Curriculum-Guided Layer Scaling for Language Model Pretraining

• URL: http://arxiv.org/abs/2506.11389v1
• Date: Fri, 13 Jun 2025 01:22:16 GMT
• Title: Curriculum-Guided Layer Scaling for Language Model Pretraining
• Authors: Karanpartap Singh, Neil Band, Ehsan Adeli,
• Abstract summary: We propose Curriculum-Guided Layer Scaling (CGLS), a framework for compute-efficient pretraining.<n>CGLS synchronizes increasing data difficulty with model growth.<n>We show that increasing model depth leads to better generalization and zero-shot performance on various downstream benchmarks.
• Score: 8.195860140972615
• License: http://creativecommons.org/licenses/by/4.0/
• Abstract: As the cost of pretraining large language models grows, there is continued interest in strategies to improve learning efficiency during this core training stage. Motivated by cognitive development, where humans gradually build knowledge as their brains mature, we propose Curriculum-Guided Layer Scaling (CGLS), a framework for compute-efficient pretraining that synchronizes increasing data difficulty with model growth through progressive layer stacking (i.e. gradually adding layers during training). At the 100M parameter scale, using a curriculum transitioning from synthetic short stories to general web data, CGLS outperforms baseline methods on the question-answering benchmarks PIQA and ARC. Pretraining at the 1.2B scale, we stratify the DataComp-LM corpus with a DistilBERT-based classifier and progress from general text to highly technical or specialized content. Our results show that progressively increasing model depth alongside sample difficulty leads to better generalization and zero-shot performance on various downstream benchmarks. Altogether, our findings demonstrate that CGLS unlocks the potential of progressive stacking, offering a simple yet effective strategy for improving generalization on knowledge-intensive and reasoning tasks.

Related papers

• Scaling DRL for Decision Making: A Survey on Data, Network, and Training Budget Strategies [66.83950068218033]
  Scaling Laws demonstrate that scaling model parameters and training data enhances learning performance.<n>Despite its potential to improve performance, the integration of scaling laws into deep reinforcement learning has not been fully realized.<n>This review addresses this gap by systematically analyzing scaling strategies in three dimensions: data, network, and training budget.
  arXiv  Detail & Related papers  (2025-08-05T08:03:12Z)
• LESA: Learnable LLM Layer Scaling-Up [57.0510934286449]
  Training Large Language Models (LLMs) from scratch requires immense computational resources, making it prohibitively expensive.<n>Model scaling-up offers a promising solution by leveraging the parameters of smaller models to create larger ones.<n>We propose textbfLESA, a novel learnable method for depth scaling-up.
  arXiv  Detail & Related papers  (2025-02-19T14:58:48Z)
• Dynamic Loss-Based Sample Reweighting for Improved Large Language Model Pretraining [55.262510814326035]
  Existing reweighting strategies primarily focus on group-level data importance.<n>We introduce novel algorithms for dynamic, instance-level data reweighting.<n>Our framework allows us to devise reweighting strategies deprioritizing redundant or uninformative data.
  arXiv  Detail & Related papers  (2025-02-10T17:57:15Z)
• SLCA++: Unleash the Power of Sequential Fine-tuning for Continual Learning with Pre-training [68.7896349660824]
  We present an in-depth analysis of the progressive overfitting problem from the lens of Seq FT. Considering that the overly fast representation learning and the biased classification layer constitute this particular problem, we introduce the advanced Slow Learner with Alignment (S++) framework. Our approach involves a Slow Learner to selectively reduce the learning rate of backbone parameters, and a Alignment to align the disjoint classification layers in a post-hoc fashion.
  arXiv  Detail & Related papers  (2024-08-15T17:50:07Z)
• Structure-aware Domain Knowledge Injection for Large Language Models [38.08691252042949]
  StructTuning is a methodology to transform Large Language Models (LLMs) into domain specialists.<n>It significantly reduces the training corpus needs to a mere 5% while achieving an impressive 100% of traditional knowledge injection performance.
  arXiv  Detail & Related papers  (2024-07-23T12:38:48Z)
• SLCA: Slow Learner with Classifier Alignment for Continual Learning on a Pre-trained Model [73.80068155830708]
  We present an extensive analysis for continual learning on a pre-trained model (CLPM) We propose a simple but extremely effective approach named Slow Learner with Alignment (SLCA) Across a variety of scenarios, our proposal provides substantial improvements for CLPM.
  arXiv  Detail & Related papers  (2023-03-09T08:57:01Z)
• Exploiting Explainable Metrics for Augmented SGD [43.00691899858408]
  There are several unanswered questions about how learning under optimization really works and why certain strategies are better than others. We propose new explainability metrics that measure the redundant information in a network's layers. We then exploit these metrics to augment the Gradient Descent (SGD) by adaptively adjusting the learning rate in each layer to improve generalization performance.
  arXiv  Detail & Related papers  (2022-03-31T00:16:44Z)
• Self-Supervised Class Incremental Learning [51.62542103481908]
  Existing Class Incremental Learning (CIL) methods are based on a supervised classification framework sensitive to data labels. When updating them based on the new class data, they suffer from catastrophic forgetting: the model cannot discern old class data clearly from the new. In this paper, we explore the performance of Self-Supervised representation learning in Class Incremental Learning (SSCIL) for the first time.
  arXiv  Detail & Related papers  (2021-11-18T06:58:19Z)
• An EM Framework for Online Incremental Learning of Semantic Segmentation [37.94734474090863]
  We propose an incremental learning strategy that can adapt deep segmentation models without catastrophic forgetting, using a streaming input data with pixel annotations on the novel classes only. We validate our approach on the PASCAL VOC 2012 and ADE20K datasets, and the results demonstrate its superior performance over the existing incremental methods.
  arXiv  Detail & Related papers  (2021-08-08T11:30:09Z)

This list is automatically generated from the titles and abstracts of the papers in this site.

This site does not guarantee the quality of this site (including all information) and is not responsible for any consequences.