INDEX

Explanations

pointing and remarking

New Auto-Interp

Configuration

Prompts (Dashboard)

16,384 prompts, 128 tokens each

Dataset (Dashboard)

monology/pile-uncopyrighted

Embeds

IFrame

Link

Not in Any Lists

No Comments

Negative Logits

 thanking

-0.13

 apology

-0.12

 apologize

-0.12

 thanked

-0.12

 apologized

-0.11

 apologies

-0.11

 apolog

-0.11

 thank

-0.10

 Claud

-0.10

SOL

-0.10

POSITIVE LOGITS

 pointing

0.23

nud

0.20

 pointed

0.17

 alerted

0.16

 commentary

0.15

 Point

0.14

 whispered

0.14

 spotting

0.14

 remarked

0.14

 identification

0.14

Activations Density 0.122%