Similar Items: Transformers with Selective Access to Early Representations
- Superposition Is Not Necessary: A Mechanistic Interpretability Analysis of Transformer Representations for Time Series Forecasting
- Aitchison Embeddings for Learning Compositional Graph Representations
- PHALAR: Phasors for Learned Musical Audio Representations
- A Unified Framework of Hyperbolic Graph Representation Learning Methods
- Pretrained Model Representations as Acquisition Signals for Active Learning of MLIPs
- Manifold Steering Reveals the Shared Geometry of Neural Network Representation and Behavior