INDEX

Explanations

words related to competitions or events

oai_token-act-pair · gpt-3.5-turbo

references to rankings or ordinal positions

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

IFrame

Link

Not in Any Lists

No Comments

Negative Logits

 Wolves

-0.68

âĻ¥

-0.60

 books

-0.60

 dismiss

-0.58

che

-0.58

________________________

-0.57

bas

-0.57

 wiki

-0.57

 walk

-0.56

 Sparkle

-0.56

POSITIVE LOGITS

inal

4.74

inals

2.79

INAL

2.31

inally

2.21

inous

1.50

ina

1.36

inary

1.32

inate

1.29

ination

1.28

inas

1.27

Activations Density 0.003%

words related to competitions or events

references to rankings or ordinal positions

No Comments

No Known Activations