INDEX

Explanations

<|start_header_id|>user<|end_header_id|>

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

.scalablytyped

-0.15

ëį°ìĿ´íĬ¸

-0.10

alion

-0.09

kli

-0.09

ï¼ŁãĢį\n\n

-0.09

roz

-0.08

Ø§ÙĦØ¥ÙĨØ¬ÙĦÙĬØ²ÙĬØ©

-0.08

gii

-0.08

.newBuilder

-0.07

ÑģÑĤÐ°ÑĤÐ¸

-0.07

POSITIVE LOGITS

chu

0.10

âĶ

0.08

 ranges

0.08

 diagn

0.08

avi

0.07

/mit

0.07

 ÄĲo

0.07

Ģ

0.07

 aforementioned

0.07

WHO

0.07

Activations Density 0.040%