INDEX

Explanations

elements described with

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

ITHER

-0.13

ither

-0.12

 withholding

-0.11

 withhold

-0.11

 withheld

-0.10

_using

-0.09

 watt

-0.08

inho

-0.08

tober

-0.08

 abruptly

-0.08

POSITIVE LOGITS

 with

0.35

 vá»Ľi

0.31

 dengan

0.28

 avec

0.28

with

0.24

\twith

0.22

 withd

0.16

avec

0.14

 Ø¨Ø§

0.14

 With

0.13

Activations Density 0.234%