INDEX

Explanations

phrases that indicate actions perceived as wrong or illegal

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

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 8-res_fs49152-jb

Configuration

jbloom/GPT2-Small-Feature-Splitting-Experiment-Layer-8/blocks.8.hook_resid_pre_49152

Prompts (Dashboard)

24,576 prompts, 128 tokens each

Dataset (Dashboard)

Skylion007/openwebtext

Features

49,152

Data Type

torch.float32

Hook Point

blocks.8.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

lime

-0.76

uli

-0.69

otti

-0.68

Luck

-0.66

calling

-0.65

cedented

-0.65

quickShipAvailable

-0.64

soDeliveryDate

-0.64

rising

-0.63

furt

-0.63

POSITIVE LOGITS

 compare

1.05

 engage

1.04

 omit

1.03

 impose

1.02

 contemplate

1.00

 introduce

0.99

ask

0.98

 celebrate

0.96

 underestimate

0.96

 incorporate

0.96

Activations Density 0.066%

phrases that indicate actions perceived as wrong or illegal

No Comments

No Known Activations