INDEX

Explanations

under human excess

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

 recycled

-0.10

ä¹¾

-0.10

forman

-0.09

 Recycling

-0.09

 depr

-0.08

zza

-0.08

 recycle

-0.08

 health

-0.08

 deprived

-0.08

 Supply

-0.08

POSITIVE LOGITS

 consumption

0.21

 consum

0.19

 poll

0.16

 Consumption

0.16

 resource

0.16

 consume

0.16

 industrial

0.15

 Consum

0.15

def

0.15

SUV

0.14

Activations Density 0.084%

No Known Activations