INDEX

Explanations

to treat

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

 amel

-0.10

avigator

-0.09

aldo

-0.09

 engineered

-0.09

abler

-0.09

emer

-0.08

YTE

-0.08

hun

-0.08

æıĲ

-0.08

POSITIVE LOGITS

 treat

0.45

 treatment

0.43

 treating

0.43

 behand

0.35

 Treat

0.34

 treats

0.34

 Treatment

0.33

 tratamiento

0.31

 treatments

0.31

 tratt

0.31

Activations Density 0.074%