INDEX

Explanations

references to people holding positions of authority or leadership

oai_token-act-pair · gpt-3.5-turbo

references to government positions or titles

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

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 0-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.0.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.0.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

aeus

-0.72

itars

-0.65

 alike

-0.65

iera

-0.65

 constitu

-0.64

NetMessage

-0.62

raq

-0.62

merce

-0.62

 kindness

-0.60

ques

-0.59

POSITIVE LOGITS

 Under

3.50

Under

2.50

 UNDER

1.93

under

1.85

 under

1.73

 Over

1.33

Und

1.16

 Purs

1.13

 Underground

1.09

 Above

1.09

Activations Density 0.022%

references to people holding positions of authority or leadership

references to government positions or titles

No Comments

No Known Activations