INDEX

Explanations

words or phrases related to a specific place or person named "Cuthbert"

oai_token-act-pair · gpt-3.5-turbo

repeated mentions of the term "auth" or similar variations, likely related to authentication or authoring content

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

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 0-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.0.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.0.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

 stalls

-0.73

 stock

-0.68

 gaming

-0.66

stay

-0.65

 scenery

-0.64

 flickering

-0.64

 compact

-0.64

MOV

-0.63

 short

-0.63

 animated

-0.63

POSITIVE LOGITS

uth

4.83

UTH

2.36

uthor

1.43

ut

1.23

uts

1.23

oth

1.18

uther

1.14

uten

1.11

auth

1.11

utt

1.10

Activations Density 0.020%

words or phrases related to a specific place or person named "Cuthbert"

repeated mentions of the term "auth" or similar variations, likely related to authentication or authoring content

No Comments

No Known Activations