INDEX

Explanations

personal references pointing to the speaker

oai_token-act-pair · gpt-3.5-turbo

references to the speaker or self

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

 Equality

-0.63

profits

-0.60

edged

-0.60

 Confederation

-0.58

 Governments

-0.58

earable

-0.57

ipedia

-0.57

 Coalition

-0.57

Args

-0.56

Us

-0.55

POSITIVE LOGITS

lees

1.17

zzo

1.11

 personally

1.09

adows

1.08

adow

1.03

andering

0.97

imei

0.95

cca

0.89

 myself

0.88

gal

0.86

Activations Density 0.096%

personal references pointing to the speaker

references to the speaker or self

No Comments

No Known Activations