INDEX

Explanations

words related to journalistic sources, particularly "Ars" and "Verge."

oai_token-act-pair · gpt-3.5-turbo

mentions of the entity "Ars."

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

IFrame

Link

Not in Any Lists

No Comments

Negative Logits

 Nationwide

-0.91

 bear

-0.68

 McMahon

-0.67

ë

-0.66

bered

-0.65

taker

-0.64

UNC

-0.64

<+

-0.64

 Wilmington

-0.63

employment

-0.63

POSITIVE LOGITS

Ars

1.10

 challeng

0.89

isoft

0.86

 Skydragon

0.86

Ã¨

0.85

 mathemat

0.82

antis

0.79

 defic

0.79

 metic

0.78

 resil

0.78

Activations Density 0.004%

words related to journalistic sources, particularly "Ars" and "Verge."

mentions of the entity "Ars."

No Comments

No Known Activations