INDEX

Explanations

demonstrating consequences of

The neuron selectively activates on the high‐salience result‐reporting phrases in scientific abstracts—especially strong verbs like “find,” “show,” or “demonstrate” (often followed by “that”) that introduce key findings.

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

 mélanger

-1.23

Temperatura

-1.22

 cliquer

-1.20

Vielleicht

-1.19

ֶּ

-1.18

 télécharger

-1.16

 consommer

-1.13

rufen

-1.09

 utiliser

-1.08

 réessayer

-1.08

POSITIVE LOGITS

 effects

1.22

 different

1.19

$\

1.18

鏖

1.15

 diverse

1.15

会出现

1.10

 even

1.07

for

1.07

 change

1.05

AndFlush

1.05

Activations Density 0.110%