INDEX

Explanations

words related to legal actions resulting in imprisonment

oai_token-act-pair · gpt-3.5-turbo

references to individuals being jailed or imprisoned

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

cession

-0.71

lder

-0.67

 Blizz

-0.66

Bra

-0.65

Bru

-0.65

Myr

-0.65

prise

-0.64

avez

-0.64

ww

-0.63

urgical

-0.63

POSITIVE LOGITS

 jailed

1.30

 sentenced

0.97

 imprisoned

0.94

 prisoners

0.93

 inmates

0.92

 incarcerated

0.86

 inmate

0.86

 imprison

0.83

 prisoner

0.82

 detain

0.79

Activations Density 0.006%

words related to legal actions resulting in imprisonment

references to individuals being jailed or imprisoned

No Comments

No Known Activations