LEAD: Iterative Data Selection for Efficient LLM Instruction Tuning
Fuente:
arXiv
Saved in:
| Main Authors: | , , , , , , |
|---|---|
| Format: | Preprint |
| Published: |
2025
|
| Subjects: | |
| Online Access: | |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| _version_ | 1866913832532705280 |
|---|---|
| author | Lin, Xiaotian Qi, Yanlin Zhu, Yizhang Palpanas, Themis Chai, Chengliang Tang, Nan Luo, Yuyu |
| author_facet | Lin, Xiaotian Qi, Yanlin Zhu, Yizhang Palpanas, Themis Chai, Chengliang Tang, Nan Luo, Yuyu |
| contents | Instruction tuning has emerged as a critical paradigm for improving the capabilities and alignment of large language models (LLMs). However, existing iterative model-aware data selection methods incur significant computational overhead, as they rely on repeatedly performing full-dataset model inference to estimate sample utility for subsequent training iterations, creating a fundamental efficiency bottleneck. In this paper, we propose LEAD, an efficient iterative data selection framework that accurately estimates sample utility entirely within the standard training loop, eliminating the need for costly additional model inference. At its core, LEAD introduces Instance-Level Dynamic Uncertainty (IDU), a theoretically grounded utility function combining instantaneous training loss, gradient-based approximation of loss changes, and exponential smoothing of historical loss signals. To further scale efficiently to large datasets, LEAD employs a two-stage, coarse-to-fine selection strategy, adaptively prioritizing informative clusters through a multi-armed bandit mechanism, followed by precise fine-grained selection of high-utility samples using IDU. Extensive experiments across four diverse benchmarks show that LEAD significantly outperforms state-of-the-art methods, improving average model performance by 6.1%-10.8% while using only 2.5% of the training data and reducing overall training time by 5-10x. |
| format | Preprint |
| id |
arxiv_https___arxiv_org_abs_2505_07437 |
| institution | arXiv |
| publishDate | 2025 |
| record_format | arxiv |
| spellingShingle | LEAD: Iterative Data Selection for Efficient LLM Instruction Tuning Lin, Xiaotian Qi, Yanlin Zhu, Yizhang Palpanas, Themis Chai, Chengliang Tang, Nan Luo, Yuyu Machine Learning Artificial Intelligence Databases Instruction tuning has emerged as a critical paradigm for improving the capabilities and alignment of large language models (LLMs). However, existing iterative model-aware data selection methods incur significant computational overhead, as they rely on repeatedly performing full-dataset model inference to estimate sample utility for subsequent training iterations, creating a fundamental efficiency bottleneck. In this paper, we propose LEAD, an efficient iterative data selection framework that accurately estimates sample utility entirely within the standard training loop, eliminating the need for costly additional model inference. At its core, LEAD introduces Instance-Level Dynamic Uncertainty (IDU), a theoretically grounded utility function combining instantaneous training loss, gradient-based approximation of loss changes, and exponential smoothing of historical loss signals. To further scale efficiently to large datasets, LEAD employs a two-stage, coarse-to-fine selection strategy, adaptively prioritizing informative clusters through a multi-armed bandit mechanism, followed by precise fine-grained selection of high-utility samples using IDU. Extensive experiments across four diverse benchmarks show that LEAD significantly outperforms state-of-the-art methods, improving average model performance by 6.1%-10.8% while using only 2.5% of the training data and reducing overall training time by 5-10x. |
| title | LEAD: Iterative Data Selection for Efficient LLM Instruction Tuning |
| topic | Machine Learning Artificial Intelligence Databases |
| url | https://arxiv.org/abs/2505.07437 |