INDEX

Explanations

phrases or sentences related to corrections or updates in a document

oai_token-act-pair · gpt-3.5-turbo

references to Reddit posts and comments

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

ãĤ¼ãĤ¦ãĤ¹

-0.83

cycles

-0.82

ometers

-0.78

zees

-0.78

phthal

-0.77

negie

-0.76

ulic

-0.75

stals

-0.75

osponsors

-0.75

olics

-0.75

POSITIVE LOGITS

 article

1.81

 statement

1.59

 tweet

1.57

 excerpt

1.57

 quote

1.57

 paragraph

1.50

 comment

1.50

 letter

1.48

 interview

1.46

 remark

1.44

Activations Density 0.422%

phrases or sentences related to corrections or updates in a document

references to Reddit posts and comments

No Comments

No Known Activations