INDEX

Explanations

determines input after "the", "active", "your", "two", "which", "it"

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

¦æĥħ

-0.13

¶Į

-0.11

łéĻ¤

-0.10

.Dictionary

-0.09

Å¡ker

-0.09

įng

-0.09

ķãĤĵ

-0.09

ogui

-0.09

³ç´°

-0.09

 dara

-0.09

POSITIVE LOGITS

 original

0.14

 earlier

0.13

 aforementioned

0.13

original

0.12

 initial

0.12

 previously

0.11

 originally

0.11

 input

0.10

åĪļæīį

0.10

 initially

0.10

Activations Density 0.038%