INDEX

Explanations

confirm understanding

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

Kre

-0.11

atego

-0.10

 McGregor

-0.09

à¤Ĺà¤°

-0.09

 friend

-0.09

 flat

-0.09

anga

-0.09

teg

-0.09

aque

-0.08

_goto

-0.08

POSITIVE LOGITS

ack

0.22

 understood

0.20

 understanding

0.20

 acknowledge

0.20

Acknowled

0.20

 acknowledged

0.20

Ack

0.19

Ack

0.19

ACK

0.18

çĲĨè§£

0.18

Activations Density 0.138%

No Known Activations