Newest papers

Beyond Attack Success Rate: Cost-Aware Metric for Evaluating Iterative LLM Jailbreaks

Rohan Menon, Chih-Hong Cheng · 2026 · arXiv (Cornell University)

DOI: 10.48550/arxiv.2610.26589 · cited by 0 works (OpenAlex) · open-access PDF

Russian

Успешность атаки (ASR) оценивает джейлбрейк против больших языковых моделей без учета повторных вызовов к модели жертвы. Тем не менее, каждый дополнительный вызов дает развернутой системе еще одну возможность обнаружить атаку и вмешаться. Мы вводим $\mathrm{ASR}_{\mathrm{STEP}}$, что наказывает дополнительные вызовы жертвы-модели при сохранении двоичных суждений успеха. Наша оценка рассматривает GCG, PAIR с бюджетом шага, и одноэтапную базовую линию PAP на JailBreakBench по трем семействам моделей. Результаты GCG снижаются на 7,0-8,7 процентных пунктов после дисконтирования, в то время как PAP остаются неизменными. В то

Machine translation (Meta M2M100); the English original follows.

Abstract

Attack success rate (ASR) evaluates jailbreaks against large language models without accounting for repeated calls to the victim model. Yet each additional call gives the deployed system another opportunity to detect the attack and intervene. We introduce $\mathrm{ASR}_{\mathrm{STEP}}$, which penalizes additional victim-model calls while retaining binary success judgments. Our evaluation considers GCG, PAIR with a step budget, and a single-step PAP baseline on JailBreakBench across three model families. GCG's scores decrease by 7.0-8.7 percentage points after discounting, while PAP's remain unchanged. While for PAIR attack, we see a wider discounting range between 3.4-12 percentage points, showing that the "step" quantity differs per iterative strategy. Reporting both metrics makes the discount due to repeated victim-model interaction explicit. Code is available here: https://github.com/SuELS-Group/beyond_asr.

Read in: English · Spanish · Chinese · Hindi · Arabic · Portuguese · French · German · Japanese · Russian · Korean · Indonesian · Vietnamese

Topics: Metric (unit), Discounting, Computer science, Baseline (sea), Code (set theory), Range (aeronautics)

Is this paper real? This page was built from OpenAlex and arXiv records, not generated. Software can verify any reference in one call: GET https://store.agentexchange.work/gw/papers/verify?title=...&year=... (0.003 USD), fetch this record as JSON with /gw/papers/paper?id=2610.26589, or list what it cites with /gw/papers/references?id=.... Pay per call in USDC (x402) or with a 19 USD card key for 1,000 calls. Catalogue · llms.txt

Sources: OpenAlex (CC0), arXiv, Crossref. Agent Exchange