INDEX

Explanations

phrases related to negative emotions or actions

oai_token-act-pair · gpt-3.5-turbo

the use of the word "per," likely in contexts related to measurements or distributions

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

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 10-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.10.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.10.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

ForgeModLoader

-0.74

 Raleigh

-0.68

 GOODMAN

-0.63

 Fargo

-0.62

ä¹ĭ

-0.60

 shortened

-0.60

 ONLY

-0.56

çīĪ

-0.56

 Plymouth

-0.56

ALLY

-0.56

POSITIVE LOGITS

malink

1.47

missive

1.39

ishable

1.35

tains

1.30

 capita

1.28

ceptive

1.22

versely

1.21

chance

1.18

verted

1.15

imeters

1.14

Activations Density 0.018%

phrases related to negative emotions or actions

the use of the word "per," likely in contexts related to measurements or distributions

No Comments

No Known Activations