INDEX

Explanations

phrases indicating advice or recommendations

oai_token-act-pair · gpt-3.5-turbo

instances of the word "you" and its variations to indicate direct address

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

obbies

-0.79

rates

-0.66

 implements

-0.64

Scroll

-0.64

 flourish

-0.64

 refinement

-0.63

 dismant

-0.62

asury

-0.62

 apologies

-0.60

IDs

-0.58

POSITIVE LOGITS

 gonna

1.23

 obligated

1.07

 likely

1.05

 probably

1.01

 lucky

1.01

 entitled

0.99

 wasting

0.98

 liable

0.97

 going

0.95

probably

0.95

Activations Density 0.151%

phrases indicating advice or recommendations

instances of the word "you" and its variations to indicate direct address

No Comments

No Known Activations