INDEX

Explanations

expressing inability or apology

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

alus

-0.10

imat

-0.10

iglia

-0.09

 exagger

-0.09

 Thoughts

-0.09

tip

-0.09

parsers

-0.08

:.:

-0.08

 Disclaimer

-0.08

 ignorance

-0.08

POSITIVE LOGITS

 earlier

0.21

 previously

0.16

 Earlier

0.13

Earlier

0.12

 previous

0.12

ä¹ĭåīį

0.12

 noticed

0.11

 originally

0.11

 Previously

0.11

 dÅĻÃŃve

0.10

Activations Density 0.048%