INDEX

Explanations

normal way then specific alternative

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

ingle

-0.09

abl

-0.09

phin

-0.09

.ErrorCode

-0.08

 Bacon

-0.08

 precinct

-0.08

quiz

-0.08

 trib

-0.08

 Harden

-0.08

urg

-0.08

POSITIVE LOGITS

 normal

0.24

normal

0.19

 Normal

0.17

æŃ£å¸¸

0.17

 regular

0.16

 conventional

0.16

 standard

0.16

Normal

0.15

 response

0.15

.normal

0.13

Activations Density 0.051%