AI Paper Atlas

Research map / Architectures and efficiency / Pruning and efficient tuning

Weight Decay Improves Language Model Plasticity

Tessa Han, Sebastian Bordt, Hanlin Zhang, Sham M. Kakade

ICML 2026 · Architectures and efficiency › Pruning and efficient tuning · approach: Parameter-efficient transfer analysis · 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
raise pretraining weight decay above the 0.1 default (to 1.0) (ablation)Llama-2 and OLMo-2 models at 20 TPP · CoT reasoning fine-tuning (six tasks, average)downstream accuracy: optimal pretraining weight decay is 1.0 versus default 0.1Improved
raise pretraining weight decay from 0.1 to 0.3 (ablation)OLMo-2-1B at 140 TPP · CoT reasoning fine-tuningdownstream accuracy: optimal pretraining weight decay is 0.3Improved
raise pretraining weight decay from 0.1 to 0.3 (ablation)OLMo-2-1B at 140 TPP · pretraining validation losscross-entropy loss: 2.6088 -> 2.6208Hurt
raise pretraining weight decay from 0.1 to 1.0 (ablation)OLMo-2-1B at 140 TPP · pretraining validation losscross-entropy loss: 2.6088 -> 2.7064Hurt
raise pretraining weight decay from 0.1 to 0.5 or 1.0 (ablation)Llama-2 20 TPP models · pretraining validation losscross-entropy loss: optimal weight decay 0.5 (0.5B, 1B), 0.6 (OLMo-2-1B), 1.0 (4B)Improved
raise pretraining weight decay to 10 (ablation)Llama-2-0.5B at 20 TPP · pretraining validation losscross-entropy lossHurt
raise pretraining weight decay from 0.1 to 0.5 (ablation)Llama-2-0.5B at 20 TPP · upstream 5-benchmark average (pretrained)accuracy: 0.40 -> 0.41No clear effect
raise pretraining weight decay from 0.1 to 1.0 (ablation)OLMo-2-1B at 20 TPP · upstream 5-benchmark average (pretrained)accuracy: 0.49 -> 0.50No clear effect
raise pretraining weight decay from 0.1 to 1.0 (ablation)OLMo-2-1B at 140 TPP · upstream 5-benchmark average (pretrained)accuracy: 0.61 -> 0.56Hurt
raise pretraining weight decay to 1.0 (ablation)OLMo-2-1B at 20 TPP · attention W_QK rankpseudo-rank: reduces rank of W_QK by roughly a factor of 2Improved
raise pretraining weight decay (ablation)OLMo-2-1B at 20 TPP · train-val gaploss gap: gap decreases monotonically with weight decayImproved
raise pretraining weight decay (ablation)all five model groups · linear probing on SST and AG Newsprobe accuracy: higher at every layerImproved

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