Data Organization Matters in Multimodal Instruction Tuning: A Controlled Study of Capability Trade-offs

• URL: http://arxiv.org/abs/2603.27744v1
• Date: Sun, 29 Mar 2026 15:54:06 GMT
• Title: Data Organization Matters in Multimodal Instruction Tuning: A Controlled Study of Capability Trade-offs
• Abstract summary: We study whether data organization affects the trade-off among general understanding, structured reasoning, and fine-grained OCR/document understanding.<n>We compare four strategies: direct mixture, curriculum training, balanced sampling, and reverse curriculum.
• Score: 0.0
• License: http://creativecommons.org/licenses/by-nc-nd/4.0/
• Abstract: Recent multimodal large language models (MLLMs) perform strongly on general visual understanding, diagram and chart reasoning, and document-centric perception. However, these abilities are learned from heterogeneous supervision sources with very different task structures and learning demands, and the effect of their temporal organization during training remains underexplored. We study whether data organization affects the trade-off among general understanding, structured reasoning, and fine-grained OCR/document understanding in multimodal instruction tuning. To isolate this factor, we use a controlled three-stage training framework in which the backbone, trainable modules, and optimization pipeline are fixed across all runs, and only the temporal arrangement of post-alignment supervision is changed. We compare four strategies: direct mixture, curriculum training, balanced sampling, and reverse curriculum. Experiments on general visual instruction following, diagram reasoning, chart reasoning, scene-text question answering, and document question answering show that data organization is a first-order design variable in multimodal adaptation. Curriculum training gives the best overall trade-off and the strongest structured reasoning performance. Balanced sampling is better for OCR-oriented capability but weakens the broader capability balance. Reverse curriculum performs worst in both final performance and optimization stability. Training-dynamics analysis further suggests that building general understanding and reasoning before introducing OCR-intensive supervision leads to smoother optimization and faster convergence. These findings highlight data scheduling as an explicit design dimension for multimodal model adaptation.

