Safe Semantics, Unsafe Interpretations: Tackling Implicit Reasoning Safety in Large Vision-Language Models

Wei Cai, Jian Zhao, Yuchu Jiang, Tianle Zhang, Xuelong Li
8/12/2025
cs.AI

No confident code implementation found for this paper. 3 possible matches can be shown below.

Abstract

Large Vision-Language Models face growing safety challenges with multimodal inputs. This paper introduces the concept of Implicit Reasoning Safety, a vulnerability in LVLMs. Benign combined inputs trigger unsafe LVLM outputs due to flawed or hidden reasoning. To showcase this, we developed Safe Semantics, Unsafe Interpretations, the first dataset for this critical issue. Our demonstrations show that even simple In-Context Learning with SSUI significantly mitigates these implicit multimodal threats, underscoring the urgent need to improve cross-modal implicit reasoning.

View on arXivView PDF

Code Implementations

No confident code match yet

We couldn't find an author-owned or strongly-evidenced community implementation for this paper. 3 weaker matches are hidden by default — verify before relying on them.

No code implementations found yet.

Know of an implementation? Let us know in the comments below!

Similar Papers

When Safe Unimodal Inputs Collide: Optimizing Reasoning Chains for Cross-Modal Safety in Multimodal Large Language Models

85% match2025

MM-PoisonRAG: Disrupting Multimodal RAG with Local and Global Poisoning Attacks

84% match1 repo162025

SaFeR-VLM: Toward Safety-aware Fine-grained Reasoning in Multimodal Models

83% match1 repo102025

SIA: Enhancing Safety via Intent Awareness for Vision-Language Models

83% match1 repo2025

SafeMT: Multi-turn Safety for Multimodal Language Models

82% match2025

Evolving Contextual Safety in Multi-Modal Large Language Models via Inference-Time Self-Reflective Memory

82% match1 repo182026

Cite this paper

@article{cai2025safe,
  title  = {Safe Semantics, Unsafe Interpretations: Tackling Implicit Reasoning Safety in Large Vision-Language Models},
  author = {Wei Cai and Jian Zhao and Yuchu Jiang and Tianle Zhang and Xuelong Li},
  year   = {2025},
  eprint = {2508.08926},
  archivePrefix = {arXiv},
  url    = {http://arxiv.org/abs/2508.08926v1}
}

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/2508.08926.svg)](https://codeofpaper.com/paper/2508.08926)

Discussion