INDEX

Explanations

can get, check, cook, study

The neuron strongly activates on modal and auxiliary constructions—especially “can,” “able,” “will,” and the infinitive marker “to”—marking phrases that express ability, possibility, or intent.

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

𝟑

-1.02

Bereich

-1.02

토

-0.98

 stratég

-0.96

gnügen

-0.94

 фак

-0.93

ffilm

-0.93

Når

-0.92

 frasco

-0.91

JAXB

-0.91

POSITIVE LOGITS

 grasp

1.06

 meanwhile

1.00

 aware

0.97

 réussite

0.93

 poursuite

0.91

 conduct

0.87

 Meanwhile

0.87

このまま

0.87

 regards

0.86

 involve

0.86

Activations Density 0.173%