INDEX

Explanations

not perfect

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

 Cord

-0.10

 Devon

-0.09

 cord

-0.09

ccione

-0.08

 Cursors

-0.08

_elt

-0.08

 pros

-0.08

undra

-0.08

 tipped

-0.08

ewire

-0.08

POSITIVE LOGITS

inf

0.44

 perfect

0.36

perfect

0.28

 Perfect

0.27

Perfect

0.24

omn

0.24

å®Į

0.22

 PERF

0.22

 perfection

0.22

 immune

0.20

Activations Density 0.078%