INDEX

Explanations

variable declaration followed by assignment/check

The neuron fires strongly on exactly three‐character tokens (e.g. “err” or “url”).

New Auto-Interp

Configuration

Prompts (Dashboard)

392,802 prompts, 256 tokens each

Dataset (Dashboard)

monology/pile-uncopyrighted

Embeds

IFrame

Link

Not in Any Lists

Negative Logits

(((

0.69

 пя

0.64

 регистрации

0.63

 করিবে

0.62

lastval

0.61

 остан

0.59

 jego

0.58

((*

0.58

 jeśli

0.58

(((

0.57

POSITIVE LOGITS

 nicht

0.69

is

0.68

 ikke

0.65

 didn

0.64

 nisu

0.59

are

0.58

Ex

0.58

är

0.57

 Coaches

0.57

in

0.56

Activations Density 0.000%