INDEX

Explanations

could

This neuron specializes in spotting speculative or hypothetical constructions, especially modal “could … be” or “one could …” phrases expressing possibility.

New Auto-Interp

Configuration

Prompts (Dashboard)

392,802 prompts, 256 tokens each

Dataset (Dashboard)

monology/pile-uncopyrighted

Embeds

IFrame

Link

Not in Any Lists

Negative Logits

ıt

0.86

kan

0.64

})`;

0.63

md

0.58

াইন

0.57

겠다

0.57

mdan

0.57

 باذن

0.57

ces

0.57

 segítség

0.56

POSITIVE LOGITS

Could

3.30

 Could

3.27

 could

3.16

could

3.16

 COULD

3.13

 могли

2.53

 pourrait

2.39

 pourraient

2.39

 podrían

2.38

 могла

2.34

Activations Density 0.472%