DOSE: Data Selection for Multi-Modal LLMs via Off-the-Shelf Models

arXiv:2604.16979v1 Announce Type: cross Abstract: High-quality and diverse multimodal data are essential for improving vision-language models (VLMs), yet existing datasets often contain noisy, redundant, and poorly aligned samples. To address these problems, data filtering is commonly used to enhance the efficiency and performance of multimodal learning, but it introduces extra computational cost because filtering models are usually trained on the same data they are meant to screen. To reduce this cost, we study DOSE, which explores whether off-the-shelf pretrained models that have never seen the target data can be used to select training samples for larger and stronger multimodal models without any task-specific training. Even without fine-tuning, these models can effectively assess text quality and image-text alignment to guide data selection. Based on this, we build a joint quality-alignment distribution and apply adaptive weighted sampling to select informative samples while maintaining long-tail diversity. This approach enhances data diversity, enabling models trained on DOSE-filtered data to match or surpass those trained on the full dataset on standard VQA and math benchmarks. Extensive experiments demonstrate its effectiveness, efficiency, and scalability.

Related

• ChartNet: A Million-Scale, High-Quality Multimodal Dataset for Robust Chart Understanding
• Reasoning Resides in Layers: Restoring Temporal Reasoning in Video-Language Models with Layer-Selective Merging
• Revisiting Compositionality in Dual-Encoder Vision-Language Models: The Role of Inference
• Seeing but Not Thinking: Routing Distraction in Multimodal Mixture-of-Experts
• Reasoning Within the Mind: Dynamic Multimodal Interleaving in Latent Space

Source: arXiv cs.CL | 2026-04-21