INDEX

Explanations

dates written in numeric format with a comma and year at the end

oai_token-act-pair · gpt-3.5-turbo

punctuation marks, particularly periods at the end of statements

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

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 11-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.11.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.11.hook_resid_pre

Architecture

standard

Context Size

128

Dataset

Skylion007/openwebtext

Hook Point Layer

Activation Function

relu

Embeds

IFrame

Link

Not in Any Lists

No Comments

Negative Logits

 downright

-0.93

 popping

-0.82

 tack

-0.80

lege

-0.79

 pesky

-0.79

 stomp

-0.76

 dangling

-0.76

 tempting

-0.75

 tempt

-0.74

 stubborn

-0.74

POSITIVE LOGITS

Its

1.37

 Originally

1.36

Contents

1.34

 Afterwards

1.27

 Shortly

1.27

 Later

1.25

 During

1.25

 Approximately

1.23

 Initially

1.21

 Following

1.21

Activations Density 0.303%

dates written in numeric format with a comma and year at the end

punctuation marks, particularly periods at the end of statements

No Comments

No Known Activations