AI Paper Atlas

Research map / Architectures and efficiency / RNNs and normalization

Recursion in Recursion: Two-Level Nested Recursion for Length Generalization with Scalability

Jishnu Ray Chowdhury, Cornelia Caragea

NeurIPS 2023 · Architectures and efficiency › RNNs and normalization · approach: Efficient network compression and training · 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
Use RIR-EBT-GRC rather than the simpler balanced-tree modelBBT-GRC · ListOps-O length gen. 900-1000accuracy: 31.5 -> 97.1Improved
Add the S4D pre-chunk processing layer (ablation)RIR-EBT-GRC · ListOps-O argument gen. 100-1000, 10 argumentsaccuracy: 63.95 -> 74.9Improved
Add beam alignment instead of concatenating beams by rank (ablation)RIR-EBT-GRC · ListOps-O length gen. 900-1000accuracy: 68.8 -> 97.1Improved
Use score-weighted beam alignment rather than uniform random alignment (ablation)RIR-EBT-GRC · ListOps-O length gen. 900-1000accuracy: 86.4 -> 97.1Improved
Keep the inner recursion unconstrained by RIR at inference (ablation)RIR-EBT-GRC · ListOps-O length gen. 900-1000accuracy: 36.9 -> 97.1Improved
Use beam size 7 rather than beam size 5 (ablation)RIR-EBT-GRC · ListOps-O length gen. 900-1000accuracy: 89.0 -> 97.1Improved
Use chunk size 30 rather than chunk size 20 (ablation)RIR-EBT-GRC · ListOps-O length gen. 900-1000accuracy: 63.7 -> 97.1Improved
Use chunk size 30 rather than chunk size 10 (ablation)RIR-EBT-GRC · ListOps-O length gen. 900-1000accuracy: 39.9 -> 97.1Improved
Use EBT-GRC rather than CRvNN as the inner recursion (ablation)RIR · ListOps-O length gen. 900-1000accuracy: 28.7 -> 97.1Improved
Use EBT-GRC rather than Ordered Memory as the inner recursion (ablation)RIR · ListOps-O length gen. 900-1000accuracy: 29.7 -> 97.1Improved
Remove S4D pre-chunk processing (ablation)RIR-EBT-GRC · LRA Textaccuracy: 87.97_{1.6} -> 88.13_{1.3}No clear effect
Use RIR inference rather than unconstrained inner-recursion inference (ablation)RIR-EBT-GRC · LRA TextaccuracyNo clear effect

Earlier methods this paper compares with

Named baselines from earlier accepted papers in this paper's results tables: better than 0 of 1 on most shared results.

Matched by method name in results tables read by AI. Authors' own tables tend to favour the newer method, and a comparison can be missed when a method is renamed.

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