INDEX

Explanations

phrases related to physical actions involving force or destruction

oai_token-act-pair · gpt-3.5-turbo

negative prefixes or terms associated with deterioration or reduction

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

IFrame

Link

Not in Any Lists

No Comments

Negative Logits

 amplified

-0.65

 haunt

-0.65

 locality

-0.65

 consolation

-0.64

 mercy

-0.64

 Vaughn

-0.63

 chalk

-0.62

 modification

-0.62

 prized

-0.62

 illuminated

-0.62

POSITIVE LOGITS

up

1.36

down

1.32

earth

1.31

upon

1.31

edge

1.27

together

1.25

sounding

1.24

enough

1.23

paren

1.22

again

1.22

Activations Density 0.047%

phrases related to physical actions involving force or destruction

negative prefixes or terms associated with deterioration or reduction

No Comments

No Known Activations