INDEX

Explanations

The neuron spikes on question‐mark tokens (and related punctuation), effectively marking interrogative sentences.

New Auto-Interp

Configuration

Prompts (Dashboard)

392,802 prompts, 256 tokens each

Dataset (Dashboard)

monology/pile-uncopyrighted

Embeds

IFrame

Link

Not in Any Lists

No Comments

Negative Logits

but

1.25

 nhưng

1.16

但在

1.16

但也

1.15

but

1.13

 tetapi

1.12

 ngunit

1.10

 mutta

1.08

しましたが

1.08

但

1.07

POSITIVE LOGITS

 🤔

1.23

🤔

1.16

 Answer

1.09

 Certainly

1.05

 Pues

1.02

Or

0.99

 Answers

0.97

 Ведь

0.96

 Atau

0.93

 Nope

0.88

Activations Density 0.306%