INDEX

Explanations

recommendations or obligations expressed with the word "should."

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

IFrame

Link

Negative Logits

itness

-1.75

selves

-1.67

ness

-1.64

Ĵ

-1.64

 lessons

-1.63

matically

-1.60

bit

-1.57

ories

-1.56

 insights

-1.46

bringing

-1.42

POSITIVE LOGITS

 replace

1.75

be

1.60

 OPINION

1.60

 ditch

1.56

 NEVER

1.54

 resign

1.53

NOT

1.52

bracket

1.48

iox

1.46

not

1.43

Activations Density 0.181%

recommendations or obligations expressed with the word "should."

No Known Activations