INDEX

Explanations

phrases indicating events or promotions featuring a special guest or performer

oai_token-act-pair · gpt-3.5-turbo

instances of the word "featuring" indicating a highlight or presentation of content

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

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 1-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.1.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.1.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

sis

-0.86

ochond

-0.82

umeric

-0.78

sett

-0.77

externalActionCode

-0.75

uters

-0.74

hood

-0.72

athy

-0.71

thritis

-0.70

 Sabha

-0.69

POSITIVE LOGITS

 prominently

1.11

 featuring

0.96

Features

0.89

 Features

0.88

 Featuring

0.82

 features

0.79

 Emin

0.77

 starring

0.75

endas

0.75

features

0.75

Activations Density 0.014%

phrases indicating events or promotions featuring a special guest or performer

instances of the word "featuring" indicating a highlight or presentation of content

No Comments

No Known Activations