INDEX

Explanations

references to something being new or novel

oai_token-act-pair · gpt-3.5-turbo

references to the concept of "new" or novelty

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

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 4-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.4.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.4.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

aminer

-0.80

ashtra

-0.71

veyard

-0.70

UMP

-0.67

ÃĥÃĤÃĥÃĤÃĥÃĤÃĥÃĤÃĥÃĤÃĥÃĤÃĥÃĤÃĥÃĤÃĥÃĤÃĥÃĤÃĥÃĤÃĥÃĤÃĥÃĤÃĥÃĤÃĥÃĤÃĥÃĤÃĥÃĤÃĥÃĤÃĥÃĤÃĥÃĤÃĥÃĤÃĥÃĤÃĥÃĤÃĥÃĤÃĥÃĤÃĥÃĤÃĥÃĤÃĥÃĤÃĥÃĤÃĥÃĤÃĥÃĤÃĥÃĤ

-0.67

ammers

-0.66

itta

-0.65

Zip

-0.64

xual

-0.63

otto

-0.63

POSITIVE LOGITS

bies

1.42

bie

1.33

 Zealand

1.16

found

0.92

foundland

0.89

 arrivals

0.88

 discoveries

0.86

born

0.85

 developments

0.84

ãĤ´

0.81

Activations Density 0.105%

references to something being new or novel

references to the concept of "new" or novelty

No Comments

No Known Activations