INDEX

Explanations

references to the Australian Labor Party

oai_token-act-pair · gpt-3.5-turbo

references to the Labor Party

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

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 7-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.7.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.7.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

 Hiroshima

-0.76

irs

-0.75

 Allaah

-0.72

eatures

-0.70

ebook

-0.67

imaru

-0.66

oxide

-0.65

Redditor

-0.63

 Cherokee

-0.60

brid

-0.59

POSITIVE LOGITS

aby

0.99

 unions

0.77

acia

0.76

DEM

0.70

acea

0.68

 Labor

0.68

UST

0.68

hurst

0.67

quist

0.67

boro

0.66

Activations Density 0.017%

references to the Australian Labor Party

references to the Labor Party

No Comments

No Known Activations