Score: 0

TeleLoRA: Teleporting Model-Specific Alignment Across LLMs

Published: March 26, 2025 | arXiv ID: 2503.20228v1

By: Xiao Lin , Manoj Acharya , Anirban Roy and more

Potential Business Impact:

Cleans harmful secrets from AI without retraining.

Business Areas:
Natural Language Processing Artificial Intelligence, Data and Analytics, Software

Mitigating Trojans in Large Language Models (LLMs) is one of many tasks where alignment data is LLM specific, as different LLMs have different Trojan triggers and trigger behaviors to be removed. In this paper, we introduce TeleLoRA (Teleporting Low-Rank Adaptation), a novel framework that synergizes model-specific alignment data across multiple LLMs to enable zero-shot Trojan mitigation on unseen LLMs without alignment data. TeleLoRA learns a unified generator of LoRA adapter weights by leveraging local activation information across multiple LLMs. This generator is designed to be permutation symmetric to generalize across models with different architectures and sizes. We optimize the model design for memory efficiency, making it feasible to learn with large-scale LLMs with minimal computational resources. Experiments on LLM Trojan mitigation benchmarks demonstrate that TeleLoRA effectively reduces attack success rates while preserving the benign performance of the models.

Page Count
16 pages

Category
Computer Science:
Machine Learning (CS)