DecoupleMix: Decoupled Ratio Search and Convex Allocation for Scalable VLM Data Recipes

• URL: http://arxiv.org/abs/2607.24516v1
• Date: Mon, 27 Jul 2026 14:55:04 GMT
• Title: DecoupleMix: Decoupled Ratio Search and Convex Allocation for Scalable VLM Data Recipes
• Abstract summary: We formulate data construction as a systematic mixture-optimization problem.<n>For inter-class allocation, we apply a multidimensional, dataset-level assessment scoring Quality and Difficulty.<n>For intra-class composition, we apply a multidimensional, dataset-level assessment scoring Quality and Difficulty.
• Score: 10.902458329410562
• License: http://creativecommons.org/licenses/by/4.0/
• Abstract: While data curation for Vision Language Models (VLMs) is increasingly active, public practice for constructing pretraining mixtures remains largely heuristic: practitioners stack datasets that pass quality filters, set cross-domain ratios by intuition, and lack a principled, attributable criterion for admitting new data, while frontier recipes remain undisclosed. We formulate data construction as a systematic mixture-optimization problem and turn it into a reproducible engineering discipline by decoupling the mixture into two orthogonal sub-problems: inter-class ratios across capabilities and intra-class ratios within a category. For inter-class allocation, we use a single-variable iterative search; for intra-class composition, we apply a multidimensional, dataset-level assessment scoring Quality and Difficulty, and formulate selection as a constrained convex optimization with a diversity objective. The DecoupleMix framework delivers two critical capabilities: guiding what data to collect next and rendering dataset validation a controlled, attributable experiment. Experiments show our approach consistently surpasses heuristic baselines. Moreover, optimal ratios discovered on small-scale proxies transfer seamlessly to larger scales without retuning. Using 80B additional multimodal continue-pretraining tokens, our VLM is competitive with strong open-source models trained with substantially larger multimodal budgets.

Related papers

• Multimodal Data Curation Through Ranked Retrieval [0.011966041609090325]
  We present a framework that improves alignment by acting on both the training pairs and the embedding model.<n>We demonstrate that this approach collapses the modality gap by over 90% on average vs base embedding experts.
  arXiv  Detail & Related papers  (2026-05-01T23:45:36Z)
• Data Mixing for Large Language Models Pretraining: A Survey and Outlook [52.66761353708619]
  Large language models (LLMs) rely on pretraining on massive and heterogeneous corpora.<n>Unlike sample-level data selection, data mixing optimize domain-level sampling weights to allocate limited budgets more effectively.<n>This paper provides a comprehensive review of data mixing for LLM pretraining.
  arXiv  Detail & Related papers  (2026-03-25T13:30:40Z)
• MoDoMoDo: Multi-Domain Data Mixtures for Multimodal LLM Reinforcement Learning [37.71233459623324]
  Reinforcement Learning with Verifiable Rewards (RLVR) has recently emerged as a powerful paradigm for post-training large language models (LLMs)<n>Applying RLVR to Multimodal LLMs (MLLMs) presents significant opportunities but is complicated by the broader, heterogeneous nature of vision-language tasks.<n>We introduce a systematic post-training framework for Multimodal LLM RLVR, featuring a rigorous data mixture problem formulation and benchmark implementation.
  arXiv  Detail & Related papers  (2025-05-30T17:59:38Z)
• Multi-Level Aware Preference Learning: Enhancing RLHF for Complex Multi-Instruction Tasks [81.44256822500257]
  RLHF has emerged as a predominant approach for aligning artificial intelligence systems with human preferences.<n> RLHF exhibits insufficient compliance capabilities when confronted with complex multi-instruction tasks.<n>We propose a novel Multi-level Aware Preference Learning (MAPL) framework, capable of enhancing multi-instruction capabilities.
  arXiv  Detail & Related papers  (2025-05-19T08:33:11Z)
• Add-One-In: Incremental Sample Selection for Large Language Models via a Choice-Based Greedy Paradigm [50.492124556982674]
  This paper introduces a novel choice-based sample selection framework.<n>It shifts the focus from evaluating individual sample quality to comparing the contribution value of different samples.<n>We validate our approach on a larger medical dataset, highlighting its practical applicability in real-world applications.
  arXiv  Detail & Related papers  (2025-03-04T07:32:41Z)
• Hybrid-Regularized Magnitude Pruning for Robust Federated Learning under Covariate Shift [2.298932494750101]
  We show that inconsistencies in client-side training distributions substantially degrade the performance of federated learning models.<n>We propose a novel FL framework using a combination of pruning and regularisation of clients' training to improve the sparsity, redundancy, and robustness of neural connections.
  arXiv  Detail & Related papers  (2024-12-19T16:22:37Z)
• Adapt-$\infty$: Scalable Continual Multimodal Instruction Tuning via Dynamic Data Selection [89.42023974249122]
  Adapt-$infty$ is a new multi-way and adaptive data selection approach for lifelong instruction tuning.<n>We construct pseudo-skill clusters by grouping gradient-based sample vectors.<n>We select the best-performing data selector for each skill cluster from a pool of selector experts.<n>This data selector samples a subset of the most important samples from each skill cluster for training.
  arXiv  Detail & Related papers  (2024-10-14T15:48:09Z)
• A Framework for Fine-Tuning LLMs using Heterogeneous Feedback [69.51729152929413]
  We present a framework for fine-tuning large language models (LLMs) using heterogeneous feedback. First, we combine the heterogeneous feedback data into a single supervision format, compatible with methods like SFT and RLHF. Next, given this unified feedback dataset, we extract a high-quality and diverse subset to obtain performance increases.
  arXiv  Detail & Related papers  (2024-08-05T23:20:32Z)
• Tackling Diverse Minorities in Imbalanced Classification [80.78227787608714]
  Imbalanced datasets are commonly observed in various real-world applications, presenting significant challenges in training classifiers. We propose generating synthetic samples iteratively by mixing data samples from both minority and majority classes. We demonstrate the effectiveness of our proposed framework through extensive experiments conducted on seven publicly available benchmark datasets.
  arXiv  Detail & Related papers  (2023-08-28T18:48:34Z)
• Learning with MISELBO: The Mixture Cookbook [62.75516608080322]
  We present the first ever mixture of variational approximations for a normalizing flow-based hierarchical variational autoencoder (VAE) with VampPrior and a PixelCNN decoder network. We explain this cooperative behavior by drawing a novel connection between VI and adaptive importance sampling. We obtain state-of-the-art results among VAE architectures in terms of negative log-likelihood on the MNIST and FashionMNIST datasets.
  arXiv  Detail & Related papers  (2022-09-30T15:01:35Z)

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.