INDEX

Explanations

proper names of individuals, particularly related to the field of security or sports

oai_token-act-pair · gpt-3.5-turbo

recurring names of notable figures

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

Hung

-0.92

quit

-0.87

gun

-0.76

Gun

-0.71

cs

-0.70

Nun

-0.70

gun

-0.69

 respir

-0.69

 Meteor

-0.68

 triglycer

-0.68

POSITIVE LOGITS

 Woodward

3.56

 Rowling

1.77

Wood

1.63

 Rove

1.60

 Wynne

1.44

 Klopp

1.27

 Steele

1.16

 Strauss

1.15

Tuc

1.14

 Ballard

1.12

Activations Density 0.049%

proper names of individuals, particularly related to the field of security or sports

recurring names of notable figures

No Comments

No Known Activations