INDEX

Explanations

the word "purgatory"

oai_token-act-pair · gpt-3.5-turbo

references to locations or specific places

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

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 10-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.10.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.10.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

 Noble

-0.85

 Breed

-0.83

 Hath

-0.82

 Sweeney

-0.81

Jos

-0.81

 Quin

-0.79

 Tend

-0.78

IDES

-0.77

 Cous

-0.76

 Brah

-0.74

POSITIVE LOGITS

wall

1.00

Wall

0.95

 Wallace

0.90

 wall

0.85

 wallpaper

0.81

 firewall

0.81

tax

0.80

illa

0.80

illery

0.78

apor

0.78

Activations Density 0.361%

the word "purgatory"

references to locations or specific places

No Comments

No Known Activations