Research map / Architectures and efficiency / Attention and transformers
Attention as Implicit Structural Inference
Ryan Singh, Christopher Buckley
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 |
|---|---|---|---|
| use two-hop Multihop Attention instead of single-layer self-attention | single-layer attention · Multihop toy problem | task performance | Improved |
| use Multihop Attention instead of two-layer self-attention | Multihop Attention · Multihop toy problem | task performance | No clear effect |
| Multihop Attention balances attention across two positions | Multihop Attention · Multihop toy problem | attention weights | Improved |
| replace uniform prior with geometric prior and Beta hyper-prior (Expanding Attention) | self-attention · Expanding toy problem | training curves | No clear effect |
| use Expanding Attention when recency bias (1/p) or feature dimension d is large | self-attention · Expanding toy problem | relative number of operations | Improved |
Closest papers by meaning
- Dissecting the Interplay of Attention Paths in a Statistical Mechanics Theory of Transformers (NeurIPS 2024 · similarity 0.89)
- High-Dimensional Learning Dynamics of Attention-Indexed Models (NeurIPS 2026 · similarity 0.89)
- Attention Mechanisms Perspective: Exploring LLM Processing of Graph-Structured Data (ICML 2025 · similarity 0.88)
- Hierarchical Self-Attention: Generalizing Neural Attention Mechanics to Multi-Scale Problems (NeurIPS 2025 · similarity 0.88)
- Attention as a Hypernetwork (ICLR 2025 · similarity 0.87)
- Bayes optimal learning of attention-indexed models (NeurIPS 2025 · similarity 0.87)
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.
