SafeGene: Reusable Adapters for Transferable Safety Alignment
Quick Answer
SafeGene introduces a reusable safety-adapter module for open-weight LLMs, enhancing safety alignment without compromising performance.
Quick Take
It effectively reduces harmful response rates across various model families while maintaining downstream task efficiency, outperforming existing safe adaptation methods in safety-utility trade-offs.
Key Points
- SafeGene decouples safety capability from task-specific updates for better adaptability.
- It utilizes aligned-degraded model discrepancies to create transferable safety vectors.
- Experiments show reduced harmful response rates while maintaining performance across tasks.
- SafeGene outperforms traditional safe adaptation methods in safety-utility trade-offs.
- The approach is applicable across multiple architecture-compatible model families.
Paper Resources
Article Content
From source RSS / original summaryarXiv:2606. 06519v1 Announce Type: new Abstract: Open-weight are increasingly fine-tuned into customized assistants, but downstream fine-tuning can weaken safety alignment and make models more vulnerable to malicious prompts, even when the training data is not intentionally harmful. This creates a recurring safety recovery problem as target models are repeatedly updated with new task data or user interactions.
We propose SafeGene, a reusable safety-adapter module designed for cross-task reuse within each architecture-compatible model family. Rather than treating safety recovery as a model-specific repair step, SafeGene treats safety capability as an independent, reusable adapter representation decoupled from task-specific updates.
This representation is obtained from aligned--degraded model discrepancies, refined into task-transferable safety vectors through data-aware layer selection, and expressed in each downstream task-adapted model via few-shot layer-wise coefficient recalibration.
Experiments across multiple model families, downstream tasks, and safety judges show that SafeGene-enhanced models reduce harmful response rates while maintaining downstream performance, outperforming representative safe adaptation methods in safety--utility trade-off.
Want this in your inbox every morning?
Daily brief at your local 8am — bilingual EN/中文, free.
More from arXiv cs.AI
See more →HOBA: Hierarchical On-Policy Bidding Agents for Adaptive Online Advertising
HOBA (Hierarchical On-policy Bidding Agents) is a novel hierarchical reinforcement learning framework that enhances online advertising bidding systems by improving adaptability and reducing hyperparameter tuning costs. It utilizes a for hyperparameter inference, a SARSA agent for expert model selection, and a dynamic expert pool for bid execution, achieving a +3.6% increase in target cost during large-scale deployment and outperforming state-of-the-art baselines on AuctionNet.