AI Paper Atlas

Research map / Architectures and efficiency / RNNs and normalization

Evolution of Concepts in Language Model Pre-Training

Xuyang Ge, Wentao Shu, Jiaxing Wu, Yunhua Zhou, Zhengfu He, Xipeng Qiu

ICLR 2026 · Architectures and efficiency › RNNs and normalization · approach: Layer-wise analysis of language models · 6 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
increase crosscoder dictionary size (ablation)crosscoder · Pythia activationsexplained variance and L0Improved
train one crosscoder over snapshots instead of per-snapshot SAEssparse dictionary learning · Pythia per-snapshot activationsexplained variance, L0No clear effect
crosscoder instead of SAE on final snapshotsparse dictionary learning · Pythia final snapshotPareto frontierImproved
later emergence stepcrosscoder features · Pythia-160Mfeature complexity score: Pearson r = 0.309Improved
ablate top-ranked attributed features (tens of features) (ablation)Pythia-160M crosscoder · SVA, induction, IOItask metricHurt
keep only top-ranked attributed features (ablation)Pythia-160M crosscoder · SVA, induction, IOItask metric recoveryImproved

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