INDEX

Explanations

phrases related to locations, specifically the Detroit-Windsor area

oai_token-act-pair · gpt-3.5-turbo

references to specific geographical locations, particularly Windsor and Essex

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

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 1-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.1.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.1.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

ramid

-1.01

ynthesis

-0.79

ļé

-0.77

mble

-0.75

alg

-0.75

ngth

-0.74

undai

-0.73

irty

-0.72

artments

-0.69

alach

-0.68

POSITIVE LOGITS

 Windsor

1.36

 Spit

1.23

 Bride

0.81

hurst

0.80

 Castle

0.80

mouth

0.75

shire

0.75

 Downs

0.74

 Loft

0.73

borough

0.72

Activations Density 0.004%

phrases related to locations, specifically the Detroit-Windsor area

references to specific geographical locations, particularly Windsor and Essex

No Comments

No Known Activations