AI Paper Atlas

Research map / Large language models / Instruction and fine-tuning

Group-wise Data Ordering: Enhancing Instruction Tuning of Large Language Models via Embedding Proximity

Yiwen Ye, Boyuan Jiang, Xiaobin Hu, Shengzhi Wang, Xiaozhong Ji, Jinghao Lin, Deli Yu, Jiale Chen, Kai WU, Haihua Yang, Yong Xia

ICML 2026 · Large language models › Instruction and fine-tuning · approach: Instruction data selection

Read the paperCheck your idea against 137,217 papers

AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.

Measured effects reported in the paper

What was changedWhereMeasuredEffect
order clusters from easy to hard using embedding-proximity grouping with mixed regions and replay buffer, versus random shufflingLLaVA-Vicuna-7B full fine-tuning · 12 multimodal sub-benchmarks avgaverage score: 64.88 -> 66.30Improved
group-based ordering alone, without mixed regions (ablation)EP-Order · 12 multimodal sub-benchmarks avgaverage score: 64.88 -> 64.70No clear effect
add mixed regions between clusters (ablation)EP-Order · 12 multimodal sub-benchmarks avgaverage score: 64.70 -> 65.53Improved
add replay buffer data from earlier clusters to mixed regions (ablation)EP-Order · 12 multimodal sub-benchmarks avgaverage score: 65.53 -> 66.30Improved
use embedding-based grouping instead of random partitions with the same mixed region and replay buffer design (ablation)EP-Order · 12 multimodal sub-benchmarks avgaverage score: 64.59 -> 66.30Improved
use HDBSCAN (15 clusters) instead of k-means (ablation)EP-Order · 12 multimodal sub-benchmarks avgaverage score: 65.53 -> 66.30 (k=15 vs HDBSCAN)Improved
use too few (k=5) k-means clusters (ablation)EP-Order · 12 multimodal sub-benchmarks avgaverage score: 65.64 (k=10) -> 64.14 (k=5)Hurt
fold-wise cluster ordering instead of global easy-to-hard (ablation)EP-Order · 12 multimodal sub-benchmarks avgaverage score: 66.30 -> 64.90 (n=5)Hurt
order clusters hard-to-easy instead of easy-to-hard (ablation)EP-Order · 12 multimodal sub-benchmarks avgaverage score: 66.30 -> 65.60Hurt
sort data by loss from the warm-up model instead of random shuffleLLaVA-Vicuna-7B full fine-tuning · 12 multimodal sub-benchmarks avgaverage score: 64.88 -> 52.58Hurt
EP-Order versus random shuffling on LoRA fine-tuningLLaVA-Vicuna-7B LoRA · 12 multimodal sub-benchmarks avgaverage score: 64.38 -> 65.76Improved
EP-Order versus random shuffling on text-only instruction tuningQwen3-4B on Tulu-3-SFT · ARC, BBH, GSM, MMLU avgaverage accuracy: 74.81 -> 79.51Improved

Closest papers by meaning

Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.

Open scime draft

Get the next scime Dispatch and a monthly field update

Free. We send a confirmation link first; every email has a one-click unsubscribe. Privacy