INDEX

Explanations

phrases related to highlighting or marking something

oai_token-act-pair · gpt-3.5-turbo

instances of the word "marked" used to indicate notable features or characteristics

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

ews

-0.76

atio

-0.69

usa

-0.66

lahoma

-0.63

estate

-0.63

iances

-0.63

leases

-0.62

ettle

-0.61

ither

-0.61

POSITIVE LOGITS

 marked

3.83

marked

2.30

 marks

1.76

 marking

1.73

 unmarked

1.57

 mark

1.54

 labeled

1.43

 labelled

1.32

marks

1.28

 highlighted

1.24

Activations Density 0.015%

phrases related to highlighting or marking something

instances of the word "marked" used to indicate notable features or characteristics

No Comments

No Known Activations