INDEX

Explanations

forms of the verbs 'do' and 'does'

oai_token-act-pair · gpt-3.5-turbo Triggered by @bot

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 8-res_fs3072-jb

Configuration

jbloom/GPT2-Small-Feature-Splitting-Experiment-Layer-8/blocks.8.hook_resid_pre_3072

Prompts (Dashboard)

24,576 prompts, 128 tokens each

Dataset (Dashboard)

Skylion007/openwebtext

Features

3,072

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

PlotsExplanationShow Test FieldDefault Test Text

IFrame

Link

Not in Any Lists

No Comments

Negative Logits

case

-0.69

Hok

-0.68

hog

-0.65

 Ages

-0.65

 Printed

-0.64

fields

-0.63

 palms

-0.62

 Entered

-0.60

 Dise

-0.60

stood

-0.60

POSITIVE LOGITS

pez

1.07

ppel

1.06

omsday

1.00

herty

0.96

not

0.95

vet

0.94

oms

0.92

NOT

0.92

zed

0.89

actic

0.89

Activations Density 0.341%

forms of the verbs 'do' and 'does'

No Comments

No Known Activations