INDEX

Explanations

offensie comment

This neuron detects rare/long subword tokens (uncommon or technical pieces such as medical measurements, drug names, and other unusual multi-character fragments).

New Auto-Interp

Configuration

Prompts (Dashboard)

238,145 prompts, 512 tokens each

Dataset (Dashboard)

lmsys + oasst1

Embeds

IFrame

Link

Not in Any Lists

Negative Logits

0.13

to

0.13

as

0.13

on

0.11

if

0.11

 with

0.11

at

0.10

 Python

0.10

is

0.10

 after

0.10

POSITIVE LOGITS

Ꭸ

0.12

徂

0.11

inghouse

0.11

荭

0.11

stown

0.11

룔

0.10

酈

0.10

 здравоохранения

0.10

潁

0.10

neyland

0.10

Activations Density 2.626%