INDEX

Explanations

official and preferred sources

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

)(((

-0.09

zel

-0.09

 scri

-0.08

Pam

-0.08

 predecessor

-0.08

 reconc

-0.08

 rounded

-0.08

 domic

-0.08

RR

-0.08

 Content

-0.08

POSITIVE LOGITS

 sources

0.19

 reliable

0.16

 official

0.16

sources

0.15

 websites

0.15

 sites

0.14

official

0.14

 online

0.13

 Ð¸ÑģÑĤÐ¾Ñĩ

0.13

 news

0.13

Activations Density 0.065%