INDEX

Explanations

phrases related to involvement with power or authority

oai_token-act-pair · gpt-3.5-turbo

the word "with" and phrases indicating associations or relationships

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

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 4-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.4.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.4.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

soft

-0.63

mAh

-0.59

obook

-0.58

press

-0.58

pad

-0.58

lish

-0.55

Angelo

-0.55

reader

-0.55

hood

-0.54

book

-0.54

POSITIVE LOGITS

 regard

1.46

 regards

1.38

stood

1.36

 respect

1.16

draw

1.13

drawn

1.06

 impunity

1.04

standing

1.00

holding

0.88

 dignity

0.84

Activations Density 0.309%

phrases related to involvement with power or authority

the word "with" and phrases indicating associations or relationships

No Comments

No Known Activations