# Multimodal Safety Evaluation in Generative Agent Social Simulations

arXiv: https://arxiv.org/abs/2510.07709 · HTML: https://mlsanity.com/paper/2510.07709/

## Key facts

- Cited by 6 papers in the map (data through 14 Oct 2026).
- Placed in Agent threats, guardrails and governance, one of 16 directions in LLM agents and tool use.
- Builds on 19 of its 33 references that are in the map.
- First posted on arXiv 9 Oct 2025.

## Placed in

- LLM agents and tool use / [Agent threats, guardrails and governance](https://mlsanity.com/agents/agent-threats-guardrails-and-governance/)

## Builds on (most cited first, 19 of 19)

- [Visual Instruction Tuning](https://mlsanity.com/paper/2304.08485/) (2023)
- [Flamingo: a Visual Language Model for Few-Shot Learning](https://mlsanity.com/paper/2204.14198/) (2022)
- [Generative Agents: Interactive Simulacra of Human Behavior](https://mlsanity.com/paper/2304.03442/) (2023)
- [AutoGen: Enabling Next-Gen LLM Applications via Multi-Agent Conversation](https://mlsanity.com/paper/2308.08155/) (2023)
- [CAMEL: Communicative Agents for "Mind" Exploration of Large Language Model Society](https://mlsanity.com/paper/2303.17760/) (2023)
- [AgentVerse: Facilitating Multi-Agent Collaboration and Exploring Emergent Behaviors](https://mlsanity.com/paper/2308.10848/) (2023)
- [FigStep: Jailbreaking Large Vision-Language Models via Typographic Visual Prompts](https://mlsanity.com/paper/2311.05608/) (2023)
- [Visual Adversarial Examples Jailbreak Aligned Large Language Models](https://mlsanity.com/paper/2306.13213/) (2023)
- [MM-SafetyBench: A Benchmark for Safety Evaluation of Multimodal Large Language Models](https://mlsanity.com/paper/2311.17600/) (2023)
- [Jailbreak in pieces: Compositional Adversarial Attacks on Multi-Modal Language Models](https://mlsanity.com/paper/2307.14539/) (2023)
- [OpenAgents: An Open Platform for Language Agents in the Wild](https://mlsanity.com/paper/2310.10634/) (2023)
- [OASIS: Open Agent Social Interaction Simulations with One Million Agents](https://mlsanity.com/paper/2411.11581/) (2024)
- [OWL: Optimized Workforce Learning for General Multi-Agent Assistance in Real-World Task Automation](https://mlsanity.com/paper/2505.23885/) (2025)
- [Evil Geniuses: Delving into the Safety of LLM-based Agents](https://mlsanity.com/paper/2311.11855/) (2023)
- [ScreenAgent: A Vision Language Model-driven Computer Control Agent](https://mlsanity.com/paper/2402.07945/) (2024)
- [On the Resilience of LLM-Based Multi-Agent Collaboration with Faulty Agents](https://mlsanity.com/paper/2408.00989/) (2024)
- [MultiTrust: A Comprehensive Benchmark Towards Trustworthy Multimodal Large Language Models](https://mlsanity.com/paper/2406.07057/) (2024)
- [AgentSense: Benchmarking Social Intelligence of Language Agents through Interactive Scenarios](https://mlsanity.com/paper/2410.19346/) (2024)
- [Building Trustworthy Multimodal AI: A Review of Fairness, Transparency, and Ethics in Vision-Language Tasks](https://mlsanity.com/paper/2504.13199/) (2025)

## Built on by (most cited first, 6 of 6)

- [SAVeS: Steering Safety Judgments in Vision-Language Models via Semantic Cues](https://mlsanity.com/paper/2603.19092/) (2026)
- [Systematizing LLM Persona Design: A Four-Quadrant Technical Taxonomy for AI Companion Applications](https://mlsanity.com/paper/2511.02979/) (2025)
- [HBEE: Human Behavioral Entropy Engine -- Pre-Registered Multi-Agent LLM Simulation of Peer-Suspicion-Based Detection Inversion](https://mlsanity.com/paper/2605.07472/) (2026)
- [Agentic Collaborative Cognition for Zero-Shot 3D Understanding](https://mlsanity.com/paper/2606.24649/) (2026)
- [SGA: Plug&Play Geometric Verification for Educational Video Synthesis](https://mlsanity.com/paper/2607.18116/) (2026)
- [APEX-Harm: Evaluating Agent Safety in Shared Professional Workspaces](https://mlsanity.com/paper/2610.13623/) (2026)

## Abstract (arXiv)

Can generative agents be trusted in multimodal environments? Despite advances in large language and vision-language models that enable agents to act autonomously and pursue goals in rich settings, their ability to reason about safety, coherence, and trust across modalities remains limited. We introduce a reproducible simulation framework for evaluating agents along three dimensions: (1) safety improvement over time, including iterative plan revisions in text-visual scenarios; (2) detection of unsafe activities across multiple categories of social situations; and (3) social dynamics, measured as interaction counts and acceptance ratios of social exchanges. Agents are equipped with layered memory, dynamic planning, multimodal perception, and are instrumented with SocialMetrics, a suite of behavioral and structural metrics that quantifies plan revisions, unsafe-to-safe conversions, and information diffusion across networks. Experiments show that while agents can detect direct multimodal contradictions, they often fail to align local revisions with global safety, reaching only a 55 percent success rate in correcting unsafe plans. Across eight simulation runs with three models - Claude, GPT-4o mini, and Qwen-VL - five agents achieved average unsafe-to-safe conversion rates of 75, 55, and 58 percent, respectively. Overall performance ranged from 20 percent in multi-risk scenarios with GPT-4o mini to 98 percent in localized contexts such as fire/heat with Claude. Notably, 45 percent of unsafe actions were accepted when paired with misleading visuals, showing a strong tendency to overtrust images. These findings expose critical limitations in current architectures and provide a reproducible platform for studying multimodal safety, coherence, and social dynamics.