Concept-skill Transferability-based Data Selection for Large Vision-Language Models
This addresses the problem of expensive training for large vision-language models, offering a scalable solution for researchers and practitioners, though it is incremental as it builds on existing data selection methods.
The paper tackles the high cost of training large vision-language models by introducing COINCIDE, a data selection technique that uses a small reference model to select diverse and transferable visual instruction tuning data, achieving comparable performance with only 20% of the LLaVA-1.5 dataset and 70% reduction in running time, and superior results with 16.7% of the Vision-Flan dataset.
Instruction tuning, or supervised finetuning on extensive task-specific data, is necessary for Large Vision-Language Models (LVLMs) to generalize well across a broad range of vision-language (VL) tasks. However, training on large VL datasets can become prohibitively expensive. In this work, we introduce COINCIDE, an effective and scalable data selection technique that uses a small model as a reference model to select visual instruction tuning data for efficient finetuning of a target LVLM, focusing on diversity and transferability. Specifically, we cluster the training data using internal activations from a small model, which identifies VL concept-skill compositions needed by a target LVLM. We then sample data from these diverse clusters by considering their density and transferability, or the ability to transfer well to other concept-skill compositions. This approach ensures the diversity of these compositions, which is vital for LVLM generalization. Extensive experiments demonstrate that COINCIDE achieves superior performance and data selection efficiency against 8 strong baselines on two distinct datasets: LLaVA-1.5 and Vision-Flan. Using only 20% of the LLaVA-1.5 dataset, COINCIDE achieves performance comparable to the LVLM finetuned on the whole dataset, with 70% reduction of the wall-clock running time. On the Vision-Flan dataset, our method achieves superior results with only 16.7% of the training data.