INDEX

Explanations

the word "culprit"

oai_token-act-pair · gpt-3.5-turbo

terms related to identifying a culprit or mastermind behind an action or event

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

PlotsExplanationShow Test FieldDefault Test Text

IFrame

Link

Not in Any Lists

No Comments

Negative Logits

ear

-0.75

iring

-0.74

upp

-0.72

formed

-0.70

cyclop

-0.69

pec

-0.68

achine

-0.67

porting

-0.64

ORT

-0.63

intend

-0.63

POSITIVE LOGITS

 culprit

1.46

 offenders

0.84

 offender

0.83

DragonMagazine

0.80

 mastermind

0.80

 perpetrator

0.71

ãĥ¼ãĥĨãĤ£

0.71

æ©

0.70

enary

0.70

 shenan

0.69

Activations Density 0.009%

the word "culprit"

terms related to identifying a culprit or mastermind behind an action or event

No Comments

No Known Activations