Research map / Architectures and efficiency / RNNs and normalization
Encoding word order in complex embeddings
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 complex-order embeddings instead of the strongest listed CNN position-embedding baseline | CNN text classifier · MR | accuracy: 0.816 -> 0.825 | Helped |
| use complex-order embeddings instead of the strongest listed CNN position-embedding baseline | CNN text classifier · SUBJ | accuracy: 0.938 -> 0.951 | Helped |
| use complex-order embeddings instead of the strongest listed CNN position-embedding baseline | CNN text classifier · CR | accuracy: 0.836 -> 0.852 | Helped |
| use complex-order embeddings instead of the strongest listed CNN position-embedding baseline | CNN text classifier · MPQA | accuracy: 0.897 -> 0.906 | Helped |
| use complex-order embeddings instead of the strongest listed CNN position-embedding baseline | CNN text classifier · SST | accuracy: 0.856 -> 0.864 | Helped |
| use complex-order embeddings instead of the strongest listed CNN position-embedding baseline | CNN text classifier · TREC | accuracy: 0.918 -> 0.939 | Helped |
| use complex-order embeddings instead of Transformer embeddings | Transformer machine translation · WMT 2016 English-German | BLEU: +1.3 BLEU | Helped |
| use complex-order embeddings instead of complex-vanilla embeddings | Transformer machine translation · WMT 2016 English-German | BLEU: +1.1 BLEU | Helped |
| use complex-order embeddings instead of the 6-layer Transformer XL counterpart | Transformer XL language model · text8 test | BPC: 1.29 -> 1.26 | Helped |
| add initial phases (ablation) | Transformer-complex-order · SST | accuracy: 0.813 -> 0.785; -0.028 | Hurt |
| share periods across dimensions (ablation) | Transformer-complex-order · SST | accuracy: 0.813 -> 0.797; -0.016 | Hurt |
| share periods across words (ablation) | Transformer-complex-order · SST | accuracy: 0.813 -> 0.805; -0.008 | Hurt |
Closest papers by meaning
- On Position Embeddings in BERT (ICLR 2021 · similarity 0.88)
- Rethinking Positional Encoding in Language Pre-training (ICLR 2021 · similarity 0.88)
- Beyond Real: Imaginary Extension of Rotary Position Embeddings for Long-Context LLMs (ICLR 2026 · similarity 0.86)
- Rethinking Addressing in Language Models via Contextualized Equivariant Positional Encoding (ICML 2025 · similarity 0.86)
- Eliminating Position Bias of Language Models: A Mechanistic Approach (ICLR 2025 · similarity 0.85)
- By Tying Embeddings You Are Assuming the Distributional Hypothesis (ICML 2024 · similarity 0.84)
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.