INDEX

Explanations

punctuation marks used for emphasis or as part of a list

oai_token-act-pair · gpt-3.5-turbo

negations or denials in statements

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

abal

-0.71

ģĸ

-0.71

adelphia

-0.68

FIG

-0.63

ifted

-0.63

MpServer

-0.62

 comprom

-0.61

alone

-0.60

 culminated

-0.60

 eleph

-0.60

POSITIVE LOGITS

sir

0.95

 Nope

0.93

 wait

0.86

 seriously

0.86

 sorry

0.84

 Sorry

0.81

not

0.80

 none

0.78

no

0.77

not

0.77

Activations Density 0.064%

punctuation marks used for emphasis or as part of a list

negations or denials in statements

No Comments

No Known Activations