INDEX

Explanations

understanding ability judgment

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

 Dawson

-0.09

 vend

-0.08

 themselves

-0.08

Ð½Ð¾Ð´

-0.08

ãĢĢãĢĢ ãĢĢ ãĢĢ ãĢĢ

-0.08

ãĥŃãĥ¼

-0.08

/lgpl

-0.08

sono

-0.08

croft

-0.08

hlen

-0.08

POSITIVE LOGITS

 ability

0.24

 willingness

0.18

 Ability

0.17

Ability

0.16

ability

0.14

 feeling

0.14

 inability

0.13

endency

0.11

 able

0.11

 awareness

0.11

Activations Density 0.178%