INDEX

Explanations

emotional states and consequences

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

ÂĢÂĢ

-0.10

TRGL

-0.10

ï¼Įå°±æĺ¯

-0.10

ceptar

-0.10

<|begin_of_text|>

-0.10

@nate

-0.09

URLException

-0.09

 ÑĩÑĤÐ¾Ð±Ñĭ

-0.09

BOOLE

-0.09

odash

-0.09

POSITIVE LOGITS

 especially

0.15

 because

0.14

especially

0.13

but

0.12

 although

0.12

 Especially

0.11

because

0.11

 despite

0.11

åıĬåħ¶

0.10

 considering

0.10

Activations Density 0.094%