INDEX

Explanations

transition to conclusion

The neuron is primarily detecting the discourse connective “Thus,” (and similar conclusion‐marking transition words).

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

of

-2.41

 эту

-1.88

 هذه

-1.80

 queste

-1.71

どのような

-1.65

 only

-1.64

 этой

-1.59

ここから

-1.58

these

-1.56

 этот

-1.56

POSITIVE LOGITS

赁

1.91

卨

1.91

 '/')

1.67

ԝ

1.65

桕

1.61

愎

1.60

om

1.59

Skocz

1.56

Aacute

1.56

 ',',

1.54

Activations Density 0.025%