INDEX

Explanations

you measure water

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

 Quadr

-0.12

 flushed

-0.09

 flushing

-0.09

gars

-0.09

 quadr

-0.09

 breathing

-0.08

gep

-0.08

 rake

-0.08

odata

-0.08

POSITIVE LOGITS

 water

0.15

 Fill

0.15

 Water

0.14

 filled

0.13

 fill

0.13

fillable

0.13

.fill

0.13

_fill

0.13

_FILL

0.12

 pouring

0.12

Activations Density 0.051%