INDEX

Explanations

words related to intense emotions or actions, such as 'detector', 'outrage', 'genius', 'inexplicable', 'love', 'trans', 'giggles', 'sly', 'bored', 'quick-trigger'

oai_token-act-pair · gpt-3.5-turbo

expressions related to humor and comedic elements

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

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 8-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.8.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.8.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

accept

-0.62

 conduc

-0.62

Sah

-0.61

ourses

-0.61

PTS

-0.61

 offshore

-0.61

 leasing

-0.60

 challeng

-0.59

 arsen

-0.59

âĵĺ

-0.58

POSITIVE LOGITS

pants

1.07

 chuckle

1.05

 understatement

1.02

 sarc

1.01

 irony

0.98

 grin

0.97

 nerd

0.97

 laughter

0.95

 jokes

0.94

 memes

0.94

Activations Density 0.774%

words related to intense emotions or actions, such as 'detector', 'outrage', 'genius', 'inexplicable', 'love', 'trans', 'giggles', 'sly', 'bored', 'quick-trigger'

expressions related to humor and comedic elements

No Comments

No Known Activations