INDEX

Explanations

answering questions

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

anus

-0.09

endon

-0.09

anged

-0.09

RS

-0.09

ne

-0.08

RS

-0.08

 tran

-0.08

 Cobb

-0.08

Hag

-0.08

POSITIVE LOGITS

 questions

0.21

 question

0.16

 Questions

0.15

questions

0.15

 Fragen

0.13

 Ð²Ð¾Ð¿ÑĢÐ¾Ñģ

0.13

 general

0.12

_questions

0.11

Questions

0.11

 Frage

0.11

Activations Density 0.078%