INDEX

Explanations

words conveying extreme levels of disbelief or ridicule

oai_token-act-pair · gpt-3.5-turbo

descriptors of absurdity and ridiculousness

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

yer

-0.88

Reviewed

-0.82

enfranch

-0.78

rien

-0.74

avers

-0.74

ribution

-0.71

maid

-0.71

builders

-0.71

rounder

-0.70

oyal

-0.70

POSITIVE LOGITS

ness

0.89

 amounts

0.89

nesses

0.89

 nonsense

0.85

 absurdity

0.85

ly

0.84

LY

0.83

 lengths

0.82

 amount

0.79

NESS

0.79

Activations Density 0.042%

words conveying extreme levels of disbelief or ridicule

descriptors of absurdity and ridiculousness

No Comments

No Known Activations