INDEX

Explanations

phrases related to old-fashioned items or concepts

oai_token-act-pair · gpt-3.5-turbo

references to traditional or retro elements

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

 nerv

-0.70

HRC

-0.65

 Mermaid

-0.64

 terminated

-0.64

 contagious

-0.64

 deductible

-0.63

 istg

-0.63

 cancellation

-0.63

 openly

-0.62

âĶĢâĶĢ

-0.62

POSITIVE LOGITS

fashioned

1.76

school

1.43

style

1.38

timer

1.37

growth

1.29

world

1.27

tim

1.26

time

1.24

money

1.20

fashion

1.18

Activations Density 0.032%

phrases related to old-fashioned items or concepts

references to traditional or retro elements

No Comments

No Known Activations