INDEX

Explanations

references to the word "soda."

oai_token-act-pair · gpt-3.5-turbo

references to soda products

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

PlotsExplanationShow Test FieldDefault Test Text

IFrame

Link

Not in Any Lists

No Comments

Negative Logits

auld

-0.76

aird

-0.74

HCR

-0.73

debian

-0.72

sure

-0.72

izations

-0.70

MSN

-0.70

blogspot

-0.67

isations

-0.67

ions

-0.66

POSITIVE LOGITS

 soda

1.30

 Soda

1.22

 Coke

0.90

Cola

0.88

 Flask

0.87

 beverage

0.86

 flask

0.84

Nit

0.83

 bottle

0.83

 dioxide

0.82

Activations Density 0.003%

references to the word "soda."

references to soda products

No Comments

No Known Activations