INDEX

Explanations

words related to undermining, demeaning, or diminishing

oai_token-act-pair · gpt-3.5-turbo

derogatory or belittling language towards individuals or groups

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

 unauthorized

-0.68

ASA

-0.66

ca

-0.66

ECB

-0.65

 auth

-0.64

 Board

-0.63

deb

-0.61

 Venice

-0.60

 admissions

-0.60

 commission

-0.60

POSITIVE LOGITS

horse

3.62

knife

1.32

ittle

1.32

truth

1.21

Knight

1.13

 ethic

1.12

lust

1.12

quit

1.08

times

1.07

loving

1.05

Activations Density 0.033%

words related to undermining, demeaning, or diminishing

derogatory or belittling language towards individuals or groups

No Comments

No Known Activations