INDEX

Explanations

words related to hidden or private things

oai_token-act-pair · gpt-3.5-turbo

variations of the prefix "und," indicating negation or the absence of a quality

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

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 10-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.10.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.10.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

 gorilla

-0.70

Ĥİ

-0.66

OUP

-0.60

BOOK

-0.60

 fart

-0.58

TION

-0.58

 speakers

-0.57

auga

-0.56

ongyang

-0.56

 twist

-0.56

POSITIVE LOGITS

etermined

1.36

aunted

1.27

oubtedly

1.27

oubted

1.26

iscovered

1.25

efined

1.25

oubt

1.16

ifferent

1.15

ecided

1.07

irect

1.05

Activations Density 0.022%

words related to hidden or private things

variations of the prefix "und," indicating negation or the absence of a quality

No Comments

No Known Activations