INDEX

Explanations

violent incidents involving harm to people

oai_token-act-pair · gpt-3.5-turbo

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 10-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.10.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.10.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

isSpecialOrderable

-0.83

ĨĴ

-0.81

 Relax

-0.71

hedon

-0.70

 Awareness

-0.70

 Applic

-0.70

yi

-0.69

 Aware

-0.68

 QUEST

-0.67

lear

-0.67

POSITIVE LOGITS

 injuring

1.56

 wounding

1.44

 destroying

1.43

 injure

1.39

 killing

1.32

 wreck

1.28

 robbing

1.23

 murdering

1.20

 cripp

1.19

 inciner

1.19

Activations Density 0.244%

violent incidents involving harm to people

No Comments

No Known Activations