Research map / Architectures and efficiency / RNNs and normalization
Ordered Memory
Read the paperCheck your idea against 108,967 papers
AI-extracted summary. Written by AI from the paper's title only (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 changed | Where | Measured | Effect |
|---|---|---|---|
| Use Ordered Memory instead of ON-LSTM | Logical inference · Logical inference, 7–12 operations | accuracy: 98 ± 0.0 vs 91; 97 ± 0.4 vs 87; 96 ± 0.5 vs 85; 94 ± 0.8 vs 81; 93 ± 0.5 vs 78; 92 ± 1.1 vs 75 | Helped |
| Use Ordered Memory instead of ON-LSTM | Logical inference · Systematic generalization A, B, and C | accuracy: 94 vs 70; 91 vs 63; 81 vs 60 | Helped |
| Replace the Gated Recursive Cell with the TreeRNN operator (ablation) | Ordered Memory · Logical inference, 7–12 operations | accuracy: 98 ± 0.0 -> 69; 97 ± 0.4 -> 67; 96 ± 0.5 -> 65; 94 ± 0.8 -> 61; 93 ± 0.5 -> 57; 92 ± 1.1 -> 53 | Hurt |
| Use Ordered Memory for ListOps | Ordered Memory · ListOps | accuracy: 99.9% | Helped |
| Use Ordered Memory to induce parses for ListOps | Ordered Memory · ListOps | parsing F1: 100% | Helped |
| Use ELMo embeddings rather than GloVe embeddings | Ordered Memory · SST-2 | accuracy: 90.4 -> 92.0 | Helped |
| Use ELMo embeddings rather than GloVe embeddings | Ordered Memory · SST-5 | accuracy: 52.2 -> 55.2 | Helped |
Closest papers by meaning
- Ordered Neurons: Integrating Tree Structures into Recurrent Neural Networks (ICLR 2019 · similarity 0.89)
- Memory Architectures in Recurrent Neural Network Language Models (ICLR 2018 · similarity 0.88)
- Learning Hierarchical Structures with Differentiable Nondeterministic Stacks (ICLR 2022 · similarity 0.85)
- Enhancing Pointer Network for Sentence Ordering with Pairwise Ordering Predictions (AAAI 2020 · similarity 0.85)
- Artificial Hippocampus Networks for Efficient Long-Context Modeling (ICML 2026 · similarity 0.85)
- A Tensor Decomposition Perspective on Second-order RNNs (ICML 2024 · similarity 0.83)
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.