SafeMol: Dual-Modality Safety Alignment for Molecular Multimodal Models
cs.LG
Submitted: 2026-09-27
Updated: 2026-09-27
License: http://creativecommons.org/licenses/by/4.0/
The gist: Molecular multimodal models support diverse understanding and generation tasks but may introduce safety vulnerabilities when handling hazardous molecules.
Terminology
Abstract
Molecular multimodal models support diverse understanding and generation tasks but may introduce safety vulnerabilities when handling hazardous molecules. In this work, We reveal substantial jailbreak vulnerabilities under both text-only and graph-conditioned settings. Our analysis further shows that safety robustness must hold across input modalities while balancing safety, over-refusal, and utility. To address these challenges, we construct SafeMolBench, a molecular multimodal safety-alignment benchmark with 3702 samples covering 618 unique hazardous molecules and safe molecular tasks, organized into hazardous-harmful, hazardous-allowed, and utility-replay subsets to support unified training and evaluation of safety, over-refusal, and utility. Based on SafeMolBench, we propose SafeMol, a parameter-efficient safety alignment framework that jointly optimizes lightweight modules across text-only and graph-conditioned inputs, uses MMD for distribution-level representation alignment to reduce modality-induced discrepancies, and explicitly models molecular hazardousness and harmful operational intent. Experiments on SafeMolBench show that SafeMol reduces attack success by several tens of percentage points while largely maintaining low over-refusal and preserving molecular-task utility.
Sources
- OR-Bench: An Over-Refusal Benchmark for Large Language Models
- SciRisk-Bench: A Risk-Dimension-Aware Benchmark for AI4Science Safety
- The Llama 3 Herd of Models
- LoRA: Low-Rank Adaptation of Large Language Models
- VLM-Guard: Safeguarding Vision-Language Models via Fulfilling Safety Alignment Gap
- A Molecular Multimodal Foundation Model Associating Molecule Graphs with Natural Language
- SMILES-Prompting: A Novel Approach to LLM Jailbreak Attacks in Chemical Synthesis
- Learning Optimal Multimodal Information Bottleneck Representations
- ChemSafetyBench: Benchmarking LLM Safety on Chemistry Domain
- Safety Fine-Tuning at (Almost) No Cost: A Baseline for Vision Large Language Models
Related papers
- Polynomial-Augmented Neural Networks (PANNs) with Weak Orthogonality Constraints for Enhanced Function and PDE Approximation
- AIRL-S: Unifying Reinforcement Learning and Search-Based Test-Time Scaling via Adversarial Inverse Reinforcement Learning
- Transformers as Bayesian In-Context Experimenters: Smoothness-Adaptive Efficient ATE Estimation
- Convergence issues in Relational Concept Analysis based on AOC-posets
- Beliefs Beyond Posteriors: The Predictive Cost of ELBO Binding
- Understanding Diffusion Models via Ratio-Based Function Approximation with SignReLU Networks