INDEX

Explanations

for the first time

The neuron strongly activates on extremely common tokens (high‐frequency, generic words and symbols) rather than content-specific terms.

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

-0.87

Lab

-0.87

Bill

-0.85

AM

-0.83

を中心に

-0.81

-0.80

mathrm

-0.78

enosis

-0.78

Ed

-0.76

 登録

-0.76

POSITIVE LOGITS

 впервые

5.25

初めて

3.94

 erstmals

3.91

首次

3.38

 first

3.34

for

3.27

 lần

3.25

 впер

3.13

 popr

3.11

第一次

3.02

Activations Density 0.166%