INDEX

Explanations

phrases related to criminal activities, particularly those involving violence or harm

oai_token-act-pair · gpt-3.5-turbo

terminology related to criminal charges and violent offenses

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

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 3-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.3.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.3.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

atch

-0.89

soDeliveryDate

-0.88

sonian

-0.88

orrow

-0.86

hill

-0.83

tower

-0.80

otle

-0.80

tz

-0.80

ublic

-0.79

bern

-0.79

POSITIVE LOGITS

 aggrav

1.07

 aggravated

1.00

 indecent

0.87

 assault

0.83

 assaults

0.83

 offenses

0.82

 offenders

0.80

 offences

0.77

 mayhem

0.74

 offence

0.72

Activations Density 0.008%

phrases related to criminal activities, particularly those involving violence or harm

terminology related to criminal charges and violent offenses

No Comments

No Known Activations