INDEX

Explanations

names of individuals, likely with associated titles or roles

oai_token-act-pair · gpt-3.5-turbo

proper nouns, particularly names and titles of people in authority

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

IFrame

Link

Not in Any Lists

No Comments

Negative Logits

adoes

-0.73

ials

-0.71

ILLE

-0.70

aceae

-0.70

 butterflies

-0.68

casts

-0.66

 hitters

-0.66

 cartoons

-0.66

.$

-0.65

LESS

-0.65

POSITIVE LOGITS

 chairman

1.44

 deputy

1.31

 Chairman

1.28

 director

1.24

 Chair

1.24

 chair

1.24

 Director

1.22

 spokesperson

1.22

 chief

1.22

 superintendent

1.20

Activations Density 0.184%

names of individuals, likely with associated titles or roles

proper nouns, particularly names and titles of people in authority

No Comments

No Known Activations