INDEX

Explanations

phrases related to law enforcement agencies and legal proceedings

oai_token-act-pair · gpt-3.5-turbo

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 10-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.10.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.10.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

oldemort

-0.81

yz

-0.75

OVER

-0.74

Reviewer

-0.73

Ô

-0.73

flows

-0.69

reddits

-0.69

gebra

-0.68

obyl

-0.68

olkien

-0.68

POSITIVE LOGITS

 office

1.50

 Office

1.40

 department

1.18

office

1.15

 offices

1.09

 deputies

1.03

 Department

1.02

Office

1.01

 aide

0.95

 mansion

0.94

Activations Density 0.063%

phrases related to law enforcement agencies and legal proceedings

No Comments

No Known Activations