INDEX

Explanations

end with specified item

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

 prefixes

-0.12

 prepend

-0.11

 prefixed

-0.11

 prefix

-0.11

 pref

-0.11

 Prefix

-0.11

 preamble

-0.11

ulan

-0.10

('('

-0.10

 Ranger

-0.10

POSITIVE LOGITS

 trailing

0.16

å°¾

0.14

suffix

0.14

(suffix

0.13

 ending

0.13

 suffix

0.13

_suffix

0.12

à¸¥à¸ĩà¸Ĺ

0.12

 trails

0.12

ending

0.12

Activations Density 0.116%