AI Paper Atlas

Research map / Architectures and efficiency / Attention and transformers

Test-Time Training Done Right

Tianyuan Zhang, Sai Bi, Yicong Hong, Kai Zhang, Fujun Luan, Songlin Yang, Kalyan Sunkavalli, William T. Freeman, Hao Tan

ICLR 2026 · Architectures and efficiency › Attention and transformers · approach: State space and recurrent sequence models · 129 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
use large-chunk TTT instead of full attention for prefillLaCT · NVS prefill 48 viewsprefill time: 16.1 s -> 1.4 sImproved
scale up fast-weight state size (ablation)LaCT · novel view synthesis and language modelingPSNR / lossImproved
use Muon test-time optimizer instead of vanilla GD or momentum (ablation)LaCT · NVS and language modelingPSNR / lossImproved
use Muon update (vs GD) (ablation)LaCT · 3B language model training throughputtokens/s: 5.0K -> 4.3KHurt
use nonlinear SwiGLU fast weights instead of linear (ablation)LaCT · NVS and language modelingPSNR / lossImproved
use large-chunk linear recurrence instead of per-token Mamba-2 with same state size (ablation)LaCT linear · novel view synthesisPSNRImproved
use linear large-chunk recurrence instead of GLA SWA with same state size (ablation)LaCT linear · language modelinglossHurt
combine large nonlinear state with Muon (ablation)LaCT · language modelingper-position loss and retrieval accuracyImproved
use LaCT instead of sliding window attentionLaCT · autoregressive video diffusiondenoising lossImproved
use LaCT instead of 3D Gaussian Splatting under sparse viewsLaCT · DL3DV scenesPSNRImproved

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