INDEX

Explanations

references to the New Yorker magazine

oai_token-act-pair · gpt-3.5-turbo

references to "The New Yorker" publication

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

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 3-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.3.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.3.hook_resid_pre

Architecture

standard

Context Size

128

Dataset

Skylion007/openwebtext

Hook Point Layer

Activation Function

relu

Embeds

PlotsExplanationShow Test FieldDefault Test Text

IFrame

Link

Not in Any Lists

No Comments

Negative Logits

smoking

-0.70

rams

-0.68

leading

-0.67

oney

-0.65

ETA

-0.64

 bell

-0.63

owing

-0.63

lower

-0.62

inant

-0.62

ground

-0.62

POSITIVE LOGITS

vier

0.94

 Yorker

0.94

 Mellon

0.84

 Magazine

0.77

 magazine

0.75

 Travels

0.75

 Wire

0.73

 Jacket

0.72

opol

0.71

 Juda

0.70

Activations Density 0.011%

references to the New Yorker magazine

references to "The New Yorker" publication

No Comments

No Known Activations