INDEX

Explanations

No Explanations Found

New Auto-Interp

Configuration

ctigges/pythia-70m-deduped__mlp-sm_processed/0-mlp-sm

Prompts (Dashboard)

32,768 prompts, 128 tokens each

Dataset (Dashboard)

monology/pile-uncopyrighted

Features

32,768

Data Type

torch.float32

Hook Name

blocks.0.hook_mlp_out

Hook Layer

Architecture

standard

Context Size

128

Dataset

EleutherAI/the_pile_deduplicated

Activation Function

relu

Embeds

PlotsExplanationShow Test FieldDefault Test Text

IFrame

Link

Not in Any Lists

No Comments

Negative Logits

onder

-1.56

enth

-1.42

dimethyl

-1.42

aston

-1.39

going

-1.39

"}](#

-1.39

vier

-1.39

'$.

-1.39

severe

-1.38

ouss

-1.37

POSITIVE LOGITS

 predecessor

1.60

 pals

1.48

 favourite

1.47

 front

1.46

nem

1.45

isans

1.45

 critics

1.45

 predecessors

1.43

 caption

1.42

Activations Density 0.214%

No Known Activations

This feature has no known activations.