INDEX

Explanations

preparing for possibilities

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

deÅŁ

-0.10

 contemporary

-0.10

amage

-0.09

cot

-0.09

ialis

-0.09

Cot

-0.09

ATALOG

-0.09

nitÅĻ

-0.08

omanip

-0.08

POSITIVE LOGITS

 possibility

0.25

 worst

0.23

 possible

0.23

 conting

0.20

 possibilities

0.19

 eventual

0.19

 Worst

0.18

possible

0.18

 Ð²Ð¾Ð·Ð¼Ð¾Ð¶

0.17

 Possible

0.17

Activations Density 0.038%