INDEX

Explanations

Code/URLsReason: MAX_ACTIVATING_TOKENS are single letters, and TOP_ACTIVATING_TEXTS commonly contain variable/parameter names in code snippets or URLs

New Auto-Interp

Configuration

Dataset (Dashboard)

Various

Features

131,072

Data Type

float32

Hook Name

blocks.19.hook_resid_post

Architecture

standard

Context Size

1,024

Dataset

monology/pile-uncopyrighted

Embeds

IFrame

Link

Negative Logits

DATA

-0.08

 adlandır

-0.07

 mentors

-0.07

	values

-0.07

名

-0.07

调整

-0.07

iqu

-0.07

 funding

-0.06

>.↵↵

-0.06

 receptors

-0.06

POSITIVE LOGITS

_MethodInfo

0.07

 Crest

0.06

.CurrentCulture

0.06

 recess

0.06

/export

0.06

 reconoc

0.06

 Caucas

0.06

 يس

0.06

onec

0.06

[DllImport

0.06

Activations Density 0.145%

No Known Activations