INDEX

Explanations

mentions of editors-in-chief of various publications

oai_token-act-pair · gpt-3.5-turbo

phrases related to editorial positions or titles

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

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 10-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.10.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.10.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

abouts

-0.81

enegger

-0.73

astic

-0.72

igans

-0.68

imore

-0.63

HT

-0.62

wcs

-0.62

igan

-0.61

TOM

-0.60

asia

-0.60

POSITIVE LOGITS

Chief

1.16

chief

1.08

 chief

1.07

 Chief

1.01

 charge

0.81

Publisher

0.80

Princ

0.79

mbuds

0.78

ashington

0.75

 chiefs

0.74

Activations Density 0.032%

mentions of editors-in-chief of various publications

phrases related to editorial positions or titles

No Comments

No Known Activations