INDEX

Explanations

phrases related to legal or political inquiries

oai_token-act-pair · gpt-3.5-turbo

phrases that indicate ongoing or repeated action

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

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 5-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.5.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.5.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

eem

-0.63

 behold

-0.60

bart

-0.58

 photograp

-0.56

 observing

-0.56

 prosecuting

-0.55

ugu

-0.55

marriage

-0.55

bats

-0.55

 recall

-0.55

POSITIVE LOGITS

 been

1.39

 undergone

1.15

 arisen

1.09

 become

1.08

 begun

1.05

been

1.00

 risen

1.00

 gone

0.99

 gotten

0.98

hed

0.93

Activations Density 0.285%

phrases related to legal or political inquiries

phrases that indicate ongoing or repeated action

No Comments

No Known Activations