INDEX

Explanations

fair and balanced manner

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

 decency

-0.11

Ð½Ð°Ð¼

-0.10

dÃ¼l

-0.09

elson

-0.09

 wholes

-0.09

 whirl

-0.09

 payload

-0.09

 motion

-0.09

nio

-0.08

POSITIVE LOGITS

 objective

0.35

 neutral

0.30

 impartial

0.30

objective

0.29

 fair

0.29

 neutrality

0.27

 neutr

0.27

 Neutral

0.26

 Objective

0.26

Neutral

0.26

Activations Density 0.176%