INDEX

Explanations

titles or honorifics preceding names or entities

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

){#

-1.83

ference

-1.80

braska

-1.50

sterdam

-1.47

rimin

-1.44

stituted

-1.43

ourier

-1.42

cticut

-1.38

etition

-1.38

 Mexico

-1.36

POSITIVE LOGITS

±

2.19

¢

2.03

cliffe

1.76

ius

1.76

¦

1.71

¡

1.64

croft

1.63

giving

1.58

 Majesty

1.49

 axes

1.47

Activations Density 0.123%

titles or honorifics preceding names or entities

No Comments

No Known Activations