INDEX

Explanations

instances of email-related words or terms

oai_token-act-pair · gpt-3.5-turbo

occurrences of the "e-" prefix in words

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

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 7-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.7.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.7.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

AVG

-0.73

 Cassidy

-0.72

Ô

-0.71

eele

-0.70

ulhu

-0.66

 OPEC

-0.64

uania

-0.63

Zup

-0.61

 Loll

-0.61

 Witches

-0.61

POSITIVE LOGITS

mail

1.29

mails

1.28

commerce

1.26

cigarette

1.12

expression

1.04

ventures

1.00

government

0.98

communication

0.97

collar

0.97

Commerce

0.93

Activations Density 0.025%

instances of email-related words or terms

occurrences of the "e-" prefix in words

No Comments

No Known Activations