concepts following the

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

Negative Logits

uin

-0.10

Ð½Ð°Ð´Ð»ÐµÐ¶

-0.08

egers

-0.08

 Provid

-0.08

ebp

-0.08

gcd

-0.08

ithmetic

-0.08

POSITIVE LOGITS

 question

0.28

 answer

0.28

answer

0.20

 exact

0.18

 extent

0.18

question

0.18

 notion

0.17

 jury

0.17

 degree

0.16

 answers

0.16

Activations Density 0.087%

No Known Activations