INDEX

Explanations

list items and structure

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

 Went

-0.11

drs

-0.09

epar

-0.09

bjerg

-0.09

reh

-0.08

 Chew

-0.08

 Shame

-0.08

 bombard

-0.08

 Ezra

-0.08

--------------------------------------------------------------------------\n

-0.08

POSITIVE LOGITS

eliac

0.09

Gur

0.08

 Blaze

0.08

@Setter

0.08

licht

0.08

unded

0.08

 crunch

0.08

_fw

0.08

ãĢĤãĢĤ\n\n

0.08

 eventual

0.08

Activations Density 0.058%