Research map / Architectures and efficiency / Attention and transformers
Attention Approximates Sparse Distributed Memory
Trenton Bricken, Cengiz Pehlevan
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 changed | Where | Measured | Effect |
|---|---|---|---|
| L2 normalize value vectors in GPT2 training (ablation) | GPT2 · language modeling | performance | No clear effect |
| Query-Key Normalization (L2 normalized queries and keys, learned beta) versus standard Attention | Transformer · five small translation datasets | performance | Improved |
| Use exact circle intersection instead of continuous approximation | SDM · n=1000, r=1,000,000, d=451 | expected neurons in intersection: 1000 -> 1071 | Improved |
Closest papers by meaning
- Transformers Learn Faster with Semantic Focus (NeurIPS 2025 · similarity 0.89)
- Associative Transformer (CVPR 2025 · similarity 0.89)
- Delta Attention: Fast and Accurate Sparse Attention Inference by Delta Correction (NeurIPS 2025 · similarity 0.87)
- The emergence of sparse attention: impact of data distribution and benefits of repetition (NeurIPS 2025 · similarity 0.87)
- Memorization Capacity of Multi-Head Attention in Transformers (ICLR 2024 · similarity 0.86)
- Neural Attention Search (NeurIPS 2025 · similarity 0.86)
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.
