Search

Xinting Huang

Discovering Interpretable Algorithms by Decompiling Transformers to RASP
Decomposing Representation Space into Interpretable Subspaces with Unsupervised Learning
Contextualize-then-Aggregate: Circuits for In-Context Learning in Gemma-2 2B
Lower Bounds for Chain-of-Thought Reasoning in Hard-Attention Transformers
A Formal Framework for Understanding Length Generalization in Transformers
InversionView: A General-Purpose Method for Reading Information from Neural Activations

Published with Wowchemy — the free, open source website builder that empowers creators.