Paper Library

Measuring What Matters: A Framework for Evaluating Safety Risks in Real-World LLM Applications

Jia Yi Goh, Shaun Khoo, Nyx Iskandar, Gabriel Chua, Leanne Tan, Jessica Foo

2025-07-13

safety

2507.09820v1

SEALGuard: Safeguarding the Multilingual Conversations in Southeast Asian Languages for LLM Software Systems

Wenliang Shan, Michael Fu, Rui Yang, Chakkrit Tantithamthavorn

2025-07-11

2507.08898v3

SEALGuard: Safeguarding the Multilingual Conversations in Southeast Asian Languages for LLM Software Systems

Wenliang Shan, Michael Fu, Rui Yang, Chakkrit Tantithamthavorn

2025-07-11

2507.08898v2

SEALGuard: Safeguarding the Multilingual Conversations in Southeast Asian Languages for LLM Software Systems

Wenliang Shan, Michael Fu, Rui Yang, Chakkrit, Tantithamthavorn

2025-07-11

2507.08898v1

A Dynamic Stackelberg Game Framework for Agentic AI Defense Against LLM Jailbreaking

Zhengye Han, Quanyan Zhu

2025-07-10

red teaming safety

2507.08207v1

GuardVal: Dynamic Large Language Model Jailbreak Evaluation for Comprehensive Safety Testing

Peiyan Zhang, Haibo Jin, Liying Kang, Haohan Wang

2025-07-10

red teaming

2507.07735v1

An attention-aware GNN-based input defender against multi-turn jailbreak on LLMs

Zixuan Huang, Kecheng Huang, Lihao Yin, Bowei He, Huiling Zhen, Mingxuan Yuan, Zili Shao

2025-07-09

red teaming

2507.07146v1

On the Robustness of Verbal Confidence of LLMs in Adversarial Attacks

Stephen Obadinma, Xiaodan Zhu

2025-07-09

red teaming

2507.06489v1

Foundation Model Self-Play: Open-Ended Strategy Innovation via Foundation Models

Aaron Dharna, Cong Lu, Jeff Clune

2025-07-09

red teaming

2507.06466v1

The bitter lesson of misuse detection

Hadrien Mariaccia, Charbel-Raphaël Segerie, Diego Dorn

2025-07-08

red teaming

2507.06282v1

CAVGAN: Unifying Jailbreak and Defense of LLMs via Generative Adversarial Attacks on their Internal Representations

Xiaohu Li, Yunfeng Ning, Zepeng Bao, Mayi Xu, Jianhao Chen, Tieyun Qian

2025-07-08

red teaming safety

2507.06043v1

TuneShield: Mitigating Toxicity in Conversational AI while Fine-tuning on Untrusted Data

Aravind Cheruvu, Shravya Kanchi, Sifat Muhammad Abdullah, Nicholas Kong, Daphne Yao, Murtuza Jadliwala, Bimal Viswanath

2025-07-08

red teaming

2507.05660v1

Response Attack: Exploiting Contextual Priming to Jailbreak Large Language Models

Ziqi Miao, Lijun Li, Yuan Xiong, Zhenhua Liu, Pengyu Zhu, Jing Shao

2025-07-07

red teaming

2507.05248v1

Trojan Horse Prompting: Jailbreaking Conversational Multimodal Models by Forging Assistant Message

Wei Duan, Li Qian

2025-07-07

red teaming

2507.04673v1

July 14 - July 20, 2025

The Devil behind the mask: An emergent safety vulnerability of Diffusion LLMs

ARMOR: Aligning Secure and Safe Large Language Models via Meticulous Reasoning

July 07 - July 13, 2025

Measuring What Matters: A Framework for Evaluating Safety Risks in Real-World LLM Applications

SEALGuard: Safeguarding the Multilingual Conversations in Southeast Asian Languages for LLM Software Systems

SEALGuard: Safeguarding the Multilingual Conversations in Southeast Asian Languages for LLM Software Systems

SEALGuard: Safeguarding the Multilingual Conversations in Southeast Asian Languages for LLM Software Systems

A Dynamic Stackelberg Game Framework for Agentic AI Defense Against LLM Jailbreaking

GuardVal: Dynamic Large Language Model Jailbreak Evaluation for Comprehensive Safety Testing

An attention-aware GNN-based input defender against multi-turn jailbreak on LLMs

On the Robustness of Verbal Confidence of LLMs in Adversarial Attacks

Foundation Model Self-Play: Open-Ended Strategy Innovation via Foundation Models

The bitter lesson of misuse detection

CAVGAN: Unifying Jailbreak and Defense of LLMs via Generative Adversarial Attacks on their Internal Representations

TuneShield: Mitigating Toxicity in Conversational AI while Fine-tuning on Untrusted Data

Response Attack: Exploiting Contextual Priming to Jailbreak Large Language Models

Trojan Horse Prompting: Jailbreaking Conversational Multimodal Models by Forging Assistant Message

June 30 - July 06, 2025

Tail-aware Adversarial Attacks: A Distributional Approach to Efficient LLM Jailbreaking

Tail-aware Adversarial Attacks: A Distributional Approach to Efficient LLM Jailbreaking

Attention Slipping: A Mechanistic Understanding of Jailbreak Attacks and Defenses in LLMs

Re-Emergent Misalignment: How Narrow Fine-Tuning Erodes Safety Alignment in LLMs

On Jailbreaking Quantized Language Models Through Fault Injection Attacks

On Jailbreaking Quantized Language Models Through Fault Injection Attacks

Adversarial Manipulation of Reasoning Models using Internal Representations

Visual Contextual Attack: Jailbreaking MLLMs with Image-Driven Context Injection