INDEX

Explanations

people or groups being punished or prosecuted

oai_token-act-pair · gpt-3.5-turbo

references to individuals and groups in the context of speech, expression, and legal consequences

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

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 11-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.11.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.11.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

ggles

-0.74

rawdownloadcloneembedreportprint

-0.74

++++++++++++++++

-0.68

atility

-0.65

 Surviv

-0.63

kefeller

-0.63

 teamwork

-0.62

 Recovery

-0.61

BuyableInstoreAndOnline

-0.61

 flashback

-0.59

POSITIVE LOGITS

 suspected

1.21

 deemed

1.19

 accused

1.15

 convicted

1.10

 whose

1.09

who

1.09

 exercising

1.05

 violating

1.01

who

0.98

 complicit

0.96

Activations Density 0.278%

people or groups being punished or prosecuted

references to individuals and groups in the context of speech, expression, and legal consequences

No Comments

No Known Activations