INDEX

Explanations

worsening or persistent negative conditions

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

XB

-0.10

ved

-0.09

nelle

-0.09

 lump

-0.09

ablo

-0.09

 Completion

-0.09

 Brett

-0.09

 suffers

-0.08

 damaging

-0.08

POSITIVE LOGITS

 severity

0.21

 severe

0.21

ä¸¥

0.18

severity

0.17

 Severity

0.17

 sever

0.16

 worse

0.15

Severity

0.15

 worst

0.13

 acute

0.12

Activations Density 0.067%