INDEX

Explanations

phrases that contain non-alphabetic characters such as parentheses, commas, and numbers within the text

oai_token-act-pair · gpt-3.5-turbo

different article or document publication details

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

 eleph

-0.95

 confir

-0.92

 exha

-0.91

 exting

-0.84

 prosec

-0.80

 tack

-0.79

 tram

-0.79

 destro

-0.78

 neighb

-0.76

 incub

-0.76

POSITIVE LOGITS

Introduction

1.77

Yesterday

1.71

Dear

1.71

Enlarge

1.69

Recently

1.67

Hello

1.62

Today

1.58

Last

1.56

toggle

1.56

Sometimes

1.55

Activations Density 0.340%

phrases that contain non-alphabetic characters such as parentheses, commas, and numbers within the text

different article or document publication details

No Comments

No Known Activations