INDEX

Explanations

can have impact, can be made

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

rani

-0.10

tach

-0.09

ä¹Łä¸į

-0.09

 mant

-0.09

acc

-0.09

åŃĿ

-0.09

_env

-0.09

 fits

-0.08

exampleInput

-0.08

inati

-0.08

POSITIVE LOGITS

 difference

0.22

difference

0.21

åĨ³å®ļ

0.20

 decide

0.19

 Difference

0.19

 decides

0.19

Difference

0.18

 deciding

0.17

 either

0.17

 determines

0.16

Activations Density 0.091%