INDEX

Explanations

removing artificial

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

 Deutsch

-0.10

 axial

-0.10

 Briggs

-0.10

 deform

-0.10

 bent

-0.10

 bends

-0.10

unm

-0.09

ä¸

-0.09

 Operation

-0.09

reeze

-0.09

POSITIVE LOGITS

 plastic

0.19

 strain

0.17

 Plastic

0.16

å¡ĳ

0.16

 Damage

0.16

 damage

0.15

Damage

0.15

 material

0.15

 strains

0.15

 constit

0.15

Activations Density 0.015%