INDEX

Explanations

compared to

The neuron detects degree‐ or magnitude‐expressing modifiers (adjectives/adverbs like “long,” “old,” “pretty,” “much,” “too [big],” “a lot,” etc.).

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

hamed

-0.92

上前

-0.89

 einander

-0.88

িয়

-0.87

 své

-0.85

 здесь

-0.84

gangenen

-0.81

 visse

-0.81

 with

-0.81

？！”

-0.79

POSITIVE LOGITS

for

1.88

 considering

1.54

 compared

1.22

 برای

1.16

banez

1.14

ebly

1.12

 surtout

1.08

Considering

1.07

rfan

1.04

 для

1.02

Activations Density 0.023%