INDEX

Explanations

words related to misconduct or inappropriate behavior

oai_token-act-pair · gpt-3.5-turbo

variations of the word "behavior."

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

 Rwanda

-0.74

 Korean

-0.71

 Panthers

-0.69

 Panther

-0.68

 Nordic

-0.66

 Panzer

-0.66

 Purg

-0.64

ãģĤ

-0.64

 Kinnikuman

-0.64

 Roof

-0.63

POSITIVE LOGITS

beh

1.45

aviour

1.29

Beh

0.96

terness

0.89

behavior

0.89

avin

0.89

Beh

0.84

abus

0.83

 behav

0.81

ilib

0.81

Activations Density 0.007%

words related to misconduct or inappropriate behavior

variations of the word "behavior."

No Comments

No Known Activations