INDEX

Explanations

phrases related to being blacklisted or listed

oai_token-act-pair · gpt-3.5-turbo

instances of being blacklisted and specific geographical terms

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

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 0-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.0.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.0.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

unia

-0.77

asa

-0.75

 Bigfoot

-0.67

bip

-0.67

 Mania

-0.67

sid

-0.66

 coax

-0.66

ern

-0.64

 constitu

-0.64

azo

-0.63

POSITIVE LOGITS

listed

2.14

lake

1.28

naire

1.22

ursed

1.21

 Material

1.20

 Pebble

1.13

kar

1.12

 Furn

1.10

Aj

1.03

carbon

1.03

Activations Density 0.048%

phrases related to being blacklisted or listed

instances of being blacklisted and specific geographical terms

No Comments

No Known Activations