INDEX

Explanations

instances of numbers or quantity-related terms being emphasized in a document

oai_token-act-pair · gpt-3.5-turbo

the phrase "nearly" followed by numerical approximations or quantities

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

soType

-0.78

agate

-0.77

kat

-0.71

ivo

-0.70

oslav

-0.67

Admin

-0.65

enter

-0.64

iw

-0.63

 Therefore

-0.63

 Instructor

-0.62

POSITIVE LOGITS

 nearly

2.92

 almost

1.95

 virtually

1.71

 roughly

1.70

almost

1.65

 practically

1.60

 Nearly

1.57

Nearly

1.56

 approximately

1.32

 barely

1.31

Activations Density 0.017%

instances of numbers or quantity-related terms being emphasized in a document

the phrase "nearly" followed by numerical approximations or quantities

No Comments

No Known Activations