analiza aiqbench
MUSE introduces a run-centric framework that tracks the full trajectory of multimodal LLM interactions, moving beyond binary success metrics to analyze multi-turn attack dynamics and nuanced refusal behaviors. The study reveals that iterative attacks are significantly more effective than direct requests, while modality switching alone does not consistently boost attack success rates. This fine-grained evaluation approach provides essential data for hardening multimodal models against sophisticated, evolving adversarial strategies.
MUSE: A Run-Centric Platform for Multimodal Unified Safety Evaluation of Large Language Models
arXiv:2603.02482v2 Announce Type: replace-cross Abstract: Safety evaluation of multimodal large language models requires tracking not only whether an attack succeeds, but also how the interaction unfolds across turns and input modalities. We present MUSE (Multimodal Unified Safety Evaluation), an open-source, browser-based, run-centric platform for multimodal safety evaluation. MUSE treats each attack run as the persistent unit of execution, inspection, and analysis, preserving its configuration, multi-turn trajectory, delivered modalities and media, target responses, and safety judgments. A five-level response taxonomy further distinguishes full Compliance from Partial Compliance and refusal behavior, yielding hard ASR, soft ASR, and gray-zone width (GZW). Across 11,700 evaluations on six multimodal LLMs, direct text-only requests yield only 3.1% macro hard ASR and 4.4% soft ASR, while iterative attack procedures are substantially more effective. Attack effectiveness also varies substantially with the attacker backbone. In contrast, Inter-Turn Modality Switching (ITMS), evaluated as a controlled delivery-modality probe, does not consistently increase attack success. These results demonstrate the value of run-centric, fine-grained evaluation for characterizing multimodal safety behavior beyond a single binary success metric.
Abstract:Safety evaluation of multimodal large language models requires tracking not only whether an attack succeeds, but also how the interaction unfolds across turns and input modalities. We present MUSE (Multimodal Unified Safety Evaluation), an open-source, browser-based, run-centric platform for multimodal safety evaluation. MUSE treats each attack run as the persistent unit of execution, inspection, and analysis, preserving its configuration, multi-turn trajectory, delivered modalities and media, target responses, and safety judgments. A five-level response taxonomy further distinguishes full Compliance from Partial Compliance and refusal behavior, yielding hard ASR, soft ASR, and gray-zone width (GZW). Across 11,700 evaluations on six multimodal LLMs, direct text-only requests yield only 3.1% macro hard ASR and 4.4% soft ASR, while iterative attack procedures are substantially more effective. Attack effectiveness also varies substantially with the attacker backbone. In contrast, Inter-Turn Modality Switching (ITMS), evaluated as a controlled delivery-modality probe, does not consistently increase attack success. These results demonstrate the value of run-centric, fine-grained evaluation for characterizing multimodal safety behavior beyond a single binary success metric.
Machine Learning (cs.LG); Computation and Language (cs.CL); Computer Vision and Pattern Recognition (cs.CV); Sound (cs.SD); Audio and Speech Processing (eess.AS)
arXiv:2603.02482 [cs.LG]
(or arXiv:2603.02482v2 [cs.LG] for this version)
Text integral sub licența deschisă a publicației · 196 cuvinte
Subiecte conexe
[AINews] not much happened today
Latent Space · 10 sept. 2026
[AINews] Reality Checks on AI News (Yegge shuts down Gas Town, Databricks’ +60% Astra cost)
Latent Space · 17 sept. 2026
NVIDIA GTC Taipei at COMPUTEX: Live Updates on What’s Next in AI
NVIDIA · 21 mai 2026
[AINews] Claude Fable/Mythos 5.1: new SOTA model, 75% cache price cut but 70% more output tokens
Latent Space · 2 sept. 2026