AI Paper Atlas

Research map / Architectures and efficiency / RNNs and normalization

Target-Oriented Pretraining Data Selection via Neuron-Activated Graph

Zijun Wang, Haoqin Tu, Weidong Zhou, Yiyang Zhou, Xiaohuan Zhou, Bingni Zhang, Weiguo Feng, Taifeng Wang, Cihang Xie, Fengze Liu

ICML 2026 · Architectures and efficiency › RNNs and normalization · approach: Layer-wise analysis of language models · 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
select pretraining data by NAG similarity instead of random sampling1.2B LM pretraining · six-benchmark average, single-targetaverage accuracy: 44.9% -> 49.8%Improved
use NAG instead of BETR embedding similaritytarget-oriented data selection · single-target averageaverage accuracy: 48.8% -> 49.8%Improved
use NAG instead of BETR under naive multi-target mixturetarget-oriented data selection · multi-target averageaverage accuracy: 44.4% -> 48.5%Improved
restrict NAG to the last layer only (ablation)NAG (Qwen3-1.7B) · six-benchmark averageaverage accuracy: 49.8% -> 45.7%Hurt
use UP-projection neurons instead of K or DOWN projection neurons (ablation)NAG · HellaSwagaccuracy: 56.69 (K), 57.99 (DOWN) -> 60.55 (UP)Improved
deactivate NAG-selected neurons (0.12% of all) in the LLM (ablation)Qwen3-1.7B-Base · six-benchmark averageaverage accuracy: 60.6% -> 37.1%Hurt
deactivate an equal number of random neurons (ablation)Qwen3-1.7B-Base · six-benchmark averageaverage accuracy: 60.6% -> 60.4%No clear effect
add NAG ranking to FineWeb-Edu quality signalFineWeb-Edu · single-target averageaverage accuracy: 47.4% -> 49.6%Improved
lower filtering rate rf (keep top 5% instead of 20%) with NAG (ablation)NAG ranking · HellaSwagaccuracy: 60.5% -> 62.3%Improved
lower filtering rate rf for random selection (ablation)random sampling · HellaSwagaccuracy: 51.6% -> 50.9%Hurt
increase layerwise neuron ratio from 0.3% to 2.1% (ablation)NAG (Qwen3-1.7B) · HellaSwagaccuracyNo clear effect
use a larger backbone (Qwen3 1.7B/4B/8B) for NAG (ablation)NAG · HellaSwagaccuracyImproved

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