INDEX

Explanations

The neuron fires on very short (mostly single-letter) identifier tokens in code.

mathematical notation and code

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

<bos>

-1.23

 Fällen

-1.19

かというと

-1.07

 dieselbe

-1.06

жом

-1.04

}$

-1.02

 Hinsicht

-0.98

 meistens

-0.98

でしょうね

-0.98

 Еще

-0.96

POSITIVE LOGITS

の方も

1.26

 déposer

1.19

になれる

1.16

 soin

1.11

and

1.10

 onderdelen

1.09

("&

1.08

 nettoyer

1.08

 Jalisco

1.07

».

1.05

Activations Density 0.011%