PulseAugur
中
实时 13:36:32
English(EN) LESSER: Post-Training Data Selection with Output-Layer Gradients

新的LESSER方法加速了LLM训练后数据选择

研究人员开发了一种名为LESSER的新方法,用于选择大型语言模型的训练后数据。该技术利用输出层梯度来近似数据选择过程,而输出层梯度的计算成本远低于全参数梯度。LESSER在监督微调方面的计算成本降低了高达9.7倍,在强化学习基准测试方面降低了3倍,同时保持了下游任务的性能。研究表明,即使单个样本在输出层梯度和全梯度之间的排名不同,该方法也能选择梯度一致的批次。 AI

影响 降低了LLM微调的计算成本,可能加速研究和开发。

排序理由 该集群描述了在arXiv上的一篇学术论文中提出的一种新方法。[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 →

新的LESSER方法加速了LLM训练后数据选择

本文如何被排名

Signal score
7 / 100
Composite score across the factors below. Higher = stronger signal that this story matters right now.
Newsworthiness bucket
Tool
该集群描述了在arXiv上的一篇学术论文中提出的一种新方法。[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
Same-day
Cluster formed today. Ranking reflects the current source set at time of score.

完整方法见我们的编辑标准。

报道来源 [1]

  1. arXiv cs.LG TIER_1 English(EN) · Lyuxin David Zhang, Eric Wong, Surbhi Goel, Anton Xue ·

    LESSER:基于输出层梯度的训练后数据选择

    arXiv:2610.03702v1 Announce Type: new Abstract: The choice of post-training data for large language models substantially affects downstream performance. Gradient-based data selection is a popular approach that ranks training data by how well their gradients align with those of a …