INDEX

Explanations

introductory phrases indicating interviews or conversations

oai_token-act-pair · gpt-3.5-turbo

mentions of interviews and discussions involving individuals or characters

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

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 11-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.11.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.11.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

 exempted

-0.70

 defund

-0.69

 retri

-0.68

arnaev

-0.68

 costing

-0.68

anooga

-0.68

 unconstitutional

-0.67

ignty

-0.66

 federally

-0.66

 repeal

-0.65

POSITIVE LOGITS

 Collider

1.40

 Variety

1.22

IGN

1.22

THR

1.18

 Entertainment

1.18

EW

1.16

 Eater

1.11

 Vanity

1.08

TMZ

1.08

MTV

1.07

Activations Density 0.219%

introductory phrases indicating interviews or conversations

mentions of interviews and discussions involving individuals or characters

No Comments

No Known Activations