INDEX

Explanations

numeric codes or identifiers that typically denote specific items or categories

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

Not in Any Lists

No Comments

Negative Logits

ilers

-1.58

 bait

-1.55

iling

-1.50

 maker

-1.50

manager

-1.46

etc

-1.41

 nobody

-1.38

ests

-1.38

 protector

-1.38

 ourselves

-1.37

POSITIVE LOGITS

¾

3.98

³

3.32

Ń

3.28

ª

2.90

Ĵ

2.87

¿

2.86

¯

2.85

ĸ´

2.83

²

2.79

Ĥ¬

2.77

Activations Density 0.117%

numeric codes or identifiers that typically denote specific items or categories

No Comments

No Known Activations