INDEX

Explanations

reference to joking or humor

oai_token-act-pair · gpt-3.5-turbo

occurrences of the letter 'j'

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

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 10-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.10.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.10.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

 DeVos

-0.66

 outgoing

-0.64

Ful

-0.61

 Magnus

-0.61

 Calder

-0.61

 Fargo

-0.61

 adjourn

-0.61

 Metall

-0.59

fry

-0.59

 Creed

-0.59

POSITIVE LOGITS

umbo

1.39

igsaw

1.38

okingly

1.34

ockey

1.29

okers

1.27

ogging

1.27

agged

1.27

oust

1.23

ocular

1.22

apan

1.22

Activations Density 0.022%

reference to joking or humor

occurrences of the letter 'j'

No Comments

No Known Activations