INDEX

Explanations

words associated with quality, value, and effectiveness

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

New Auto-Interp

Configuration

ctigges/pythia-70m-deduped__res-sm_processed/2-res-sm

Prompts (Dashboard)

32,768 prompts, 128 tokens each

Dataset (Dashboard)

monology/pile-uncopyrighted

Features

32,768

Data Type

torch.float32

Hook Name

blocks.2.hook_resid_post

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

²

-2.15

·¸

-1.99

ĻĤ

-1.99

Ĥ¬

-1.92

Ĳ

-1.92

½

-1.73

Ĵ

-1.63

Īĺ

-1.62

riber

-1.59

openh

-1.58

POSITIVE LOGITS

 itself

1.82

 advantage

1.65

ulous

1.55

[$

1.45

 against

1.43

 remains

1.43

otal

1.43

[@

1.40

 constituents

1.39

 consisted

1.39

Activations Density 0.887%

words associated with quality, value, and effectiveness

No Comments

No Known Activations