INDEX

Explanations

references to editing or modifications in online content

oai_token-act-pair · gpt-3.5-turbo

instances of the word "Edit" 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 @ 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

 baskets

-0.70

 heck

-0.70

 clutch

-0.67

 buckets

-0.65

 basket

-0.63

 hoops

-0.62

 rightfully

-0.62

ibaba

-0.58

 thunder

-0.58

Harris

-0.57

POSITIVE LOGITS

 Edit

4.06

edit

2.05

 edit

1.88

Edit

1.81

".[

1.38

).[

1.28

.[

1.26

:[

1.24

Contents

1.22

Â¶

1.18

Activations Density 0.018%

references to editing or modifications in online content

instances of the word "Edit" in the text

No Comments

No Known Activations