INDEX

Explanations

Despite followed by contrasting phrases

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

rena

-0.10

amon

-0.10

-0.09

uga

-0.09

yon

-0.08

thon

-0.08

antha

-0.08

ije

-0.08

ookie

-0.08

itre

-0.08

POSITIVE LOGITS

 these

0.24

è¿ĻäºĽ

0.19

these

0.18

 These

0.16

 THESE

0.15

These

0.15

 ÑįÑĤÐ¸Ñħ

0.13

ses

0.13

 ìĿ´ëŁ¬íķľ

0.13

 estas

0.13

Activations Density 0.008%