INDEX

Explanations

strings related to "AD", such as account names or designations

oai_token-act-pair · gpt-3.5-turbo

references to 'AD' and related acronyms or disorders

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

PlotsExplanationShow Test FieldDefault Test Text

IFrame

Link

Not in Any Lists

No Comments

Negative Logits

 Strip

-0.75

iku

-0.72

oppers

-0.68

 Trop

-0.68

 Berm

-0.67

 Nieto

-0.65

ippi

-0.65

 Squid

-0.63

 grou

-0.63

eday

-0.62

POSITIVE LOGITS

AD

3.71

AD

2.01

ADC

1.80

ADS

1.48

AG

1.36

CE

1.34

AUD

1.33

BC

1.32

AH

1.31

AS

1.29

Activations Density 0.013%

strings related to "AD", such as account names or designations

references to 'AD' and related acronyms or disorders

No Comments

No Known Activations