INDEX

Explanations

references to investigations or scrutiny related to potential dishonesty or corruption

oai_token-act-pair · gpt-3.5-turbo

sections of text that indicate strong opinions or criticisms

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

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 5-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.5.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.5.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

xual

-0.78

 withdraw

-0.73

ignt

-0.70

ertodd

-0.67

 conclude

-0.65

ignty

-0.65

 abandon

-0.64

 Aval

-0.63

 Calder

-0.62

 dismantle

-0.62

POSITIVE LOGITS

Unless

0.91

Seriously

0.85

Especially

0.78

Specifically

0.78

Contents

0.77

Except

0.75

Think

0.75

Sure

0.75

Statistics

0.73

Writer

0.72

Activations Density 0.497%

references to investigations or scrutiny related to potential dishonesty or corruption

sections of text that indicate strong opinions or criticisms

No Comments

No Known Activations