INDEX

Explanations

phrases related to authority or commands

oai_token-act-pair · gpt-3.5-turbo

instances of personal statements or direct quotes regarding societal expectations and legal matters

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

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 5-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.5.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.5.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

 partName

-0.74

 argues

-0.68

 sqor

-0.62

geoning

-0.61

PCR

-0.61

 confines

-0.60

 Auditor

-0.57

esters

-0.57

Attempts

-0.57

 understandably

-0.56

POSITIVE LOGITS

"#

0.78

 undermin

0.75

ĪĴ

0.71

 evils

0.65

 superiority

0.65

 someday

0.62

 immoral

0.61

 fitt

0.60

 imminent

0.60

 TAMADRA

0.60

Activations Density 1.447%

phrases related to authority or commands

instances of personal statements or direct quotes regarding societal expectations and legal matters

No Comments

No Known Activations