INDEX

Explanations

appeal or resonate

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

 Hancock

-0.10

awk

-0.09

anga

-0.09

 McInt

-0.09

ADM

-0.09

ãģ£

-0.09

asn

-0.09

 aroused

-0.08

ze

-0.08

unm

-0.08

POSITIVE LOGITS

 appeal

0.47

 Appeal

0.39

 appeals

0.35

 appealed

0.29

 Appeals

0.28

appe

0.28

Appe

0.27

 appealing

0.26

 reson

0.25

 appe

0.19

Activations Density 0.173%