INDEX

Explanations

safe to say

This neuron detects hedging or qualified assertion phrases (e.g. “safe to say,” “fair to say”) that introduce generalizations.

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

to

-2.19

You

-1.74

 Finally

-1.61

 Here

-1.59

 sorts

-1.52

So

-1.45

use

-1.40

zzarella

-1.39

 Test

-1.36

 Build

-1.35

POSITIVE LOGITS

 that

1.71

 kesel

1.64

atteindre

1.52

 veta

1.50

Веду

1.47

 kemer

1.47

وئ

1.46

↓↓↓

1.45

 pakan

1.45

 kaos

1.44

Activations Density 0.009%