INDEX

Explanations

deduce, deduced, vengeance

The neuron tends to fire on rare or domain-specific tokens—uncommon words, proper names, technical jargon, or low-frequency morphemes.

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

 étab

-2.41

遑

-2.30

苦笑

-2.30

-2.25

蘗

-2.22



-2.20

 exceptionally

-2.19

on

-2.19

齟

-2.17

⚌

-2.17

POSITIVE LOGITS

ſſel

3.78

↵↵

2.78

 melden

2.48

</td>

2.41

琑

2.39

 podem

2.33

Of

2.31

<td>

2.28

orail

2.28

 είναι

2.27

Activations Density 0.055%