INDEX

Explanations

words related to official recommendations or approvals

oai_token-act-pair · gpt-3.5-turbo

mentions of endorsements

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

STEM

-0.71

arters

-0.71

ester

-0.68

 Smithsonian

-0.67

rooms

-0.66

isen

-0.65

 Rapt

-0.64

Pom

-0.64

Hen

-0.64

 Hamp

-0.63

POSITIVE LOGITS

 endorsements

1.57

 endorsement

1.56

 endorse

1.26

 endorsing

1.22

 endors

1.22

 endorsed

1.20

indal

0.87

 loyalty

0.83

eering

0.83

 recommendation

0.80

Activations Density 0.008%

words related to official recommendations or approvals

mentions of endorsements

No Comments

No Known Activations