Research map / Architectures and efficiency / Attention and transformers
On the Dynamics of Training Attention Models
Haoye Lu, Yongyi Mao, Amiya Nayak
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 the full attention model instead of freezing topic-word scores (ablation) | attnTC · Synthetic training | training loss decrease speed | Improved |
| Use the full attention model instead of fixing topic-word embeddings (ablation) | attnTC · Synthetic training | training loss decrease speed | Improved |
| Use a lower-capacity classifier | attention models · Synthetic dataset | topic-word score and embedding norm | Improved |
| Train with a trainable query rather than a fixed query | attention models · Synthetic dataset | positive score–embedding norm relationship | No clear effect |
| Add parallel CNN preprocessing for embeddings and keys | attnTC · Synthetic word-pair task | positive score–embedding norm relationship | No clear effect |
Closest papers by meaning
- Training Dynamics of Transformers to Recognize Word Co-occurrence via Gradient Flow Analysis (NeurIPS 2024 · similarity 0.88)
- Attention with Trained Embeddings Provably Selects Important Tokens (NeurIPS 2025 · similarity 0.88)
- Training Dynamics of In-Context Learning in Linear Attention (ICML 2025 · similarity 0.88)
- When Attention Sink Emerges in Language Models: An Empirical View (ICLR 2025 · similarity 0.87)
- A Phase Transition between Positional and Semantic Learning in a Solvable Model of Dot-Product Attention (NeurIPS 2024 · similarity 0.87)
- Asymptotics of SGD in Sequence-Single Index Models and Single-Layer Attention Networks (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.
