INDEX

Explanations

references to various types of "ills" or illnesses

oai_token-act-pair · gpt-4o-mini

Triggered by @bot

words related to bills or medical conditions

oai_token-act-pair · gpt-3.5-turbo

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

IFrame

Link

Negative Logits

TAG

-0.85

oÄŁ

-0.66

 IEEE

-0.65

bal

-0.63

 affili

-0.62

 Malays

-0.62

 prov

-0.61

members

-0.60

pai

-0.59

Barn

-0.59

POSITIVE LOGITS

sburg

1.07

iard

1.01

creen

0.96

hops

0.93

iflower

0.91

chool

0.89

inger

0.87

hift

0.86

hooting

0.85

ibrary

0.85

Activations Density 0.009%

references to various types of "ills" or illnesses

words related to bills or medical conditions

No Known Activations