INDEX

Explanations

comment markers before code elements

The neuron is picking out numeric literals when they appear inside comment text.

New Auto-Interp

Configuration

Prompts (Dashboard)

24,576 prompts, 128 tokens each

Dataset (Dashboard)

monology/pile-uncopyrighted

Embeds

IFrame

Link

Negative Logits

 hâte

-1.16

 hés

-1.15

 considér

-1.12

いろんな

-1.09

 negocia

-1.08

感じです

-1.08

 célè

-1.07

 сьогодні

-1.05

 ſame

-1.04

ようになる

-1.04

POSITIVE LOGITS

 until

1.15

 from

1.13

for

1.10

re

1.10

ïde

1.07

on

1.03

 using

1.01

at

1.01

 after

0.99

0.98

Activations Density 0.000%