Research map / Probabilistic and scientific modeling / Molecules and proteins
Transformers learn factored representations
Adam Shai, Loren Amdahl-Culleton, Casper L. Christensen, Henry R Bigelow, Fernando Rosas, Alexander Boyd, Eric A. Alt, Kyle J Ray, Paul M. Riechers
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 |
|---|---|---|---|
| train on next-token prediction (vs initialization) | GPT-2 style transformer · Five independent factors | dimensions for 95% variance: ~100 -> 11 | Improved |
| increase d_model to 480 (ablation) | transformer · Five independent factors | dimensionality | No clear effect |
| add noisy channel (epsilon>0) breaking conditional independence (ablation) | transformer · Five factors with noise | dimensionality over training | Mixed |
| use small epsilon (ablation) | transformer · Noisy-channel factors | dimensionality | No clear effect |
| use large epsilon (ablation) | transformer · Noisy-channel factors | dimensionality | Improved |
| train for longer on independent factors | transformer · Vary-one factor subspaces | sum of per-factor dims vs union: sum falls to 14, union 12 | Improved |
| replace transformer with RNN/LSTM (ablation) | RNN, LSTM · Five independent factors | CEV curve factored-ness | Mixed |
Closest papers by meaning
- Effective Context in Transformers: An Analysis of Fragmentation and Tokenization (NeurIPS 2026 · similarity 0.88)
- Transformers Represent Belief State Geometry in their Residual Stream (NeurIPS 2024 · similarity 0.88)
- Constrained Belief Updates Explain Geometric Structures in Transformer Representations (ICML 2025 · similarity 0.87)
- How do Transformers Perform In-Context Autoregressive Learning ? (ICML 2024 · similarity 0.86)
- Transformers Converge to Invariant Algorithmic Cores (NeurIPS 2026 · similarity 0.86)
- How Do Transformers Learn In-Context Beyond Simple Functions? A Case Study on Learning with Representations (ICLR 2024 · 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.
