Research map / Architectures and efficiency / Attention and transformers
Transformers are uninterpretable with myopic methods: a case study with bounded Dyck grammars
Kaiyue Wen, Yuchen Li, Bingbin Liu, Andrej Risteski
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 |
|---|---|---|---|
| Fix attention weights to 0, producing uniform attention | Two-layer Transformer · Dyck | accuracy: > 99% accuracy | Improved |
| Remove position embedding | Two-layer Transformer · Dyck₂,₄ validation | accuracy: ≥ 97% accuracy | No clear effect |
| Add linear position embedding | Two-layer Transformer · Prefix [[[[]]]](((()))) | average attention variation: 2.27 -> 2.20 | No clear effect |
| Use a minimal first layer with the specified embedding | Two-layer Transformer · Prefix [[[[]]]](((()))) | average attention variation: 2.20 -> 0.24 | Hurt |
| Add contrastive regularization to reduce balance violation | Transformer trained on Dyck · Dyck prefixes of length 400 to 500 | accuracy | Improved |
| Train with standard training rather than contrastive regularization | Transformer with a minimal first layer · Dyck prefixes of length 400 to 500 | accuracy | Hurt |
Closest papers by meaning
- Beyond Components: Singular Vector-Based Interpretability of Transformer Circuits (NeurIPS 2025 · similarity 0.90)
- Separations in the Representational Capabilities of Transformers and Recurrent Architectures (NeurIPS 2024 · similarity 0.89)
- Algorithmic Capabilities of Random Transformers (NeurIPS 2024 · similarity 0.89)
- Learning Linear Attention in Polynomial Time (NeurIPS 2025 · similarity 0.89)
- How Transformers Learn Structured Data: Insights From Hierarchical Filtering (ICML 2025 · similarity 0.88)
- Transformers need glasses! Information over-squashing in language tasks (NeurIPS 2024 · similarity 0.88)
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.
