INDEX

Explanations

the word "tower" and related terms

oai_token-act-pair · gpt-3.5-turbo

references to various types of towers

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

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 3-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.3.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.3.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

ãĤ±

-0.83

dit

-0.81

 Lives

-0.76

furt

-0.76

Ago

-0.72

vous

-0.71

 Rowling

-0.71

esville

-0.69

 Breed

-0.69

 Snap

-0.68

POSITIVE LOGITS

 towers

1.32

 tower

1.30

tower

1.10

 Towers

1.03

 skysc

0.92

 heights

0.89

 erected

0.86

 crane

0.85

 Tower

0.81

 blocks

0.79

Activations Density 0.007%

the word "tower" and related terms

references to various types of towers

No Comments

No Known Activations