INDEX

Explanations

references to historical figures with titles of authority

oai_token-act-pair · gpt-3.5-turbo

references to emperors and their titles

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

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 5-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.5.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.5.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

ties

-0.83

esville

-0.75

crew

-0.74

ter

-0.72

ledge

-0.71

chel

-0.70

solid

-0.69

asket

-0.67

icult

-0.66

stairs

-0.66

POSITIVE LOGITS

peror

1.16

 Nero

1.08

perors

0.99

 Napoleon

0.94

 Augustus

0.87

 Claud

0.87

 Marshal

0.84

 Emperor

0.83

 pengu

0.83

 Hiro

0.80

Activations Density 0.020%

references to historical figures with titles of authority

references to emperors and their titles

No Comments

No Known Activations