INDEX

Explanations

height measurements

oai_token-act-pair · gpt-3.5-turbo

references to high ratings or quality assessments

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

 Ragnarok

-0.68

arium

-0.67

 Canad

-0.63

Ain

-0.63

 Miko

-0.60

 Muhammad

-0.60

 Kinnikuman

-0.59

 Kung

-0.59

Sin

-0.59

 healed

-0.59

POSITIVE LOGITS

lighting

1.01

lights

0.95

high

0.91

nesses

0.87

ened

0.81

rising

0.80

stakes

0.74

baugh

0.73

low

0.72

school

0.71

Activations Density 0.008%

height measurements

references to high ratings or quality assessments

No Comments

No Known Activations