INDEX

Explanations

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

-0.10

 Hansen

-0.09

 Entrance

-0.09

 Tick

-0.08

Dro

-0.08

aghan

-0.08

Nas

-0.08

cha

-0.08

Alo

-0.08

|:

-0.08

POSITIVE LOGITS

 approach

0.56

 approached

0.53

 approaches

0.52

 Approach

0.50

appro

0.49

Appro

0.46

 approaching

0.45

 Appro

0.42

 appro

0.34

_appro

0.33

Activations Density 0.082%