INDEX

Explanations

phrases related to the ability to do something

oai_token-act-pair · gpt-3.5-turbo

the word "can" in various contexts

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

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 0-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.0.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.0.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

amaz

-0.71

appointed

-0.58

aign

-0.58

intent

-0.56

lighting

-0.55

udging

-0.54

tones

-0.54

majority

-0.53

ded

-0.53

adden

-0.52

POSITIVE LOGITS

can

2.79

 cannot

1.94

can

1.93

 could

1.72

CAN

1.63

Can

1.63

may

1.55

Can

1.53

 will

1.42

 must

1.41

Activations Density 0.178%

phrases related to the ability to do something

the word "can" in various contexts

No Comments

No Known Activations