paper-with-me

홈 › Papers

Revisiting Replay and Gradient Alignment for Continual Pre-Training of Large Language Models

2025-08-03 · Istabrak Abbes, Gopeshh Subbaraj, Matthew Riemer, Nizar Islah, Benjamin Therien, Tsuguchika Tabaru, Hiroaki Kingetsu, Sarath Chandar, Irina Rish arxiv

Training large language models (LLMs) typically involves pre-training on massive corpora, only to restart the process entirely when new data becomes available. A more efficient and resource-conserving approach would be continual pre-training, where models are updated with new data rather than retraining from scratch. However, the introduction of new data often causes distribution shifts, leading to performance degradation on previously learned tasks. In this paper, we take a deeper look at two popular proposals for addressing this distribution shift within the continual learning literature: experience replay and gradient alignment. We consider continual pre-training of models within the Llama family of architectures at a large scale across languages with 100 billion tokens of training data in each language, finding that both replay and gradient alignment lead to more stable learning without forgetting. This conclusion holds both as we vary the model scale and as we vary the number and diversity of tasks. Moreover, we are the first to demonstrate the effectiveness of gradient alignment techniques in the context of LLM pre-training and propose an efficient implementation of meta-experience replay (MER) that imbues experience replay with the benefits of gradient alignment despite negligible compute and memory overhead. Our scaling analysis across model sizes and replay rates indicates that small rates of replaying old examples are definitely a more valuable use of compute than investing in model size, but that it is more compute efficient to scale the size of the model than invest in high rates of replaying old examples.

📄 PDF Abstract BibTeX arXiv:2508.01908

Code (0)

등록된 구현이 없습니다.

Tasks

Continual Learning

Similar Papers 제목 키워드 기반

Revisiting Softmax Masking: Stop Gradient for Enhancing Stability in Replay-based Continual Learning

2023-09-26 · Hoyong Kim, Minchan Kwon, Kangil Kim

In replay-based methods for continual learning, replaying input samples in episodic memory has shown its effectiveness in alleviating catastrophic forgetting. However, the potential key factor of cross-entropy loss with …

Continual LearningIncremental Learning

Task-Agnostic Federated Continual Learning via Replay-Free Gradient Projection

2025-09-25 · Seohyeon Cha, Huancheng Chen, Haris Vikalo arxiv

Federated continual learning (FCL) enables collaborative model training across distributed clients on sequentially arriving tasks without revisiting past data. However, existing approaches often suffer from catastrophic …

Continual Learning

SHARP: Sparsity and Hidden Activation RePlay for Neuro-Inspired Continual Learning

2023-05-29 · Mustafa Burak Gurbuz, Jean Michael Moorman, Constantine Dovrolis

Deep neural networks (DNNs) struggle to learn in dynamic environments since they rely on fixed datasets or stationary environments. Continual learning (CL) aims to address this limitation and enable DNNs to accumulate kn…

Allclass-incremental learningClass Incremental LearningContinual Learning+1

GeRe: Towards Efficient Anti-Forgetting in Continual Learning of LLM via General Samples Replay

2025-08-06 · Yunan Zhang, Shuoran Jiang, Mengchen Zhao, Yuefeng Li 외 arxiv

The continual learning capability of large language models (LLMs) is crucial for advancing artificial general intelligence. However, continual fine-tuning LLMs across various domains often suffers from catastrophic forge…

Continual Learning

Learning to Learn without Forgetting by Maximizing Transfer and Minimizing Interference

2018-10-29 · ICLR 2019 5 · Matthew Riemer, Ignacio Cases, Robert Ajemian, Miao Liu 외

Lack of performance when it comes to continual learning over non-stationary distributions of data remains a major challenge in scaling neural network learning to more human realistic settings. In this work we propose a n…

Continual LearningMeta-LearningReinforcement Learning