INDEX

Explanations

references to negative emotions or unfavorable situations

oai_token-act-pair · gpt-3.5-turbo

references to the concept of shame

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

arbon

-0.71

irements

-0.70

opez

-0.68

ancial

-0.67

igion

-0.66

resp

-0.65

auri

-0.65

efficients

-0.64

mental

-0.64

thus

-0.63

POSITIVE LOGITS

 shame

1.36

 Shame

1.21

fully

1.07

faced

0.94

 ashamed

0.90

ously

0.89

 pity

0.86

uously

0.84

 shaming

0.83

imaru

0.79

Activations Density 0.008%

references to negative emotions or unfavorable situations

references to the concept of shame

No Comments

No Known Activations