INDEX

Explanations

the [entity/concept]

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

noinspection

-0.10

llib

-0.09

aston

-0.09

 Midlands

-0.08

 Hancock

-0.08

Ïį

-0.08

-Cs

-0.08

 Jays

-0.08

 Reich

-0.08

 yere

-0.08

POSITIVE LOGITS

 United

0.15

United

0.11

ses

0.09

 Politico

0.09

 Barnett

0.09

 Federal

0.09

bah

0.09

US

0.09

 Narendra

0.08

Activations Density 0.211%