INDEX

Explanations

purpose and context

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

for

-0.11

ï¼¿ï¼¿ï¼¿ï¼¿

-0.09

urn

-0.09

 dafÃ¼r

-0.09

oyal

-0.09

QUARE

-0.09

kola

-0.09

 Hoff

-0.09

 Geile

-0.08

POSITIVE LOGITS

 sake

0.42

 purposes

0.40

æĿ¥è¯´

0.17

 reasons

0.13

 purpose

0.12

 ÙĪÙĦ

0.12

 ÑģÐ»ÑĥÑĩ

0.11

 PURPOSE

0.11

ouri

0.10

pur

0.10

Activations Density 0.394%