AI Paper Atlas

Research map / Large language models / Language, safety and interpretability

OPUS: Towards Efficient and Principled Data Selection in Large Language Model Pre-training in Every Iteration

Shaobo Wang, Xuan Ouyang, Tianyi Xu, Yuzheng Hu, Jialin Liu, Guo Chen, Tianyu Zhang, Junhao Zheng, Kexin Yang, Xingzhang Ren, Dayiheng Liu, Linfeng Zhang

ICML 2026 · Large language models › Language, safety and interpretability · approach: LLM pruning and scaling data · 0 citations (OpenAlex)

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
use OPUS instead of random selectionGPT-2 XL, Muon · FineWeb 30Baverage accuracy: 40.29 -> 41.75Improved
use OPUS instead of random selectionGPT-2 Large, Muon · FineWeb 30Baverage accuracy: 39.72 -> 40.25Improved
use OPUS instead of random selectionGPT-2 Large, AdamW · FineWeb 30Baverage accuracy: 39.29 -> 41.43Improved
use OPUS instead of random selectionGPT-2 XL, AdamW · FineWeb 30Baverage accuracy: 40.36 -> 41.01Improved
select with OPUS from score-3 pool instead of random from score-3 poolGPT-2 XL, Muon · FineWeb-Edu 30Baverage accuracy: 41.81 -> 44.99Improved
select with OPUS from score-3 pool instead of random from score-3 poolGPT-2 Large, Muon · FineWeb-Edu 30Baverage accuracy: 41.17 -> 41.72Improved
OPUS selecting from lower-quality score-3 pool instead of score 4+5 poolGPT-2 XL, Muon · FineWeb-Edu 30Baverage accuracy: 42.42 -> 44.99Improved
OPUS selecting from lower-quality score-3 pool instead of score 4+5 poolGPT-2 Large, Muon · FineWeb-Edu 30Baverage accuracy: 42.04 -> 41.72Hurt
use OPUS instead of random selectionGPT-2 XL · OOD benchmarksaverage accuracy: 38.09 -> 40.07Improved
use OPUS instead of random selectionGPT-2 XL · BBHexact match: 9.87 -> 11.02Improved
use OPUS instead of random selectionGPT-2 · FineWeb-Edu domain perplexityaverage perplexity: 3.38 -> 3.35Improved
use OPUS instead of Random trained for 60B tokens (token budget)GPT-2 XL · FineWeb-Edu validation losstokens to reach loss: 60B -> 17BImproved

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