INDEX

Explanations

references to government officials, particularly ministers

oai_token-act-pair · gpt-3.5-turbo

references to government officials or ministers

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

ASED

-0.70

 Raven

-0.68

 Odyssey

-0.68

 Samurai

-0.68

thia

-0.66

RANT

-0.65

Raven

-0.63

 Rouse

-0.63

facing

-0.63

 Conquest

-0.62

POSITIVE LOGITS

 ministers

1.40

 Ministers

1.13

 minister

1.03

 ministerial

0.96

hip

0.95

hips

0.92

 ministries

0.90

 secretaries

0.90

esses

0.89

afety

0.84

Activations Density 0.008%

references to government officials, particularly ministers

references to government officials or ministers

No Comments

No Known Activations