MixAtlas: Uncertainty-aware Data Mixture Optimization for Multimodal LLM Midtraining

• URL: http://arxiv.org/abs/2604.14198v1
• Date: Fri, 03 Apr 2026 04:26:43 GMT
• Title: MixAtlas: Uncertainty-aware Data Mixture Optimization for Multimodal LLM Midtraining
• Abstract summary: MixAtlas produces benchmark-targeted data recipes that can be inspected, adapted, and transferred to new corpora.<n>We evaluate on 10 benchmarks spanning visual understanding, document reasoning, and multimodal reasoning.
• Score: 13.388276899120065
• License: http://arxiv.org/licenses/nonexclusive-distrib/1.0/
• Abstract: Domain reweighting can improve sample efficiency and downstream generalization, but data-mixture optimization for multimodal midtraining remains largely unexplored. Current multimodal training recipes tune mixtures along a single dimension, typically data format or task type. We introduce MixAtlas, a method that produces benchmark-targeted data recipes that can be inspected, adapted, and transferred to new corpora. MixAtlas decomposes the training corpus along two axes: image concepts (10 visual-domain clusters discovered via CLIP embeddings) and task supervision (5 objective types including captioning, OCR, grounding, detection, and VQA). Using small proxy models (Qwen2-0.5B) paired with a Gaussian-process surrogate and GP-UCB acquisition, MixAtlas searches the resulting mixture space with the same proxy budget as regression-based baselines but finds better-performing mixtures. We evaluate on 10 benchmarks spanning visual understanding, document reasoning, and multimodal reasoning. On Qwen2-7B, optimized mixtures improve average performance by 8.5%-17.6% over the strongest baseline; on Qwen2.5-7B, gains are 1.0%-3.3%. Both settings reach baseline-equivalent training loss in up to 2 times fewer steps. Recipes discovered on 0.5B proxies transfer to 7B-scale training across Qwen model families.

Related papers

• FastMix: Fast Data Mixture Optimization via Gradient Descent [55.109758242985556]
  Fastmix is a novel framework that automates data mixture discovery while training only a single proxy model.<n>We show that optimizing mixture ratios is mathematically equivalent to assigning per-source loss weights under uniform source sampling.<n>FASTmix outperforms baselines while drastically reducing search cost across pre-intensive and post-training.
  arXiv  Detail & Related papers  (2026-06-12T21:46:57Z)
• Decouple Searching from Training: Scaling Data Mixing via Model Merging for Large Language Model Pre-training [16.022416196267937]
  We propose Decouple Searching from Training Mix (DeMix), a novel framework that leverages model merging to predict optimal data ratios.<n>We show that DeMix breaks the trade-off between sufficiency, accuracy and efficiency, obtaining the optimal mixture with higher benchmark performance at lower search cost.
  arXiv  Detail & Related papers  (2026-01-31T14:27:46Z)
• MergeMix: Optimizing Mid-Training Data Mixtures via Learnable Model Merging [72.00014675808228]
  textbfMix determines optimal data mixing ratios by repurposing model merging weights as a high-fidelity, low-cost performance proxy.<n>Experiments on models with 8B and 16B parameters validate that MergeMix achieves performance comparable to or surpassing exhaustive manual tuning.
  arXiv  Detail & Related papers  (2026-01-25T14:31:57Z)
• CLIMB: CLustering-based Iterative Data Mixture Bootstrapping for Language Model Pre-training [63.07024608399447]
  We propose an automated framework that discovers, evaluates, and refines data mixtures in a pre-training setting.<n>We introduce ClimbLab, a filtered 1.2-trillion-token corpus with 20 clusters as a research playground, and ClimbMix, a compact yet powerful 400-billion-token dataset.
  arXiv  Detail & Related papers  (2025-04-17T17:58:13Z)
• C3PO: Critical-Layer, Core-Expert, Collaborative Pathway Optimization for Test-Time Expert Re-Mixing [21.119495676190127]
  Mixture-of-Experts (MoE) Large Language Models (LLMs) suffer from severely sub-optimal expert pathways.<n> naive expert selection learned from pretraining leaves a surprising 10-20% accuracy gap for improvement.<n>We develop a novel class of test-time optimization methods to re-weight or "re-mixing" the experts in different layers jointly for each test sample.
  arXiv  Detail & Related papers  (2025-04-10T17:59:56Z)
• Aioli: A Unified Optimization Framework for Language Model Data Mixing [74.50480703834508]
  We show that no existing method consistently outperforms a simple stratified sampling baseline in terms of average test perplexity.<n>We derive a new online method named Aioli, which directly estimates the mixing law parameters throughout training and uses them to dynamically adjust proportions.
  arXiv  Detail & Related papers  (2024-11-08T17:50:24Z)
• Harnessing Hard Mixed Samples with Decoupled Regularizer [69.98746081734441]
  Mixup is an efficient data augmentation approach that improves the generalization of neural networks by smoothing the decision boundary with mixed data. In this paper, we propose an efficient mixup objective function with a decoupled regularizer named Decoupled Mixup (DM) DM can adaptively utilize hard mixed samples to mine discriminative features without losing the original smoothness of mixup.
  arXiv  Detail & Related papers  (2022-03-21T07:12:18Z)
• Jigsaw Clustering for Unsupervised Visual Representation Learning [68.09280490213399]
  We propose a new jigsaw clustering pretext task in this paper. Our method makes use of information from both intra- and inter-images. It is even comparable to the contrastive learning methods when only half of training batches are used.
  arXiv  Detail & Related papers  (2021-04-01T08:09:26Z)

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.