AI Paper Atlas

Research map / Architectures and efficiency / RNNs and normalization

L-SR1: Learned Symmetric-Rank-One Preconditioning

Gal Lifshitz, Shahar Zuler, Ori Fouks, Dan Raviv

ICML 2026 · Architectures and efficiency › RNNs and normalization · approach: Differentiable optimization layers · 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 secant penalty (PGSM) during meta-training (ablation)L-SR1 · 3DPW testPA-MPJPE (mm): 52.80 -> 51.74Improved
replace LGD with L-SR1 at 4 stepslearned optimization HMR · 3DPW testPA-MPJPE (mm): 55.90 -> 51.74Improved
use optimal-horizon steps instead of 4 stepsL-SR1 · 3DPW testPA-MPJPE (mm): 51.74 -> 51.58 (364 ms -> 1183 ms)Improved
use element-wise instead of scalar learning rate (ablation)L-SR1 · 3DPWPA-MPJPE (mm): 51.80 -> 51.58Improved
use element-wise instead of scalar learning rate (ablation)L-SR1 · Quadratic N=10mean inner loss: -0.86 -> 7.33Hurt
add secant penalty during meta-training (ablation)L-SR1 · Quadratic test, buffer sizes 2/4/8test loss over first 20 iterationsImproved
increase hidden dimension from 64 to 128 (ablation)L-SR1 · Quadratic validationvalidation loss: -1.90 -> -2.56Improved
add p and q encoder inputs and d (all five inputs) (ablation)L-SR1 · Quadratic validationvalidation loss: -2.09 -> -2.56Improved
add p and q inputs to x,d,g (ablation)L-SR1 · Quadratic validationvalidation loss: -2.16 -> -2.56Improved
add d input to x,p,g,q (ablation)L-SR1 · Quadratic validationvalidation loss: -2.51 -> -2.56Improved
train on 10% instead of 80% of AMASSL-SR1 · 3DPWPA-MPJPE (mm): 51.67 -> 53.03Hurt
train on 1% instead of 10% of AMASSL-SR1 · 3DPWPA-MPJPE (mm): 53.03 -> 56.14Hurt

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