Saved in:
Bibliographic Details
Main Authors: Chen, Jiankang, Zhang, Tianke, Liu, Changyi, Ding, Haojie, Shi, Yaya, Cheng, Feng, Xiao, Huihui, Wen, Bin, Yang, Fan, Gao, Tingting, Zhang, Di
Format: Preprint
Published: 2025
Subjects:
Online Access:https://arxiv.org/abs/2502.09925
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866912232055504896
author Chen, Jiankang
Zhang, Tianke
Liu, Changyi
Ding, Haojie
Shi, Yaya
Cheng, Feng
Xiao, Huihui
Wen, Bin
Yang, Fan
Gao, Tingting
Zhang, Di
author_facet Chen, Jiankang
Zhang, Tianke
Liu, Changyi
Ding, Haojie
Shi, Yaya
Cheng, Feng
Xiao, Huihui
Wen, Bin
Yang, Fan
Gao, Tingting
Zhang, Di
contents Multimodal visual language models are gaining prominence in open-world applications, driven by advancements in model architectures, training techniques, and high-quality data. However, their performance is often limited by insufficient task-specific data, leading to poor generalization and biased outputs. Existing efforts to increase task diversity in fine-tuning datasets are hindered by the labor-intensive process of manual task labeling, which typically produces only a few hundred task types. To address this, we propose TaskGalaxy, a large-scale multimodal instruction fine-tuning dataset comprising 19,227 hierarchical task types and 413,648 samples. TaskGalaxy utilizes GPT-4o to enrich task diversity by expanding from a small set of manually defined tasks, with CLIP and GPT-4o filtering those that best match open-source images, and generating relevant question-answer pairs. Multiple models are employed to ensure sample quality. This automated process enhances both task diversity and data quality, reducing manual intervention. Incorporating TaskGalaxy into LLaVA-v1.5 and InternVL-Chat-v1.0 models shows substantial performance improvements across 16 benchmarks, demonstrating the critical importance of task diversity. TaskGalaxy is publicly released at https://github.com/Kwai-YuanQi/TaskGalaxy.
format Preprint
id arxiv_https___arxiv_org_abs_2502_09925
institution arXiv
publishDate 2025
record_format arxiv
spellingShingle TaskGalaxy: Scaling Multi-modal Instruction Fine-tuning with Tens of Thousands Vision Task Types
Chen, Jiankang
Zhang, Tianke
Liu, Changyi
Ding, Haojie
Shi, Yaya
Cheng, Feng
Xiao, Huihui
Wen, Bin
Yang, Fan
Gao, Tingting
Zhang, Di
Computer Vision and Pattern Recognition
Artificial Intelligence
Multimodal visual language models are gaining prominence in open-world applications, driven by advancements in model architectures, training techniques, and high-quality data. However, their performance is often limited by insufficient task-specific data, leading to poor generalization and biased outputs. Existing efforts to increase task diversity in fine-tuning datasets are hindered by the labor-intensive process of manual task labeling, which typically produces only a few hundred task types. To address this, we propose TaskGalaxy, a large-scale multimodal instruction fine-tuning dataset comprising 19,227 hierarchical task types and 413,648 samples. TaskGalaxy utilizes GPT-4o to enrich task diversity by expanding from a small set of manually defined tasks, with CLIP and GPT-4o filtering those that best match open-source images, and generating relevant question-answer pairs. Multiple models are employed to ensure sample quality. This automated process enhances both task diversity and data quality, reducing manual intervention. Incorporating TaskGalaxy into LLaVA-v1.5 and InternVL-Chat-v1.0 models shows substantial performance improvements across 16 benchmarks, demonstrating the critical importance of task diversity. TaskGalaxy is publicly released at https://github.com/Kwai-YuanQi/TaskGalaxy.
title TaskGalaxy: Scaling Multi-modal Instruction Fine-tuning with Tens of Thousands Vision Task Types
topic Computer Vision and Pattern Recognition
Artificial Intelligence
url https://arxiv.org/abs/2502.09925