9 citations · 17 across the 7 of their papers we have counts for
7 papers
Vector Arithmetic in Concept and Token Subspaces
Sheridan Feucht, Byron Wallace, David Bau
In order to predict the next token, LLMs must represent semantic and surface-level information about the current word. Previous work identified two types of attention heads that di…
The Dual-Route Model of Induction
Sheridan Feucht, Eric Todd, Byron Wallace +1
Prior work on in-context copying has shown the existence of induction heads, which attend to and promote individual tokens during copying. In this work we discover a new type of in…
Elucidating Mechanisms of Demographic Bias in LLMs for Healthcare
Hiba Ahsan, Arnab Sen Sharma, Silvio Amir +2
We know from prior work that LLMs encode social biases, and that this manifests in clinical tasks. In this work we adopt tools from mechanistic interpretability to unveil sociodemo…
NNsight and NDIF: Democratizing Access to Open-Weight Foundation Model Internals
Jaden Fiotto-Kaufman, Alexander R. Loftus, Eric Todd +17
We introduce NNsight and NDIF, technologies that work in tandem to enable scientific study of the representations and computations learned by very large neural networks. NNsight is…
Token Erasure as a Footprint of Implicit Vocabulary Items in LLMs
Sheridan Feucht, David Atkinson, Byron Wallace +1
LLMs process text as sequences of tokens that roughly correspond to words, where less common words are represented by multiple tokens. However, individual tokens are often semantic…
Future Lens: Anticipating Subsequent Tokens from a Single Hidden State
Koyena Pal, Jiuding Sun, Andrew Yuan +2
We conjecture that hidden state vectors corresponding to individual input tokens encode information sufficient to accurately predict several tokens ahead. More concretely, in this…
Function Vectors in Large Language Models
Eric Todd, Millicent L. Li, Arnab Sen Sharma +3
We report the presence of a simple neural mechanism that represents an input-output function as a vector within autoregressive transformer language models (LMs). Using causal media…