INDEX

Explanations

code structure and identifiers

The neuron strongly activates on code identifier tokens (variable and parameter names) in programming snippets.

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

 Pří

-1.27

dekat

-1.10

 住宅

-1.09

筈

-1.08

 hadden

-1.07

слуги

-1.07

 sabias

-1.05

ようですが

-1.05

gents

-1.05

 felf

-1.04

POSITIVE LOGITS

 first

1.32

1.20

。。。。

1.18

ard

1.11

bo

1.10

\_

1.09

 даже

1.05

 dieselbe

1.04

ancy

1.03

Activations Density 0.010%