INDEX

Explanations

words related to people's names or titles

oai_token-act-pair · gpt-3.5-turbo

instances of the letter "r"

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

 EntityItem

-0.81

£ı

-0.68

æĸ¹

-0.65

lace

-0.65

EStream

-0.64

 Fram

-0.64

 reusable

-0.62

 Contra

-0.61

 GOODMAN

-0.61

CPR

-0.60

POSITIVE LOGITS

abbit

1.18

agh

1.15

agnar

1.14

angers

1.10

acing

1.07

anger

1.07

ach

1.03

iddled

1.03

ussia

1.02

outine

1.02

Activations Density 0.023%

words related to people's names or titles

instances of the letter "r"

No Comments

No Known Activations