INDEX

Explanations

is

New Auto-Interp

Configuration

Dataset (Dashboard)

Various

Features

131,072

Data Type

float32

Hook Name

blocks.27.hook_resid_post

Architecture

standard

Context Size

1,024

Dataset

monology/pile-uncopyrighted

Embeds

IFrame

Link

Negative Logits

vice

-0.06

calling

-0.06

 prosecutors

-0.06

quite

-0.06

isters

-0.06

icts

-0.06

belum

-0.05

 πα

-0.05

myp

-0.05

 transformations

-0.05

POSITIVE LOGITS

 deutsche

0.07

э

0.07

BOSE

0.07

-campus

0.06

 Suche

0.06

ophe

0.06

.forRoot

0.06

 typingsJapgolly

0.06

SW

0.06

*>(&

0.06

Activations Density 0.027%

No Known Activations