INDEX

Explanations

phrases related to explosive devices

oai_token-act-pair · gpt-3.5-turbo

references to bombs and bombing 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

anse

-0.84

emale

-0.78

ean

-0.76

eanor

-0.71

eway

-0.71

disciplinary

-0.69

riott

-0.68

Customer

-0.68

insula

-0.67

bred

-0.67

POSITIVE LOGITS

 bombs

3.82

 Bombs

2.48

 bombings

2.13

 bombers

2.06

 bomb

2.05

 grenades

1.95

bomb

1.81

 explosives

1.79

 bombing

1.74

 Bomb

1.70

Activations Density 0.009%

phrases related to explosive devices

references to bombs and bombing incidents

No Comments

No Known Activations