INDEX

Explanations

connection is strong

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

ulur

-0.09

coli

-0.09

inka

-0.09

mue

-0.09

alli

-0.08

FromBody

-0.08

 underst

-0.08

 Welt

-0.08

cac

-0.08

undef

-0.08

POSITIVE LOGITS

yes

0.12

 numer

0.11

 harmon

0.10

Tar

0.10

patibility

0.10

 blinds

0.10

blocks

0.10

Cap

0.09

 energies

0.09

 Numer

0.09

Activations Density 0.042%