INDEX

Explanations

the following

The neuron is selectively activating on numeric literals and related tokens (e.g. seed indices, configuration values) in code‐like or configuration contexts.

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

by

-1.80

 保護

-1.23

 cordones

-1.07

 televisor

-1.06

>);

-1.06

 μὲν

-1.05

 увлаж

-1.04

ancier

-1.04

 Pokies

-1.02

----+

-1.02

POSITIVE LOGITS

 çünkü

1.27

küche

1.15

くらいで

1.13

<bos>

1.09

 한국

1.08

 forêts

1.07

 variétés

1.05

 كذلك

1.04

beeren

1.02

 لأنه

1.01

Activations Density 0.001%