AI Paper Atlas

Research map / Architectures and efficiency / Attention and transformers

FutureFill: Fast Generation from Convolutional Sequence Models

Naman Agarwal, Xinyi Chen, Evan Dogariu, Devan Shah, Hubert Strauss, Vladimir Feinberg, Daniel Suo, Peter Bartlett, Elad Hazan

ICLR 2026 · Architectures and efficiency › Attention and transformers · approach: Language model pretraining and induction · 5 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 Epoched-FutureFill instead of naive online convolution for generation without prefill at 126976 tokensFlashSTU-T STU-only · Generation time, no prefillspeedup: 1.7× speedupImproved
use Epoched-FutureFill instead of naive online convolution for generation without prefill at 126976 tokensFlashSTU-T hybrid · Generation time, no prefillspeedup: 1.5× speedupImproved
use Epoched-FutureFill with a long prefill (32768) instead of naive cached convolutionFlashSTU-T STU-only · Generation time with prefillinference time (s): 165.13 -> 75.66Improved
use Epoched-FutureFill with a short prompt (512) and total length 4096FlashSTU-T STU-only 670.75M · Generation time, prompt 512inference time (s): 12.68 -> 16.33Hurt
use Epoched-FutureFill with a short prompt (512) and total length 131072FlashSTU-T STU-only 670.75M · Generation time, prompt 512inference time (s): 1837.16 -> 1097.50Improved
set epoch length K to theoretical optimum 1024 instead of smaller K at L_gen 65536 (ablation)Epoched-FutureFill · Generation time vs cache size Kinference timeImproved

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