INDEX

Explanations

references to criminal activities, particularly focusing on incidents involving crime and violence

oai_token-act-pair · gpt-3.5-turbo

references to homicide or related incidents

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

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 0-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.0.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.0.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

stals

-0.69

Tokens

-0.67

oga

-0.65

stream

-0.65

 tune

-0.63

yn

-0.63

bler

-0.62

qual

-0.61

 Appl

-0.61

 Allows

-0.60

POSITIVE LOGITS

 homicide

3.82

 homicides

2.76

 manslaughter

1.98

 murder

1.84

 murders

1.61

 crime

1.49

 suicide

1.46

 burglary

1.45

 killings

1.42

 Murder

1.39

Activations Density 0.008%

references to criminal activities, particularly focusing on incidents involving crime and violence

references to homicide or related incidents

No Comments

No Known Activations