INDEX

Explanations

demonstratives like 'this' and 'these'

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

-ÑĤÐ°ÐºÐ¸

-0.10

Ä±rak

-0.10

 Ø¢ÙĨÚĨÙĩ

-0.09

 bunch

-0.09

vester

-0.09

ãģĭãģªãģĦ

-0.09

à¹Ģà¸Ĥà¸²

-0.09

ocaly

-0.09

 Ð»Ð¸Ð±Ð¾

-0.08

POSITIVE LOGITS

 this

0.17

 these

0.16

ï¼īãģ¯

0.14

this

0.14

ients

0.13

äººãģ¯

0.13

any

0.11

 such

0.11

 most

0.11

these

0.11

Activations Density 0.238%