INDEX

Explanations

phrases related to receiving emails or containing specific content

oai_token-act-pair · gpt-3.5-turbo

phrases that describe items or substances that are contained within something else

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

imeo

-0.74

rek

-0.68

rians

-0.66

ime

-0.63

udi

-0.62

reprene

-0.62

aughs

-0.61

soc

-0.59

maxwell

-0.57

eda

-0.57

POSITIVE LOGITS

 containing

3.37

containing

2.04

 contain

1.97

 consisting

1.86

 contains

1.80

 Contains

1.62

 comprising

1.61

 contained

1.61

 depicting

1.50

 featuring

1.41

Activations Density 0.019%

phrases related to receiving emails or containing specific content

phrases that describe items or substances that are contained within something else

No Comments

No Known Activations