INDEX

Explanations

proper nouns ending in 'ner'

oai_token-act-pair · gpt-3.5-turbo

references to individuals or their surnames

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

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 1-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.1.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.1.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

Introduced

-0.76

 varying

-0.68

 Occupations

-0.66

UTC

-0.64

 intervening

-0.61

 undefined

-0.61

PKK

-0.60

 twisted

-0.59

 disparate

-0.59

 taxing

-0.59

POSITIVE LOGITS

ner

1.33

ners

1.19

wagen

1.05

gie

0.98

cair

0.93

lund

0.92

getic

0.91

amic

0.90

strom

0.89

ocious

0.88

Activations Density 0.013%

proper nouns ending in 'ner'

references to individuals or their surnames

No Comments

No Known Activations