INDEX

Explanations

find or approach

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

å«Į

-0.10

dab

-0.10

entifier

-0.09

 usher

-0.09

rep

-0.09

 vain

-0.09

 Mock

-0.08

 RECEIVE

-0.08

unt

-0.08

 MOCK

-0.08

POSITIVE LOGITS

 approach

0.16

 approaches

0.16

 approached

0.16

appro

0.15

 corner

0.14

 waited

0.14

 Approach

0.13

 wait

0.13

 proposition

0.13

 approaching

0.13

Activations Density 0.095%