INDEX

Explanations

proper nouns related to news or media

oai_token-act-pair · gpt-3.5-turbo

mentions of the letter 'M'

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

IFrame

Link

Not in Any Lists

No Comments

Negative Logits

 Wolves

-0.69

 thumbs

-0.67

 pony

-0.67

 Mane

-0.66

 firsthand

-0.65

 exacerbated

-0.64

 hottest

-0.64

sto

-0.63

Bil

-0.63

 undercut

-0.63

POSITIVE LOGITS

ountain

1.39

umbai

1.33

useum

1.32

unicip

1.30

ikhail

1.29

arijuana

1.24

ovies

1.23

eredith

1.22

igration

1.21

ISSION

1.21

Activations Density 0.041%

proper nouns related to news or media

mentions of the letter 'M'

No Comments

No Known Activations