INDEX

Explanations

references to the country South Korea

oai_token-act-pair · gpt-3.5-turbo

mentions of South Korea

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

IFrame

Link

Negative Logits

 "$:/

-0.76

inventoryQuantity

-0.73

TED

-0.69

INESS

-0.67

ACA

-0.66

 captcha

-0.63

LEASE

-0.62

ograp

-0.61

ovie

-0.61

ICAN

-0.60

POSITIVE LOGITS

wark

1.50

western

1.28

 Carolina

1.18

 Africa

1.10

ward

1.08

 Dakota

1.08

ampton

1.05

 Korea

1.05

 Sudan

1.04

 Africans

0.99

Activations Density 0.037%

references to the country South Korea

mentions of South Korea

No Known Activations