arXiv cs.CL· Peng Sun, Yi Yang, Mingshuo Chen, Chunxiao Li, Yao Zhu, Huawen Shen, Tianfan Fu·· 11 小时前AI 评分31
QMAC：将数据选择重新定义为语义流形上的分布匹配
Rethinking Data Selection as Distribution Matching on Semantic Manifolds
AI 导读
QMAC 提出一种质量感知的多频带特征函数匹配方法，将数据选择建模为分布匹配与质量过滤的联合优化问题，在多个频率尺度上匹配所选子集与全量数据的经验特征函数。研究覆盖 4 种数据类型、8 个目标模型和 44 个 benchmark，QMAC 在标准图文指令数据上优于现有数据选择基线，并在文本指令、图文推理和文本推理数据上大幅超过随机选择，在多种模型与预算组合下超过全量数据训练。
正文 · 原文
Abstract:Instruction tuning of large language models and large vision language models relies on massive training data, which incurs substantial computational and storage costs. Meanwhile, training data often contains redundancy, noise, and low quality samples. Existing data selection methods mostly rely on diversity, importance, or a combination of both, without explicitly constraining the distributional consistency between the selected subset and the full data in the embedding space. We propose QMAC, a quality aware multiband characteristic function matching method that formulates data selection as a joint optimization problem of distribution matching and quality filtering. QMAC matches the empirical characteristic functions of the selected subset and the full data across multiple frequency scales, capturing both global semantic coverage and local perturbations, while using task related quality scores to filter noisy and low quality samples. We systematically evaluate QMAC on 4 data types, 8 target models, and 44 benchmarks. Experimental results show that QMAC outperforms existing data selection baselines on standard image-text instruction data, substantially surpasses random selection on standard text instruction, image-text reasoning, and text reasoning data, and exceeds full data training in multiple model and budget combinations. The code will be made publicly available upon acceptance of the paper.
| Subjects: | Machine Learning (cs.LG); Computation and Language (cs.CL) |
| Cite as: | arXiv:2611.08495 [cs.LG] |
| (or arXiv:2611.08495v1 [cs.LG] for this version) | |
| https://doi.org/10.48550/arXiv.2611.08495 arXiv-issued DOI via DataCite |
Submission history
From: Peng Sun [view email]
[v1]
Wed, 5 Aug 2026 04:07:16 UTC (17,414 KB)
来源：arXiv cs.CL · arxiv.org