paper-with-me

홈 › Papers

Unlearning Trojans in Large Language Models: A Comparison Between Natural Language and Source Code

2024-08-22 · Mahdi Kazemi, Aftab Hussain, Md Rafiqul Islam Rabin, Mohammad Amin Alipour, Sen Lin

This work investigates the application of Machine Unlearning (MU) for mitigating the impact of trojans embedded in conventional large language models of natural language (Text-LLMs) and large language models of code (Code-LLMs) We propose a novel unlearning approach, LYA, that leverages both gradient ascent and elastic weight consolidation, a Fisher Information Matrix (FIM) based regularization technique, to unlearn trojans from poisoned models. We compare the effectiveness of LYA against conventional techniques like fine-tuning, retraining, and vanilla gradient ascent. The subject models we investigate are BERT and CodeBERT, for sentiment analysis and code defect detection tasks, respectively. Our findings demonstrate that the combination of gradient ascent and FIM-based regularization, as done in LYA, outperforms existing methods in removing the trojan's influence from the poisoned model, while preserving its original functionality. To the best of our knowledge, this is the first work that compares and contrasts MU of trojans in LLMs, in the NL and Coding domain.

📄 PDF Abstract BibTeX arXiv:2408.12416

Code (0)

등록된 구현이 없습니다.

Tasks

Defect DetectionMachine UnlearningSentiment Analysis

Methods 이 논문이 사용한 방법론

Refunds@Expedia|||How do I get a full refund from Expedia? “How do I get a full refund from Expedia? How do I get a full refund from Expedia? – Call ☎️ +1-(888) 829 (0881) or +1-805-330-4056 or +1-805-330-4056 for Quick Help &…
Softmax The Softmax output function transforms a previous layer's output into a vector of probabilities. It is commonly used for multiclass classification. Given an input vector $x$…
Linear Layer A Linear Layer is a projection $\mathbf{XW + b}$.
Dropout Dropout is a regularization technique for neural networks that drops a unit (along with connections) at training time with a specified probability $p$ (a common value is…
WordPiece 설명 없음
Residual Connection 설명 없음
Multi-Head Attention 설명 없음
Linear Warmup With Linear Decay Linear Warmup With Linear Decay is a learning rate schedule in which we increase the learning rate linearly for $n$ updates and then linearly decay afterwards.

Similar Papers 제목 키워드 기반

UCD: Unlearning in LLMs via Contrastive Decoding

2025-06-12 · Vinith M. Suriyakumar, Ayush Sekhari, Ashia Wilson

Machine unlearning aims to remove specific information, e.g. sensitive or undesirable content, from large language models (LLMs) while preserving overall performance. We propose an inference-time unlearning algorithm tha…

Machine Unlearning

Trojans in Large Language Models of Code: A Critical Review through a Trigger-Based Taxonomy

2024-05-05 · Aftab Hussain, Md Rafiqul Islam Rabin, Toufique Ahmed, Bowen Xu 외

Large language models (LLMs) have provided a lot of exciting new capabilities in software development. However, the opaque nature of these models makes them difficult to reason about and inspect. Their opacity gives rise…

If You Don't Understand It, Don't Use It: Eliminating Trojans with Filters Between Layers

2024-07-08 · Adriano Hernandez

Large language models (LLMs) sometimes exhibit dangerous unintended behaviors. Finding and fixing these is challenging because the attack surface is massive -- it is not tractable to exhaustively search for all possible …

Data Poisoning

Oblivionis: A Lightweight Learning and Unlearning Framework for Federated Large Language Models

2025-08-12 · Fuyao Zhang, Xinyu Yan, Tiantong Wu, Wenjie Li 외 arxiv

Large Language Models (LLMs) increasingly leverage Federated Learning (FL) to utilize private, task-specific datasets for fine-tuning while preserving data privacy. However, while federated LLM frameworks effectively ena…

Federated Learning

Concept-ROT: Poisoning Concepts in Large Language Models with Model Editing

2024-12-17 · Keltin Grimes, Marco Christiani, David Shriver, Marissa Connor

Model editing methods modify specific behaviors of Large Language Models by altering a small, targeted set of network weights and require very little data and compute. These methods can be used for malicious applications…

MisinformationModel Editing