INDEX

Explanations

phrases related to having permission or ability to obtain something

oai_token-act-pair · gpt-3.5-turbo

phrases indicating access or permission to resources or information

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

 Kern

-0.60

Jou

-0.59

 scra

-0.56

 Krug

-0.56

 Lange

-0.55

 Sending

-0.54

Ki

-0.54

 farewell

-0.52

 Goodbye

-0.52

POSITIVE LOGITS

ibility

1.22

ibly

1.17

 thereto

1.10

ible

1.01

ibilities

1.00

to

0.98

ibl

0.95

ion

0.95

ioned

0.95

IBLE

0.92

Activations Density 0.045%

phrases related to having permission or ability to obtain something

phrases indicating access or permission to resources or information

No Comments

No Known Activations