English(EN) Clustering-Based Balanced Sampling and Allocation with Data Parallelism for High-Performance Fine-Tuning

新的CluSTER框架将LLM微调时间缩短70%

作者 PulseAugur 编辑部 · [1 个来源] · 2026-09-14 04:00

一个名为CluSTER的新框架已被开发出来，以提高大型语言模型（LLM）微调的效率。该方法使用梯度空间聚类来创建一个代表性的、缩减的数据集，解决了典型指令调优数据集中存在的冗余和不平衡问题。CluSTER确保在数据并行设置中跨不同数据集群和工作程序的平衡覆盖，从而在不牺牲模型质量的情况下将训练时间最多缩短69.6%。 AI

影响降低LLM训练时间和计算成本，可能加速模型开发和部署。

排序理由该条目是一篇研究论文，详细介绍了一种用于LLM微调的新框架。[lever_c_demoted from research: ic=1 ai=1.0]

在 arXiv cs.LG 阅读 →

AI 生成摘要 · Google Gemini · 来自 1 个来源。我们如何撰写摘要 →

本文如何被排名

Signal score: 24 / 100; Composite score across the factors below. Higher = stronger signal that this story matters right now.
Newsworthiness bucket: Tool; 该条目是一篇研究论文，详细介绍了一种用于LLM微调的新框架。[lever_c_demoted from research: ic=1 ai=1.0]
Source corroboration: Single-source cluster; Only one publisher covered this so far. Single-source stories can still rank when the publisher is high-authority, but they lack cross-source corroboration.
Topics: paper, infra; Editorial topic classification. Feeds into how the story surfaces on /topic/<slug> hub pages and into the per-entity coverage mix.
AI-industry relevance: High; Clearly on-topic for AI-industry coverage.
Story freshness: Breaking (< 6h); Fresh story with cross-source coverage still developing. Ranking may shift as more sources report.

完整方法见我们的编辑标准。

报道来源 [1]

arXiv cs.LG TIER_1 English(EN) · Hyunjin Kim, Youngeun Nam, Jaemin Han, Wonhyeok Choi, Jae-Gil Lee · 2026-09-14 04:00

基于聚类的平衡采样与分配结合数据并行实现高性能微调

arXiv:2609.12584v1 Announce Type: new Abstract: Instruction-tuning datasets for large language models (LLMs) are often large, redundant, and imbalanced, limiting efficient adaptation. Naive large-batch fine-tuning repeatedly includes overrepresented sample groups while weakly cov…

本文如何被排名

报道来源 [1]

基于聚类的平衡采样与分配结合数据并行实现高性能微调

相关实体

相关话题