INDEX

Explanations

phrases related to schemes and efforts to influence or disrupt events or individuals

oai_token-act-pair · gpt-3.5-turbo

phrases indicating attempts or efforts related to influence or control

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

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 8-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.8.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.8.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

checked

-0.70

avia

-0.68

 Collider

-0.66

 Steps

-0.66

Typ

-0.65

etus

-0.63

 Printed

-0.63

 Chili

-0.62

 Needs

-0.61

Choice

-0.61

POSITIVE LOGITS

 promote

1.32

 deceive

1.29

 discredit

1.28

 maximize

1.27

 intimidate

1.27

 divert

1.26

 capitalize

1.26

 deprive

1.26

 minimize

1.25

 undermine

1.23

Activations Density 0.272%

phrases related to schemes and efforts to influence or disrupt events or individuals

phrases indicating attempts or efforts related to influence or control

No Comments

No Known Activations