INDEX

Explanations

phrases related to actions or events which involve forcing someone to do something

oai_token-act-pair · gpt-3.5-turbo

phrases indicating coercion or compulsion

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

 Used

-0.73

interstitial

-0.72

ergy

-0.69

paren

-0.68

 Loud

-0.68

initialized

-0.66

 Allows

-0.65

FM

-0.65

vice

-0.65

foreseen

-0.64

POSITIVE LOGITS

 endure

1.02

 reconsider

0.97

 undertake

0.94

 confront

0.90

 wait

0.90

 abandon

0.87

 ponder

0.86

 hold

0.86

 wear

0.85

 reckon

0.84

Activations Density 0.099%

phrases related to actions or events which involve forcing someone to do something

phrases indicating coercion or compulsion

No Comments

No Known Activations