INDEX

Explanations

mentions of the politician Nancy Pelosi

oai_token-act-pair · gpt-3.5-turbo

mentions of the name "Nancy" and related titles

oai_token-act-pair · gpt-4o-mini Triggered by @bot

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 9-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.9.hook_resid_pre

Prompts (Dashboard)

24,576 prompts, 128 tokens each

Dataset (Dashboard)

Skylion007/openwebtext

Features

24,576

Data Type

torch.float32

Hook Point

blocks.9.hook_resid_pre

Architecture

standard

Context Size

128

Dataset

Skylion007/openwebtext

Hook Point Layer

Activation Function

relu

Embeds

IFrame

Link

Not in Any Lists

No Comments

Negative Logits

=-=-

-0.77

 guiActiveUnfocused

-0.73

Redditor

-0.72

 eleph

-0.71

=-=-=-=-

-0.71

abol

-0.69

ombo

-0.68

angers

-0.68

orically

-0.67

ocene

-0.67

POSITIVE LOGITS

 Pelosi

1.39

 Reagan

0.95

 Drew

0.90

 Grace

0.90

vier

0.89

 Gibbs

0.84

Lan

0.80

 Lieberman

0.78

 Schultz

0.76

Ker

0.75

Activations Density 0.030%

mentions of the politician Nancy Pelosi

mentions of the name "Nancy" and related titles

No Comments

No Known Activations