INDEX

Explanations

cause or reason

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

 Ð¿Ð¾Ð±Ð°

-0.10

hani

-0.08

iem

-0.08

afia

-0.08

CHAN

-0.08

mut

-0.08

çĽĳåĲ¬é¡µéĿ¢

-0.08

imed

-0.08

 PROFITS

-0.08

(crate

-0.08

POSITIVE LOGITS

due

0.41

 because

0.31

due

0.30

 debido

0.26

Due

0.26

åĽłä¸º

0.25

because

0.25

 kvÅ¯li

0.25

Due

0.24

 wegen

0.24

Activations Density 0.128%