Mix, Don't Tune: Bilingual Pre-Training Outperforms Hyperparameter Search in Data-Constrained Settings

• URL: http://arxiv.org/abs/2605.13225v1
• Date: Wed, 13 May 2026 09:17:51 GMT
• Title: Mix, Don't Tune: Bilingual Pre-Training Outperforms Hyperparameter Search in Data-Constrained Settings
• Abstract summary: We show how to improve language model pre-training in a data-constrained domain.<n>We use Arabic as the low-resource target and English as the auxiliary.<n>We quantify how much mixing helps: it boosts performance by an amount equivalent to 2--3$times$ the unique target data on validation loss and 2--13$times$ on downstream task accuracy.
• Score: 24.462817377406754
• License: http://creativecommons.org/licenses/by/4.0/
• Abstract: For most languages of the world, language model pre-training operates in a data-constrained regime where models must repeat their training data many times, degrading generalization. Two remedies exist: aggressive hyperparameter tuning such as high weight decay, and mixing in data from a high-resource auxiliary language to directly aid the low-resource target. While hyperparameter tuning regularizes the model by shrinking weights to restrict network capacity, auxiliary data mixing uses a tunable mixing ratio to expand the training distribution and diversify the training signal with new knowledge. Both offer a principled way to improve training in a data-constrained domain. We compare these levers systematically across four model scales from 150M to 1.43B parameters, using Arabic as the low-resource target and English as the auxiliary, over approximately 1000 pre-training runs. Three findings emerge. First, mixing yields larger improvements than hyperparameter tuning on both validation loss and downstream task accuracy, and the gap grows with model size. Second, we quantify how much mixing helps: it boosts performance by an amount equivalent to 2--3$\times$ the unique target data on validation loss and 2--13$\times$ on downstream task accuracy, with the gain scaling steeply with model size. Third, this divergence reveals that target-language validation loss systematically underestimates mixing's value. Mixing regularizes by diversifying the training signal and contributes knowledge the repeated target corpus cannot supply; validation loss captures only the first effect. Our practical recommendations are: mix in a high-resource language, prioritize the mixing ratio over hyperparameter tuning, and transfer hyperparameters from a small proxy model via $μ$P.

Related papers

• Data-Constrained Language Model Pretraining: Improved Regularization and Scaling Laws [41.26396818761427]
  We study data-constrained pretraining along two axes, regularization and scaling.<n>We propose SoftQ, a scaling law that couples model size and data size to capture their interaction under repeated data.
  arXiv  Detail & Related papers  (2026-06-05T04:10:09Z)
• TREX: Tokenizer Regression for Optimal Data Mixture [10.917621429052183]
  Tokenizer Regression for Optimal Data MiXture (TREX) is a regression-based framework that efficiently predicts the optimal data mixture for tokenizer training.<n>TREX trains small-scale proxy tokenizers on random mixtures, gathers their compression statistics, and learns to predict compression performance from data mixtures.<n>TReX's predicted mixtures outperform mixtures based on LLaMA3 and uniform distributions by up to 12% in both inand out-of-distribution compression efficiency.
  arXiv  Detail & Related papers  (2026-01-20T04:41:09Z)
• Diffusion Language Models are Super Data Learners [61.721441061210896]
  When unique data is limited, diffusion language models (DLMs) consistently surpass autoregressive (AR) models by training for more epochs.<n>We attribute the gains to three compounding factors: (1) any-order modeling, (2) super-dense compute from iterative bidirectional denoising, and (3) built-in Monte Carlo augmentation.
  arXiv  Detail & Related papers  (2025-11-05T08:17:42Z)
• Optimal Sparsity of Mixture-of-Experts Language Models for Reasoning Tasks [17.067788440109137]
  Mixture-of-Experts (MoE) models are now standard in state-of-the-art systems.<n>We investigate how MoE sparsity influences two distinct capability regimes: memorization skills and reasoning skills.
  arXiv  Detail & Related papers  (2025-08-26T04:31:28Z)
• SPIRE: Conditional Personalization for Federated Diffusion Generative Models [7.8583640700306585]
  Shared Backbone Personal Identity Representation Embeddings (SPIRE) is a framework that casts per client diffusion based generation as conditional generation in FL.<n>SPIRE factorizes the network into (i) a high capacity global backbone that learns a population level score function and (ii) lightweight, learnable client embeddings that encode local data statistics.<n>Our analysis also hints at how client embeddings act as biases that steer a shared score network toward personalized distributions.
  arXiv  Detail & Related papers  (2025-06-14T01:40:31Z)
• What Do Learning Dynamics Reveal About Generalization in LLM Reasoning? [83.83230167222852]
  We find that a model's generalization behavior can be effectively characterized by a training metric we call pre-memorization train accuracy. By connecting a model's learning behavior to its generalization, pre-memorization train accuracy can guide targeted improvements to training strategies.
  arXiv  Detail & Related papers  (2024-11-12T09:52:40Z)
• Data Mixing Laws: Optimizing Data Mixtures by Predicting Language Modeling Performance [55.872926690722714]
  We study the predictability of model performance regarding the mixture proportions in function forms.<n>We propose nested use of the scaling laws of training steps, model sizes, and our data mixing law.<n>Our method effectively optimize the training mixture of a 1B model trained for 100B tokens in RedPajama.
  arXiv  Detail & Related papers  (2024-03-25T17:14:00Z)
• Retraining-free Model Quantization via One-Shot Weight-Coupling Learning [41.299675080384]
  Mixed-precision quantization (MPQ) is advocated to compress the model effectively by allocating heterogeneous bit-width for layers. MPQ is typically organized into a searching-retraining two-stage process. In this paper, we devise a one-shot training-searching paradigm for mixed-precision model compression.
  arXiv  Detail & Related papers  (2024-01-03T05:26:57Z)
• Adapted Multimodal BERT with Layer-wise Fusion for Sentiment Analysis [84.12658971655253]
  We propose Adapted Multimodal BERT, a BERT-based architecture for multimodal tasks. adapter adjusts the pretrained language model for the task at hand, while the fusion layers perform task-specific, layer-wise fusion of audio-visual information with textual BERT representations. In our ablations we see that this approach leads to efficient models, that can outperform their fine-tuned counterparts and are robust to input noise.
  arXiv  Detail & Related papers  (2022-12-01T17:31:42Z)
• Parameter-Efficient Sparsity for Large Language Models Fine-Tuning [63.321205487234074]
  We propose a. sparse-efficient Sparse Training (PST) method to reduce the number of trainable parameters during sparse-aware training. Experiments with diverse networks (i.e., BERT, RoBERTa and GPT-2) demonstrate PST performs on par or better than previous sparsity methods.
  arXiv  Detail & Related papers  (2022-05-23T02:43:45Z)
• Improving Neural Machine Translation by Denoising Training [95.96569884410137]
  We present a simple and effective pretraining strategy Denoising Training DoT for neural machine translation. We update the model parameters with source- and target-side denoising tasks at the early stage and then tune the model normally. Experiments show DoT consistently improves the neural machine translation performance across 12 bilingual and 16 multilingual directions.
  arXiv  Detail & Related papers  (2022-01-19T00:11:38Z)

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.