MUSE: A Run-Centric Platform for Multimodal Unified Safety Evaluation of Large Language Models

TL;DR

MUSE is an open-source platform for comprehensive safety evaluation of large multimodal language models, enabling systematic testing across text, audio, image, and video inputs with novel attack strategies and metrics.

Contribution

The paper introduces MUSE, a run-centric, multimodal safety evaluation platform with new attack algorithms, dual-metric framework, and modality switching techniques for cross-modal safety assessment.

Findings

Multi-turn strategies achieve up to 100% attack success rate.

Inter-turn modality switching accelerates attack convergence.

Modality effects vary across model families, not universally.

Abstract

Safety evaluation and red-teaming of large language models remain predominantly text-centric, and existing frameworks lack the infrastructure to systematically test whether alignment generalizes to audio, image, and video inputs. We present MUSE (Multimodal Unified Safety Evaluation), an open-source, run-centric platform that integrates automatic cross-modal payload generation, three multi-turn attack algorithms (Crescendo, PAIR, Violent Durian), provider-agnostic model routing, and an LLM judge with a five-level safety taxonomy into a single browser-based system. A dual-metric framework distinguishes hard Attack Success Rate (Compliance only) from soft ASR (including Partial Compliance), capturing partial information leakage that binary metrics miss. To probe whether alignment generalizes across modality boundaries, we introduce Inter-Turn Modality Switching (ITMS), which augments multi-turn attacks with per-turn modality rotation. Experiments across six multimodal LLMs from four providers show that multi-turn strategies can achieve up to 90-100% ASR against models with near-perfect single-turn refusal. ITMS does not uniformly raise final ASR on already-saturated baselines, but accelerates convergence by destabilizing early-turn defenses, and ablation reveals that the direction of modality effects is model-family-specific rather than universal, underscoring the need for provider-aware cross-modal safety testing.