INDEX

Explanations

words related to negative conditions, such as illness or malice

oai_token-act-pair · gpt-3.5-turbo

references to illness or being unwell

oai_token-act-pair · gpt-4o-mini Triggered by @bot

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 2-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.2.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.2.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

EStream

-0.70

EStreamFrame

-0.67

ICAN

-0.66

*/(

-0.65

BOOK

-0.64

ļéĨĴ

-0.63

âĹ¼

-0.63

lished

-0.63

ourke

-0.63

oother

-0.61

POSITIVE LOGITS

uminati

1.25

inois

1.24

umin

1.03

iberal

1.00

nesses

0.97

ison

0.95

ogical

0.93

awar

0.93

iquid

0.89

usive

0.87

Activations Density 0.005%

words related to negative conditions, such as illness or malice

references to illness or being unwell

No Comments

No Known Activations