Adversarial Reasoning at Jailbreaking Time
Fuente:
arXiv
Salvato in:
| Autori principali: | , , , , , |
|---|---|
| Natura: | Preprint |
| Pubblicazione: |
2025
|
| Soggetti: | |
| Accesso online: | |
| Tags: |
Aggiungi Tag
Nessun Tag, puoi essere il primo ad aggiungerne!!
|
| _version_ | 1866913910816243712 |
|---|---|
| author | Sabbaghi, Mahdi Kassianik, Paul Pappas, George Singer, Yaron Karbasi, Amin Hassani, Hamed |
| author_facet | Sabbaghi, Mahdi Kassianik, Paul Pappas, George Singer, Yaron Karbasi, Amin Hassani, Hamed |
| contents | As large language models (LLMs) are becoming more capable and widespread, the study of their failure cases is becoming increasingly important. Recent advances in standardizing, measuring, and scaling test-time compute suggest new methodologies for optimizing models to achieve high performance on hard tasks. In this paper, we apply these advances to the task of model jailbreaking: eliciting harmful responses from aligned LLMs. We develop an adversarial reasoning approach to automatic jailbreaking that leverages a loss signal to guide the test-time compute, achieving SOTA attack success rates against many aligned LLMs, even those that aim to trade inference-time compute for adversarial robustness. Our approach introduces a new paradigm in understanding LLM vulnerabilities, laying the foundation for the development of more robust and trustworthy AI systems. |
| format | Preprint |
| id |
arxiv_https___arxiv_org_abs_2502_01633 |
| institution | arXiv |
| publishDate | 2025 |
| record_format | arxiv |
| spellingShingle | Adversarial Reasoning at Jailbreaking Time Sabbaghi, Mahdi Kassianik, Paul Pappas, George Singer, Yaron Karbasi, Amin Hassani, Hamed Machine Learning Artificial Intelligence As large language models (LLMs) are becoming more capable and widespread, the study of their failure cases is becoming increasingly important. Recent advances in standardizing, measuring, and scaling test-time compute suggest new methodologies for optimizing models to achieve high performance on hard tasks. In this paper, we apply these advances to the task of model jailbreaking: eliciting harmful responses from aligned LLMs. We develop an adversarial reasoning approach to automatic jailbreaking that leverages a loss signal to guide the test-time compute, achieving SOTA attack success rates against many aligned LLMs, even those that aim to trade inference-time compute for adversarial robustness. Our approach introduces a new paradigm in understanding LLM vulnerabilities, laying the foundation for the development of more robust and trustworthy AI systems. |
| title | Adversarial Reasoning at Jailbreaking Time |
| topic | Machine Learning Artificial Intelligence |
| url | https://arxiv.org/abs/2502.01633 |