INDEX

Explanations

directions

The neuron is picking out uncommon or “odd” subword tokens—things like numbers, punctuation clusters, or otherwise rare/irregular token pieces.

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

块

-1.05

♥♥

-1.04

日時

-1.00

ælp

-1.00

 intermédiaire

-1.00

 suivantes

-0.98

颔

-0.98

這

-0.98

 televisão

-0.97

masalah

-0.97

POSITIVE LOGITS

 Json

1.09

books

1.09

utenants

1.03

);

1.02

0.99

about

0.98

 Bluetooth

0.95

my

0.93

fare

0.93

Activations Density 0.025%