The Side Effects of Being Smart: Safety Risks in MLLMs' Multi-Image Reasoning

Renmiao Chen, Yida Lu, Shiyao Cui, Xuan Ouyang, Victor Shea-Jay Huang +4 more
1/20/2026
cs.CVcs.CL

Official code: thu-coai/MIR-SafetyBench (4 stars, MIT, last commit 2026-01-22).

Abstract

As Multimodal Large Language Models (MLLMs) acquire stronger reasoning capabilities to handle complex, multi-image instructions, this advancement may pose new safety risks. We study this problem by introducing MIR-SafetyBench, the first benchmark focused on multi-image reasoning safety, which consists of 2,676 instances across a taxonomy of 9 multi-image relations. Our extensive evaluations on 19 MLLMs reveal a troubling trend: models with more advanced multi-image reasoning can be more vulnerable on MIR-SafetyBench. Beyond attack success rates, we find that many responses labeled as safe are superficial, often driven by misunderstanding or evasive, non-committal replies. We further observe that unsafe generations exhibit lower attention entropy than safe ones on average. This internal signature suggests a possible risk that models may over-focus on task solving while neglecting safety constraints. Our code and data are available at https://github.com/thu-coai/MIR-SafetyBench.

View on arXivView PDF

Code Implementations(1)

thu-coai/MIR-SafetyBenchOfficial100%
MIT

MIR-SafetyBench: Evaluating Multi-image Reasoning Safety of Multimodal Large Language Models

41PythonJan 19, 20269 months agoMIT

Similar Papers

SafeMLRM: Demystifying Safety in Multi-modal Large Reasoning Models

88% match2025

Can't See the Forest for the Trees: Benchmarking Multimodal Safety Awareness for Multimodal LLMs

88% match2025

When Helpers Become Hazards: A Benchmark for Analyzing Multimodal LLM-Powered Safety in Daily Life

88% match1 repo22026

Rethinking Bottlenecks in Safety Fine-Tuning of Vision Language Models

88% match1 repo142025

Multimodal Situational Safety

87% match2024

Are Smarter LLMs Safer? Exploring Safety-Reasoning Trade-offs in Prompting and Fine-Tuning

87% match2025

Cite this paper

@article{chen2026side,
  title  = {The Side Effects of Being Smart: Safety Risks in MLLMs' Multi-Image Reasoning},
  author = {Renmiao Chen and Yida Lu and Shiyao Cui and Xuan Ouyang and Victor Shea-Jay Huang and Shumin Zhang and Chengwei Pan and Han Qiu and Minlie Huang},
  year   = {2026},
  eprint = {2601.14127},
  archivePrefix = {arXiv},
  url    = {http://arxiv.org/abs/2601.14127v1}
}

Add a badge to your README

Shows whether official code is linked for this paper.

Code of Paper badge
[![Code of Paper](https://codeofpaper.com/badge/2601.14127.svg)](https://codeofpaper.com/paper/2601.14127)

Discussion