DataFlex:面向大语言模型以数据为中心训练的统一基准与评估平台

data-mixture 开源项目推荐(8 个)
data-mixture 话题在 GitHub 上没有填写简介,这里按 Star 数汇总最相关的热门仓库。
热门项目 全部 8
GitHub 共 8 个仓库,已全部展示DataFlex is a data-centric training framework that enhances model performance by either selecting the most influential samples, optimizing their weights, or adjusting their mixing ratios.
A Data-Centric RL and On-Policy Distillation Framework for LLM Post-Training
(ACL 2026 Main) LLMSurgeon recovers the pretraining data mixture of any LLM from only its generated text — no weights, no training data. A calibrated domain classifier plus label-shift correction de-blurs biased predictions. Ships with LLMScan, a benchmark on 8 open-source LLMs.
Repository containing paper's code, results, and LaTeX source.
Unofficial reproduction of Nemotron-CLIMB (CLustering-based Iterative Data Mixture Bootstrapping, NeurIPS 2025, arXiv:2504.13161) on Ascend NPUs — 5 rounds, ≈3,900 NPU-hours; learned mixtures beat every fixed-ratio baseline (+0.014–0.043 STEM); the selection mechanism is the weak link — fixed and validated live.
Exact token-weighted cluster totals and reproducible calibration for NVIDIA Nemotron-ClimbMix
LLM pretraining data mixture research insights
关于这个话题
- 清单取自 GitHub 话题页并按 Star 数排序,每次展示前 8 个项目;原始页面见 https://github.com/topics/data-mixture。
- 该话题在 GitHub 上共有 8 个仓库, 本站收录的是其中最热门的 8 个,不是全量列表。
- Star / Fork 每日自动刷新一次,实时数据以 GitHub 为准。
- 当前 8 个项目累计 3473 Star,覆盖 1 种主要语言,数据更新于 2026-11-08。
- 中文介绍来自 GitHub 官方话题说明(github/explore)的机器翻译,可点「查看原文」对照;译文缺失时会先展示原文。