INDEX

Explanations

concluding phrases

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

'gc

-0.12

 this

-0.10

 à¸ģà¸£à¸ģ

-0.10

ren

-0.09

atts

-0.09

 podrob

-0.08

 initWithStyle

-0.08

oku

-0.08

pbs

-0.08

UsageId

-0.08

POSITIVE LOGITS

 although

0.20

èĻ½çĦ¶

0.18

although

0.17

 Although

0.17

Although

0.15

èĻ½

0.15

 ÑħÐ¾ÑĤÑı

0.14

 While

0.14

 while

0.14

 answer

0.14

Activations Density 0.008%