Related papers

• Self-Supervised Visual Representation Learning: Pretrain-Finetuning or Joint Training? [0.07340017786387766]
  Self-supervision is a powerful technique for learning visual representations from unlabeled data.<n>Existing techniques adopt a two-stage approach for self-supervised learning (SSL): a pretraining stage on unlabeled data followed by a finetuning stage on labeled data.<n>We investigate whether jointly optimizing the self-supervised and supervised objectives during training provides a better alternative.
  arXiv  Detail & Related papers  (2026-07-14T18:45:04Z)
• DocSeeker: Structured Visual Reasoning with Evidence Grounding for Long Document Understanding [63.257540233507626]
  We propose a paradigm that requires the model to execute a structured Analysis, Localization and Reasoning workflow.<n>We show it robustly generalizes from short-page training to ultra-long documents and is naturally synergistic with visual Retrieval-Augmented Generation systems.
  arXiv  Detail & Related papers  (2026-04-14T14:39:26Z)
• Agentic Planning with Reasoning for Image Styling via Offline RL [66.10749901925941]
  Direct prompt-based editing fails on complex transformations because vague and subjective prompts often require nuanced understanding of what should be changed in the image.<n>We present a tool-based agentic RL post-training framework that addresses this through structured planning with chain-of-thought reasoning.
  arXiv  Detail & Related papers  (2026-03-07T11:14:37Z)
• Start Small, Think Big: Curriculum-based Relative Policy Optimization for Visual Grounding [23.138205646078536]
  Chain-of-Thought (CoT) prompting has recently shown significant promise across various NLP and computer vision tasks.<n>We find that reinforcement learning (RL)-based fine-tuned CoT reasoning can paradoxically degrade performance in Visual Grounding tasks.<n>We propose Curriculum-based Relative Policy Optimization (CuRPO), a novel training strategy that leverages CoT length and generalized Intersection over Union rewards.
  arXiv  Detail & Related papers  (2025-11-17T21:22:50Z)
• Revisit the Imbalance Optimization in Multi-task Learning: An Experimental Analysis [44.410446932443]
  Multi-task learning (MTL) aims to build general-purpose vision systems by training a single network to perform multiple tasks jointly.<n>While promising, its potential is often hindered by "unbalanced optimization"<n>This paper presents a systematic experimental analysis to dissect the factors contributing to this persistent problem.
  arXiv  Detail & Related papers  (2025-09-28T14:40:06Z)
• Exploring the Effect of Reinforcement Learning on Video Understanding: Insights from SEED-Bench-R1 [53.894789613838654]
  We introduce SEED-Bench-R1, a benchmark designed to evaluate post-training methods for MLLMs in video understanding.<n>It includes intricate real-world videos and complex everyday planning tasks in the format of multiple-choice questions.<n>Using Qwen2-VL-Instruct-7B as a base model, we compare RL with supervised fine-tuning (SFT)<n>Our detailed analysis reveals that RL enhances visual perception but often produces less coherent reasoning chains.
  arXiv  Detail & Related papers  (2025-03-31T17:55:23Z)
• Reason-RFT: Reinforcement Fine-Tuning for Visual Reasoning of Vision Language Models [42.75418134743927]
  Reason-RFT is a two-stage reinforcement fine-tuning framework for visual reasoning.<n>First,Supervised Fine-Tuning (SFT) with curated CoT data activates the reasoning potential of Vision-Language Models (VLMs)<n>Second, reinforcement learning based on Group Relative Policy Optimization (GRPO) generates multiple reasoning-response pairs to enhance adaptability to domain shifts.
  arXiv  Detail & Related papers  (2025-03-26T17:38:06Z)
• Will Pre-Training Ever End? A First Step Toward Next-Generation Foundation MLLMs via Self-Improving Systematic Cognition [89.50068130832635]
  Self-Improving cognition (SIcog) is a self-learning framework for constructing next-generation foundation MLLMs by multimodal knowledge.<n>We propose Chain-of-Description for step-by-step visual understanding and integrate structured Chain-of-Thought (CoT) reasoning to support in-depth multimodal reasoning.<n>Experiments demonstrate SIcog's effectiveness in developing MLLMs with enhanced multimodal cognition.
  arXiv  Detail & Related papers  (2025-03-16T00:25:13Z)
• RaCT: Ranking-aware Chain-of-Thought Optimization for LLMs [30.216174551427443]
  Large language models (LLMs) have demonstrated remarkable potential in text reranking tasks.<n> conventional supervised fine-tuning approaches for specializing LLMs in ranking tasks often lead to significant degradation of the models' general-purpose abilities.<n>This paper presents a novel methodology that strategically combines Chain-of-Thought (CoT) prompting techniques with an innovative two-stage training pipeline.
  arXiv  Detail & Related papers  (2024-12-18T23:24:15Z)
• Deciphering Cross-Modal Alignment in Large Vision-Language Models with Modality Integration Rate [118.37653302885607]
  We present the Modality Integration Rate (MIR), an effective, robust, and generalized metric to indicate the multi-modal pre-training quality of Large Vision Language Models (LVLMs) MIR is indicative about training data selection, training strategy schedule, and model architecture design to get better pre-training results.
  arXiv  Detail & Related papers  (2024-10-09T17:59:04Z)
• In-context Demonstration Matters: On Prompt Optimization for Pseudo-Supervision Refinement [71.60563181678323]
  Large language models (LLMs) have achieved great success across diverse tasks, and fine-tuning is sometimes needed to further enhance generation quality.<n>To handle these challenges, a direct solution is to generate high-confidence'' data from unsupervised downstream tasks.<n>We propose a novel approach, pseudo-supervised demonstrations aligned prompt optimization (PAPO) algorithm, which jointly refines both the prompt and the overall pseudo-supervision.
  arXiv  Detail & Related papers  (2024-10-04T03:39:28Z)
• Hierarchical Decomposition of Prompt-Based Continual Learning: Rethinking Obscured Sub-optimality [55.88910947643436]
  Self-supervised pre-training is essential for handling vast quantities of unlabeled data in practice. HiDe-Prompt is an innovative approach that explicitly optimize the hierarchical components with an ensemble of task-specific prompts and statistics. Our experiments demonstrate the superior performance of HiDe-Prompt and its robustness to pre-training paradigms in continual learning.
  arXiv  Detail & Related papers  (2023-10-11T06:51:46Z)

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.