INDEX

Explanations

the word "connected" and its variations indicating relationships or links

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

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

IFrame

Link

Not in Any Lists

No Comments

Negative Logits

called

-1.67

"`

-1.51

 sweat

-1.51

 plagued

-1.51

dried

-1.50

woke

-1.48

 dominated

-1.44

»¿

-1.44

 blessed

-1.44

 poor

-1.44

POSITIVE LOGITS

endar

1.94

tip

1.89

iliary

1.80

iom

1.74

appendix

1.73

etable

1.72

icine

1.70

ieval

1.69

icense

1.68

emente

1.65

Activations Density 0.010%

the word "connected" and its variations indicating relationships or links

No Comments

No Known Activations