INDEX

Explanations

the word "not" followed by a strong emphasis on the subsequent word or phrase

oai_token-act-pair · gpt-3.5-turbo

negations or phrases indicating refusal

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

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 9-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.9.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.9.hook_resid_pre

Architecture

standard

Context Size

128

Dataset

Skylion007/openwebtext

Hook Point Layer

Activation Function

relu

Embeds

PlotsExplanationShow Test FieldDefault Test Text

IFrame

Link

Not in Any Lists

No Comments

Negative Logits

å¥

-0.85

ounters

-0.73

cano

-0.69

æ©

-0.67

ãĤ¼

-0.65

 Hots

-0.65

ously

-0.65

 peaks

-0.64

oise

-0.63

çļ

-0.63

POSITIVE LOGITS

 bothered

1.29

 ashamed

1.25

 afraid

1.24

 interested

1.22

 gonna

1.21

 kidding

1.21

 aware

1.10

 necessarily

1.10

 fooled

1.06

 worried

1.05

Activations Density 0.112%

the word "not" followed by a strong emphasis on the subsequent word or phrase

negations or phrases indicating refusal

No Comments

No Known Activations