INDEX

Explanations

the word "we"

oai_token-act-pair · gpt-3.5-turbo

references to a collective group or community

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

 exception

-0.65

 sidx

-0.61

 amusement

-0.59

 value

-0.58

 reserved

-0.58

 Marriott

-0.57

 citation

-0.56

 exceptions

-0.56

 firsthand

-0.56

 embarrassment

-0.55

POSITIVE LOGITS

we

3.97

WE

1.87

We

1.44

wer

1.36

they

1.35

our

1.34

well

1.34

you

1.32

1.30

weed

1.29

Activations Density 0.008%

the word "we"

references to a collective group or community

No Comments

No Known Activations