INDEX

Explanations

questions about human nature

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

azen

-0.10

 Quinn

-0.09

ainen

-0.09

actionDate

-0.08

ï¼¾

-0.08

/INFO

-0.08

/DD

-0.08

Buffers

-0.08

aze

-0.08

ç¿

-0.08

POSITIVE LOGITS

 whether

0.22

 question

0.20

whether

0.18

 Whether

0.17

 topic

0.17

 issue

0.17

 WHETHER

0.17

 opinions

0.17

æĺ¯åĲ¦

0.16

 debate

0.15

Activations Density 0.195%