INDEX

Explanations

turn or reversal

The neuron is primarily sensitive to non-word symbols and numeric or technical tokens—e.g. digits, punctuation marks, mathematical symbols, and unit notations.

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

GRAM

-0.82

 рав

-0.76

کار

-0.73

 ספר

-0.72

皓

-0.71

underline

-0.71

 поко

-0.71

taş

-0.70

offset

-0.70

whatsapp

-0.70

POSITIVE LOGITS

 turnaround

2.14

 reversal

2.03

 volte

1.88

 reversing

1.69

 flip

1.51

 revers

1.49

 reverses

1.46

1.45

 reversed

1.45

 turn

1.42

Activations Density 0.023%