INDEX

Explanations

text referring to different parts (top, bottom, head, base, eastern) of physical structures

oai_token-act-pair · gpt-3.5-turbo

references to positions in a document

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

PlotsExplanationShow Test FieldDefault Test Text

IFrame

Link

Not in Any Lists

No Comments

Negative Logits

SHARE

-0.66

ulia

-0.62

 McGee

-0.62

tons

-0.58

ullivan

-0.56

manship

-0.55

Live

-0.54

vich

-0.54

 alone

-0.54

asha

-0.53

POSITIVE LOGITS

 bottom

1.42

end

1.29

 outset

1.22

top

1.19

 beginning

1.14

 ends

1.08

 forefront

1.07

 junction

1.05

bottom

1.05

 periphery

1.03

Activations Density 0.108%

text referring to different parts (top, bottom, head, base, eastern) of physical structures

references to positions in a document

No Comments

No Known Activations