INDEX

Explanations

names of authors or creators

oai_token-act-pair · gpt-3.5-turbo

parentheses in the text

oai_token-act-pair · gpt-4o-mini Triggered by @bot

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 10-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.10.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.10.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

 consist

-0.80

 indul

-0.79

 normalized

-0.78

 everyday

-0.77

 increment

-0.76

 conformity

-0.76

 entitle

-0.76

met

-0.76

 treat

-0.75

 equival

-0.75

POSITIVE LOGITS

who

1.54

pictured

1.41

formerly

1.40

pron

1.36

whose

1.33

aka

1.31

Assistant

1.30

CEO

1.22

sic

1.22

University

1.21

Activations Density 0.069%

names of authors or creators

parentheses in the text

No Comments

No Known Activations