+biases+to+elicit+irrational+choices+of+llms%2E+ArXiv+pr)
39. Kai Ye, Hongyi Zhou, Jin Zhu, Francesco Quinzan, and Chengchun Shi. 2025. Robust reinforcement learning from human feedback for large language models fine-tuning. ArXiv preprint. [find](https://scholar.google.com/scholar?q=Kai+Ye%2C+Hongyi+Zhou%2C+Jin+Zhu%2C+Francesco+Quinzan%2C+and+Chengchun+Shi%2E+2025%2E+Robust+reinforcement+learning+from+human+feedback+for+large+language+models+fine%2Dtuning%2E+ArXiv+preprint%2E)
40. Jinghao Zhang, Yuting Liu, Qiang Liu, Shu Wu, Guibing Guo, and Liang Wang. 2024. Stealthy attack on large language model based recommendation. In Proceedings of the 62nd Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers), pages 5839–5857. [find](https://scholar.google.com/scholar?q=Jinghao+Zhang%2C+Yuting+Liu%2C+Qiang+Liu%2C+Shu+Wu%2C+Guibing+Guo%2C+and+Liang+Wang%2E+2024%2E+Stealthy+attack+on+large+language+model+based+recommendation%2E+In+Proceedings+of+the+62nd+Annual+Meeting+of+the+Associat)
41. Guanghao Zhou, Panjia Qiu, Cen Chen, Hongyu Li, Jason Chu, Xin Zhang, and Jun Zhou. 2025. Lssf: Safety alignment for large language models through low-rank safety subspace fusion. In Proceedings of the 63rd Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers). [find](https://scholar.google.com/scholar?q=Guanghao+Zhou%2C+Panjia+Qiu%2C+Cen+Chen%2C+Hongyu+Li%2C+Jason+Chu%2C+Xin+Zhang%2C+and+Jun+Zhou%2E+2025%2E+Lssf%3A+Safety+alignment+for+large+language+models+through+low%2Drank+safety+subspace+fusion%2E+In+Proceedings+of+th)
42. Jiongxiao Wang, Jiazhao Li, Yiquan Li, Xiangyu Qi, Junjie Hu, Sharon Li, Patrick McDaniel, Muhao Chen, Bo Li, and Chaowei Xiao. 2024a. Backdooralign: Mitigating fine-tuning based jailbreak attack with backdoor enhanced safety alignment. In Advances in Neural Information Processing Systems 38: Annual… [find](https://scholar.google.com/scholar?q=Jiongxiao+Wang%2C+Jiazhao+Li%2C+Yiquan+Li%2C+Xiangyu+Qi%2C+Junjie+Hu%2C+Sharon+Li%2C+Patrick+McDaniel%2C+Muhao+Chen%2C+Bo+Li%2C+and+Chaowei+Xiao%2E+2024a%2E+Backdooralign%3A+Mitigating+fine%2Dtuning+based+jailbreak+attack+with)
43. Shijie Wang, Wenqi Fan, Yue Feng, Lin Shanru, Xinyu Ma, Shuaiqiang Wang, and Dawei Yin. 2025. [Knowledge graph retrieval-augmented generation for llm-based recommendation](https://md.research.elamur.ai/free/en/papers/wang-2025-knowledge-graph-retrieval-augmented-generation/). In Proceedings of the 63rd Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers).
44. Yihan Wang, Zhouxing Shi, Andrew Bai, and Cho-Jui Hsieh. 2024b. Defending llms against jailbreaking attacks via backtranslation. ArXiv preprint. [find](https://scholar.google.com/scholar?q=Yihan+Wang%2C+Zhouxing+Shi%2C+Andrew+Bai%2C+and+Cho%2DJui+Hsieh%2E+2024b%2E+Defending+llms+against+jailbreaking+attacks+via+backtranslation%2E+ArXiv+preprint%2E)
45. Shirley Wu, Shiyu Zhao, Michihiro Yasunaga, Kexin Huang, Kaidi Cao, Qian Huang, Vassilis N. Ioannidis, Karthik Subbian, James Y. Zou, and Jure Leskovec. 2024. Stark: Benchmarking LLM retrieval on textual and relational knowledge bases. In Advances in [find](https://scholar.google.com/scholar?q=Shirley+Wu%2C+Shiyu+Zhao%2C+Michihiro+Yasunaga%2C+Kexin+Huang%2C+Kaidi+Cao%2C+Qian+Huang%2C+Vassilis+N%2E+Ioannidis%2C+Karthik+Subbian%2C+James+Y%2E+Zou%2C+and+Jure+Leskovec%2E+2024%2E+Stark%3A+Benchmarking+LLM+retrieval+on+text)

*Updated 1 November 2026*