ðŠī Berwin Gan
Search
Search
Dark mode
Light mode
Explorer
Notes ðïļ
Algorithm ð§Đ
Geometric Folding Algorithms: Linkages, Origami, Polhedra
Machine Learning ðĪ
GPU Mode
FlexOlmo: Open Language Models for Flexible Data Use
Luminal - Search-Based Deep Learning Compilers
Modern Position Encodings in Transformers (RoPE/Yarn and PaTH)
Research
Holding
REOrdering Patches Improves Vision Models
Mechanistic Interpretability
How Attention Sinks Keep Language Models Stable
Optimizer
Muon - An optimizer for hidden layers in neural networks
Distillation Robustifies Unlearning
H-Net
Hierarchical Reasoning Model
Language Models in Plato's Cave
Learning Compositional Models of the World
STP: Self-Play LLM Theorem Provers with Iterative Conjecturing and Proving
Aggregate Voting Rank ðģïļ
Covering Discs and Orthants ð
Lambda Calculus ð§Ū
Home
âŊ
Notes ðïļ
âŊ
Machine Learning ðĪ
âŊ
GPU Mode
âŊ
Modern Position Encodings in Transformers (RoPE/Yarn and PaTH)
Modern Position Encodings in Transformers (RoPE/Yarn and PaTH)
Aug 29, 2025
1 min read
unordered set
â
ordered set
absolute positional coding
sinosodiul coding
rotary position embedding
Graph View