INDEX

Explanations

character names followed by the word "Act"

oai_token-act-pair · gpt-3.5-turbo

references to acts or actions, particularly in a formal or dramatic context

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

xual

-0.71

hower

-0.71

 violet

-0.67

 custody

-0.67

 hypers

-0.65

 constitu

-0.65

agara

-0.65

»Ĵ

-0.64

 grip

-0.63

cemic

-0.63

POSITIVE LOGITS

Act

1.10

ional

1.05

ual

0.97

0.91

wip

0.88

soDeliveryDate

0.87

ership

0.82

ing

0.81

sun

0.80

acebook

0.79

Activations Density 0.010%

character names followed by the word "Act"

references to acts or actions, particularly in a formal or dramatic context

No Comments

No Known Activations