INDEX

Explanations

phrases indicating something that has never been done or seen before

oai_token-act-pair · gpt-3.5-turbo

phrases indicating unprecedented occurrences or experiences

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

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 11-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.11.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.11.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

 Torrent

-0.66

ument

-0.61

 deletion

-0.61

 Bulg

-0.60

 Mandatory

-0.59

soType

-0.58

 Emails

-0.57

 Carly

-0.57

 Bullets

-0.56

 Moral

-0.56

POSITIVE LOGITS

 dreamed

1.08

 imaginable

1.06

before

1.05

 imagined

1.02

 before

1.00

 previously

0.99

 hitherto

0.97

seen

0.94

¥µ

0.93

 existed

0.92

Activations Density 0.168%

phrases indicating something that has never been done or seen before

phrases indicating unprecedented occurrences or experiences

No Comments

No Known Activations