INDEX

Explanations

proper nouns or names, potentially of individuals

oai_token-act-pair · gpt-3.5-turbo

the letter 'W'

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

psy

-0.71

EVE

-0.66

 consensual

-0.66

 satisfied

-0.66

 unaff

-0.65

 fewer

-0.65

 earth

-0.64

 reputable

-0.63

 conscientious

-0.62

 scraping

-0.61

POSITIVE LOGITS

iesel

1.35

oj

1.29

irth

1.28

ahl

1.25

essel

1.25

asser

1.24

ohl

1.24

augh

1.21

ither

1.19

orthy

1.19

Activations Density 0.029%

proper nouns or names, potentially of individuals

the letter 'W'

No Comments

No Known Activations