INDEX

Explanations

numerical references and coded data

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

ĸ´

-3.20

İ

-3.10

ĭ

-3.01

ī

-2.87

ħ

-2.83

¶

-2.79

ı

-2.71

¨

-2.70

¤

-2.66

Ī

-2.64

POSITIVE LOGITS

 respectively

1.75

 matter

1.46

***

1.36

bond

1.33

 traveled

1.30

.—

1.29

^](#

1.27

related

1.25

 transmitted

1.25

hores

1.23

Activations Density 6.567%

numerical references and coded data

No Comments

No Known Activations