INDEX

Explanations

percentages expressed in numbers

oai_token-act-pair · gpt-3.5-turbo

references to percentage values

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

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 0-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.0.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.0.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

 Rain

-0.73

 tire

-0.70

Sal

-0.65

 Melvin

-0.64

Haf

-0.64

 rain

-0.62

 Bashar

-0.62

 Andre

-0.62

 honoring

-0.61

 Eliot

-0.61

POSITIVE LOGITS

pc

4.41

pc

2.19

PC

2.10

pb

1.36

pn

1.28

pd

1.27

mc

1.26

bp

1.23

pg

1.22

jc

1.18

Activations Density 0.010%

percentages expressed in numbers

references to percentage values

No Comments

No Known Activations