AI Paper Atlas

Research map / Architectures and efficiency / Pruning and efficient tuning

Weight Updates as Activation Shifts: A Principled Framework for Steering

Dyah Adila, John Cooper, Alexander Yun, Avi Trost, Frederic Sala

ICML 2026 · Architectures and efficiency › Pruning and efficient tuning · approach: Low-rank compression and merging · 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
add orthogonality constraint to joint LoRA+adapter training (ablation)joint weight+activation adaptation · GSM8K (Gemma-3-1b)accuracy: 22.8 -> 24.4Improved
add orthogonality constraint to joint training (ablation)joint adaptation · GSM8K (Qwen 3 4B)accuracy: 37.9 -> 38.8Improved
add orthogonality constraint to joint training (ablation)joint adaptation · Winogrande (Llama-3.2-1B)accuracy: 56.9 -> 59.3Improved
add orthogonality constraint to joint training (ablation)joint adaptation · Winogrande (Gemma-3-1b)accuracy: 51.3 -> 50.9No clear effect
add orthogonality constraint to joint training (ablation)joint adaptation · GSM8K (Llama-3.2-1B)accuracy: 31.1 -> 31.1No clear effect
use post-block steering instead of ReFT at same 0.04% budgetpost-block steering · ListOps (Llama-3.2-1B)accuracy: 49.3 -> 65.4Improved
use post-block steering instead of ReFTsteering adapter · Winogrande (Qwen 3 4B)accuracy: 64.9 -> 80.6Improved
use post-block steering instead of LoRA with 11x fewer parameterssteering adapter · Six-task average (Llama-3.2-1B)average accuracy difference vs SFT: -2.2 -> -0.2Improved
use nonlinear instead of linear post-block adapter (r=16) (ablation)post-block steering · AlpacaEval 2.0 (Llama-3.1-8B)LC win rate: 11.00 -> 11.34Improved
use nonlinear instead of linear adapter (ablation)post-block steering · GSM8K RL (R1-Distill-Qwen-1.5B)Pass@1: 84.3 -> 84.7No clear effect
use post-block steering instead of LoRA in GRPO RLpost-block steering · GSM8K (R1-Distill-Qwen-1.5B)Pass@1: 81.5 -> 84.7Improved
use post-block steering instead of LoRA r=8post-block steering · AlpacaEval 2.0LC win rate: 10.52 -> 11.00Improved

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