INDEX

Explanations

possessing qualities or attributes

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

wy

-0.09

Jog

-0.09

stra

-0.09

awei

-0.09

 automat

-0.09

 Intr

-0.08

 Neuroscience

-0.08

 Estr

-0.08

 unpack

-0.08

 Kral

-0.08

POSITIVE LOGITS

 ability

0.19

 capabilities

0.18

 capability

0.17

 knowledge

0.16

 personality

0.15

 abilities

0.14

 access

0.14

 memory

0.14

 intelligence

0.13

 Knowledge

0.13

Activations Density 0.141%