INDEX

Explanations

code and error locations

The neuron is detecting numeric tokens (digits or multi‐digit numbers), such as line numbers, literals, or positional indices.

New Auto-Interp

Configuration

Prompts (Dashboard)

24,576 prompts, 128 tokens each

Dataset (Dashboard)

monology/pile-uncopyrighted

Embeds

IFrame

Link

Not in Any Lists

Negative Logits

 both

-1.28

of

-1.28

 either

-1.23

all

-1.22

his

-1.22

for

-1.21

him

-1.11

 where

-1.09

-1.07

💯

-1.06

POSITIVE LOGITS

Garten

1.38

Zdrav

1.26

izient

1.24

 продолжает

1.23

 écl

1.23

itored

1.20

 colorear

1.19

Probl

1.16

Reak

1.16

Kval

1.16

Activations Density 0.003%