R&B: Domain Regrouping and Data Mixture Balancing for Efficient Foundation Model Training

• URL: http://arxiv.org/abs/2505.00358v1
• Date: Thu, 01 May 2025 07:08:19 GMT
• Title: R&B: Domain Regrouping and Data Mixture Balancing for Efficient Foundation Model Training
• Authors: Albert Ge, Tzu-Heng Huang, John Cooper, Avi Trost, Ziyi Chu, Satya Sai Srinath Namburi GNVV, Ziyang Cai, Kendall Park, Nicholas Roberts, Frederic Sala,
• Abstract summary: We introduce R&B, a framework that re- Partitions training data based on semantic similarity to create finer-grained domains.<n>Unlike prior works, R&B removes the need for additional compute to obtain evaluation information such as losses or gradients.<n>We demonstrate the effectiveness of R&B on five diverse datasets ranging from natural language to reasoning and multimodal tasks.
• Score: 11.213419356901005
• License: http://creativecommons.org/licenses/by/4.0/
• Abstract: Data mixing strategies have successfully reduced the costs involved in training language models. While promising, such methods suffer from two flaws. First, they rely on predetermined data domains (e.g., data sources, task types), which may fail to capture critical semantic nuances, leaving performance on the table. Second, these methods scale with the number of domains in a computationally prohibitive way. We address these challenges via R&B, a framework that re-partitions training data based on semantic similarity (Regroup) to create finer-grained domains, and efficiently optimizes the data composition (Balance) by leveraging a Gram matrix induced by domain gradients obtained throughout training. Unlike prior works, it removes the need for additional compute to obtain evaluation information such as losses or gradients. We analyze this technique under standard regularity conditions and provide theoretical insights that justify R&B's effectiveness compared to non-adaptive mixing approaches. Empirically, we demonstrate the effectiveness of R&B on five diverse datasets ranging from natural language to reasoning and multimodal tasks. With as little as 0.01% additional compute overhead, R&B matches or exceeds the performance of state-of-the-art data mixing strategies.

Related papers

• DIDS: Domain Impact-aware Data Sampling for Large Language Model Training [41.86545248261005]
  We present Domain Impact-aware Data Sampling (DIDS) to optimize domain-level sampling strategies. DIDS achieves 3.4% higher average performance while maintaining comparable training efficiency.
  arXiv  Detail & Related papers  (2025-04-17T13:09:38Z)
• 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)
• Optimizing Pretraining Data Mixtures with LLM-Estimated Utility [52.08428597962423]
  Large Language Models improve with increasing amounts of high-quality training data. We find token-counts outperform manual and learned mixes, indicating that simple approaches for dataset size and diversity are surprisingly effective. We propose two complementary approaches: UtiliMax, which extends token-based $200s by incorporating utility estimates from reduced-scale ablations, achieving up to a 10.6x speedup over manual baselines; and Model Estimated Data Utility (MEDU), which leverages LLMs to estimate data utility from small samples, matching ablation-based performance while reducing computational requirements by $simx.
  arXiv  Detail & Related papers  (2025-01-20T21:10:22Z)
• Efficient Ensembles Improve Training Data Attribution [12.180392191924758]
  Training data attribution methods aim to quantify the influence of individual data points on model predictions, with broad applications in data-centric AI. Existing methods in this field, which can be categorized as retraining-based and gradient-based methods, have struggled with naive trade-off attribution efficacy. Recent research has shown that augmenting gradient-based methods with ensembles of multiple independently trained models can achieve significantly better attribution.
  arXiv  Detail & Related papers  (2024-05-27T15:58:34Z)
• BiMix: A Bivariate Data Mixing Law for Language Model Pretraining [47.77701041534746]
  The impact of pretraining data composition on model performance remains poorly understood. $textbfBiMix$ provides a systematic framework for understanding and optimizing data mixtures. Our work contributes both theoretical insights into data mixing dynamics and practical tools for enhancing LLM training efficiency.
  arXiv  Detail & Related papers  (2024-05-23T09:44:02Z)
• Group Distributionally Robust Dataset Distillation with Risk Minimization [17.05513836324578]
  We introduce an algorithm that combines clustering with the minimization of a risk measure on the loss to conduct DD.<n>We provide a theoretical rationale for our approach and demonstrate its effective generalization and robustness across subgroups.
  arXiv  Detail & Related papers  (2024-02-07T09:03:04Z)
• SALUDA: Surface-based Automotive Lidar Unsupervised Domain Adaptation [62.889835139583965]
  We introduce an unsupervised auxiliary task of learning an implicit underlying surface representation simultaneously on source and target data. As both domains share the same latent representation, the model is forced to accommodate discrepancies between the two sources of data. Our experiments demonstrate that our method achieves a better performance than the current state of the art, both in real-to-real and synthetic-to-real scenarios.
  arXiv  Detail & Related papers  (2023-04-06T17:36:23Z)
• Continual Contrastive Finetuning Improves Low-Resource Relation Extraction [34.76128090845668]
  Relation extraction has been particularly challenging in low-resource scenarios and domains. Recent literature has tackled low-resource RE by self-supervised learning. We propose to pretrain and finetune the RE model using consistent objectives of contrastive learning.
  arXiv  Detail & Related papers  (2022-12-21T07:30:22Z)
• Cluster-level pseudo-labelling for source-free cross-domain facial expression recognition [94.56304526014875]
  We propose the first Source-Free Unsupervised Domain Adaptation (SFUDA) method for Facial Expression Recognition (FER) Our method exploits self-supervised pretraining to learn good feature representations from the target data. We validate the effectiveness of our method in four adaptation setups, proving that it consistently outperforms existing SFUDA methods when applied to FER.
  arXiv  Detail & Related papers  (2022-10-11T08:24:50Z)
• DRFLM: Distributionally Robust Federated Learning with Inter-client Noise via Local Mixup [58.894901088797376]
  federated learning has emerged as a promising approach for training a global model using data from multiple organizations without leaking their raw data. We propose a general framework to solve the above two challenges simultaneously. We provide comprehensive theoretical analysis including robustness analysis, convergence analysis, and generalization ability.
  arXiv  Detail & Related papers  (2022-04-16T08:08:29Z)
• UDALM: Unsupervised Domain Adaptation through Language Modeling [79.73916345178415]
  We introduce UDALM, a fine-tuning procedure, using a mixed classification and Masked Language Model loss. Our experiments show that performance of models trained with the mixed loss scales with the amount of available target data can be effectively used as a stopping criterion. Our method is evaluated on twelve domain pairs of the Amazon Reviews Sentiment dataset, yielding $91.74%$ accuracy, which is an $1.11%$ absolute improvement over the state-of-versathe-art.
  arXiv  Detail & Related papers  (2021-04-14T19:05:01Z)

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.