INDEX

Explanations

phrases that indicate a listing or enumeration of items

oai_token-act-pair · gpt-3.5-turbo

the word "an"

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

'an' or ' an' token

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

•single token firings

Fires on " an", "an", " An" "nan" on a SINGLE token, but not always in a longer context?

No Comments

Negative Logits

 Jagu

-0.71

 Borders

-0.68

 Bris

-0.65

 Shake

-0.64

 branches

-0.64

 Howe

-0.63

stones

-0.62

 Greene

-0.61

 possessions

-0.61

 McKay

-0.61

POSITIVE LOGITS

abolic

0.96

alysed

0.93

notation

0.92

onymous

0.91

cients

0.88

notations

0.85

onym

0.83

olin

0.81

alyses

0.80

ointed

0.80

Activations Density 0.022%

phrases that indicate a listing or enumeration of items

the word "an"

'an' or ' an' token

No Comments

No Known Activations