INDEX

Explanations

concrete and practical

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

Negative Logits

 ÑĦÑĥÐ½Ð´Ð°

-0.10

ner

-0.10

 sophistication

-0.10

ichen

-0.09

INTR

-0.09

AbsolutePath

-0.09

 intellectual

-0.09

 intellect

-0.08

 Intr

-0.08

 Matth

-0.08

POSITIVE LOGITS

 practical

0.29

 Practical

0.24

 concrete

0.20

 Concrete

0.19

 exper

0.17

 empirical

0.17

gut

0.16

Concrete

0.16

 simpler

0.15

 physical

0.15

Activations Density 0.146%

No Known Activations