Cover Image

MOSAIC: Composable Safety Alignment with Modular Control Tokens

Fuente: arXiv

Saved in:

Bibliographic Details
Main Authors:	Peng, Jingyu, Chen, Hongyu, Dong, Jiancheng, Wang, Maolin, Li, Wenxi, Li, Yuchen, Zhang, Kai, Zhao, Xiangyu
Format:	Preprint
Published:	2026
Subjects:	Artificial Intelligence
Online Access:	Acceder al recurso
Tags:	Add Tag No Tags, Be the first to tag this record!

Similar Items

Logic Jailbreak: Efficiently Unlocking LLM Safety Restrictions Through Formal Logical Expression
by: Peng, Jingyu, et al.
Published: (2025)

AdaSwitch: Balancing Exploration and Guidance in Knowledge Distillation via Adaptive Switching
by: Peng, Jingyu, et al.
Published: (2025)

MetaLoRA: Tensor-Enhanced Adaptive Low-Rank Fine-tuning
by: Wang, Maolin, et al.
Published: (2025)

MOSAIC: Modular Orchestration for Structured Agentic Intelligence and Composition
by: Bao, Yifan, et al.
Published: (2026)

Stepwise Reasoning Error Disruption Attack of LLMs
by: Peng, Jingyu, et al.
Published: (2024)

Controllable Safety Alignment: Inference-Time Adaptation to Diverse Safety Requirements
by: Zhang, Jingyu, et al.
Published: (2024)

TMCIR: Token Merge Benefits Composed Image Retrieval
by: Wang, Chaoyang, et al.
Published: (2025)

LSSF: Safety Alignment for Large Language Models through Low-Rank Safety Subspace Fusion
by: Zhou, Guanghao, et al.
Published: (2026)

Safety Alignment Should Be Made More Than Just a Few Tokens Deep
by: Qi, Xiangyu, et al.
Published: (2024)

Empowering Denoising Sequential Recommendation with Large Language Model Embeddings
by: Wu, Tongzhou, et al.
Published: (2025)

Palette: A Modular, Controllable, and Efficient Framework for On-demand Authorized Safety Alignment Relaxation in LLMs
by: Tan, Qitao, et al.
Published: (2026)

NavComposer: Composing Language Instructions for Navigation Trajectories through Action-Scene-Object Modularization
by: He, Zongtao, et al.
Published: (2025)

Automatic Synthetic Data and Fine-grained Adaptive Feature Alignment for Composed Person Retrieval
by: Liu, Delong, et al.
Published: (2023)

Data Efficient Adaptation in Large Language Models via Continuous Low-Rank Fine-Tuning
by: Han, Xiao, et al.
Published: (2025)

SafeWorld: Geo-Diverse Safety Alignment
by: Yin, Da, et al.
Published: (2024)

Speculative Prefill: Turbocharging TTFT with Lightweight and Training-Free Token Importance Estimation
by: Liu, Jingyu, et al.
Published: (2025)

Dual Relation Alignment for Composed Image Retrieval
by: Jiang, Xintong, et al.
Published: (2023)

Token Alignment via Character Matching for Subword Completion
by: Athiwaratkun, Ben, et al.
Published: (2024)

Not All Preferences Are Created Equal: Stability-Aware and Gradient-Efficient Alignment for Reasoning Models
by: Wu, Hui, et al.
Published: (2026)

Large Multimodal Model Compression via Efficient Pruning and Distillation at AntGroup
by: Wang, Maolin, et al.
Published: (2023)

SciSafeEval: A Comprehensive Benchmark for Safety Alignment of Large Language Models in Scientific Tasks
by: Li, Tianhao, et al.
Published: (2024)

Reimagining Safety Alignment with An Image
by: Xia, Yifan, et al.
Published: (2025)

Self-supervised Attribute-aware Dynamic Preference Ranking Alignment
by: Yang, Hongyu, et al.
Published: (2025)

Safety Alignment for Vision Language Models
by: Liu, Zhendong, et al.
Published: (2024)

What Matters For Safety Alignment?
by: Li, Xing, et al.
Published: (2026)

STAR-S: Improving Safety Alignment through Self-Taught Reasoning on Safety Rules
by: Wu, Di, et al.
Published: (2026)

On-Policy Self-Evolution via Failure Trajectories for Agentic Safety Alignment
by: Yin, Bo, et al.
Published: (2026)

Behavior-Equivalent Token: Single-Token Replacement for Long Prompts in LLMs
by: Dong, Jiancheng, et al.
Published: (2025)

More is Less: The Pitfalls of Multi-Model Synthetic Preference Data in DPO Safety Alignment
by: Wang, Yifan, et al.
Published: (2025)

VisuoAlign: Safety Alignment of LVLMs with Multimodal Tree Search
by: Li, MingSheng, et al.
Published: (2025)

Cat-DPO: Category-Adaptive Safety Alignment
by: Yang, Tiankai, et al.
Published: (2026)

Improving Safety Alignment via Balanced Direct Preference Optimization
by: Zhao, Shiji, et al.
Published: (2026)

2D Gaussian Splatting with Semantic Alignment for Image Inpainting
by: Li, Hongyu, et al.
Published: (2025)

Tensorized Hypergraph Neural Networks
by: Wang, Maolin, et al.
Published: (2023)

MOSAIC: Masked Objective with Selective Adaptation for In-domain Contrastive Learning
by: Pavlova, Vera, et al.
Published: (2025)

Encyclo-K: Evaluating LLMs with Dynamically Composed Knowledge Statements
by: Liang, Yiming, et al.
Published: (2025)

Meta-Soft: Leveraging Composable Meta-Tokens for Context-Preserving KV Cache Compression
by: Luo, Wei, et al.
Published: (2026)

Towards Linguistically-informed Representations for English as a Second or Foreign Language: Review, Construction and Application
by: Li, Wenxi, et al.
Published: (2026)

HPS: Hard Preference Sampling for Human Preference Alignment
by: Zou, Xiandong, et al.
Published: (2025)

Disentangling Intent from Role: Adversarial Self-Play for Persona-Invariant Safety Alignment
by: Li, Jiajia, et al.
Published: (2026)