INDEX

Explanations

pronoun followed by verb/state

New Auto-Interp

Top Features by Cosine Similarity

Configuration

Prompts (Dashboard)

10,000 prompts, 128 tokens each

Dataset (Dashboard)

lmsys/lmsys-chat-1m

Embeds

IFrame

Link

Not in Any Lists

No Comments

Negative Logits

 Brennan

-0.10

BERS

-0.09

ersen

-0.09

 extinct

-0.09

Couldn

-0.09

alten

-0.09

è

-0.09

-0.08

empre

-0.08

POSITIVE LOGITS

 originally

0.19

 ÄĳÃ£

0.16

 previously

0.16

æĽ¾

0.16

 telah

0.15

ve

0.14

 Originally

0.12

 olmuÅŁtur

0.12

’;ve

0.11

 earlier

0.11

Activations Density 0.163%