Similar Items: Beyond Perplexity: A Geometric and Spectral Study of Low-Rank Pre-Training
- Efficient Pre-Training with Token Superposition
- Long Context Pre-Training with Lighthouse Attention
- Geometric Factual Recall in Transformers
- MatryoshkaLoRA: Learning Accurate Hierarchical Low-Rank Representations for LLM Fine-Tuning
- Self-Attention as Transport: Limits of Symmetric Spectral Diagnostics
- Fuzzy Fingerprinting Encoder Pre-trained Language Models for Emotion Recognition in Conversations: Human Assessment and Validity Study