INDEX

Explanations

phrases indicating the reader should continue to follow or stay updated on a particular topic

oai_token-act-pair · gpt-3.5-turbo

commands or phrases indicating continuous attention or alerts to stay informed

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

IFrame

Link

Not in Any Lists

No Comments

Negative Logits

-1.01

 muster

-0.74

NetMessage

-0.74

ule

-0.71

 resh

-0.70

»Ĵ

-0.69

iop

-0.69

«ĺ

-0.68

 decor

-0.65

rel

-0.63

POSITIVE LOGITS

 Stay

1.05

 Winner

1.00

Stay

0.98

 Winners

0.87

 Spoiler

0.86

eous

0.84

Ago

0.81

selves

0.77

 Awards

0.77

 Favorite

0.76

Activations Density 0.041%

phrases indicating the reader should continue to follow or stay updated on a particular topic

commands or phrases indicating continuous attention or alerts to stay informed

No Comments

No Known Activations