AI Paper Atlas

Research map / Architectures and efficiency / RNNs and normalization

PRISM: Parallel Residual Iterative Sequence Model

Jie Jiang, Ke Cheng, Xin Xu, Mengyang Pang, Tianhao Lu, Jiaheng Li, Yue Liu, Yuan Wang, Jun Zhang, Huan Yu, Zhouchen Lin

ICML 2026 · Architectures and efficiency › RNNs and normalization · approach: Distributed LLM inference systems · 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 iterative refinement (L>0) to the injection operator (ablation)PRISM · Amazon ElectronicsAUC: 0.6805 -> 0.7134Improved
add iterative refinement (L>0) to the injection operator (ablation)PRISM · Amazon ElectronicsHit@200: 0.1155 -> 0.1409Improved
add GELU non-linearity in the refinement loop (ablation)PRISM · Amazon ElectronicsAUC: 0.7047 -> 0.7134Improved
add ShortConv anchor (ablation)PRISM · Amazon ElectronicsAUC: 0.7076 -> 0.7134Improved
add ShortConv anchor (ablation)PRISM · Amazon ElectronicsHit@200: 0.1406 -> 0.1409No clear effect
add gain predictor (ablation)PRISM · Amazon ElectronicsAUC: 0.7098 -> 0.7134Improved
add gain predictor (ablation)PRISM · Amazon ElectronicsHit@500: 0.2406 -> 0.2613Improved
replace serial TTT iteration with PRISM input-anchored parallel unrollingTTT · Training throughput on H20 GPUthroughput: 0.34K tokens/s -> 57-61K tokens/sImproved
use Rank-L parallel PRISM instead of explicit serial TTTTTT · Recommendation benchmarksMean Rank: 5.25 -> 2.62Improved
use PRISM instead of Gated DeltaNet (Rank-1)Gated DeltaNet · Recommendation benchmarksMean Rank: 5.12 -> 2.62Improved
Rank-L PRISM injection with L refinement projectionsPRISM · Training throughputthroughput overhead: <16% overhead vs Rank-1 baselinesHurt
use iterative non-linear PRISM instead of linear multi-expert MoMMoM / LA · XOR and Parity probing (D=16)accuracy: 0.49-0.50 -> 1.00Improved

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