Research map / Architectures and efficiency / Attention and transformers
Two (narrow) heads are better than (an arbitrarily wide) one
Amanuel Tesfaye, Zeno Kujawa, Rajmohan Rajaraman, Ravi Sundaram
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 |
|---|---|---|---|
| add a second attention head (ablation) | 1-head attention-only transformer · ESP on complete digraph | accuracy: 75% -> 100% for 2-head | Improved |
| add a second attention head (ablation) | 1-head attention-only transformer · ESP on graphs with |MFAS| > 1 | accuracy | Improved |
| add a bounded-width FFN to a single head (ablation) | 1-head transformer · ESP on complete digraph | accuracy | Improved |
| increase embedding dimension of single head (d >> n) (ablation) | 1-head attention-only transformer · ESP on complete digraph | accuracy | No clear effect |
Closest papers by meaning
- Transformers Struggle to Learn to Search (ICLR 2025 · similarity 0.90)
- Towards Principled Graph Transformers (NeurIPS 2024 · similarity 0.90)
- Transformers need glasses! Information over-squashing in language tasks (NeurIPS 2024 · similarity 0.89)
- Separations in the Representational Capabilities of Transformers and Recurrent Architectures (NeurIPS 2024 · similarity 0.89)
- Understanding Transformer Reasoning Capabilities via Graph Algorithms (NeurIPS 2024 · similarity 0.89)
- Unveiling Induction Heads: Provable Training Dynamics and Feature Learning in Transformers (NeurIPS 2024 · similarity 0.89)
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.
