INDEX

Explanations

leading to or perpetuating outcomes

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

 undermining

-0.14

 undermined

-0.13

 undermine

-0.13

 undermines

-0.12

ambi

-0.10

 jeopard

-0.10

 crippled

-0.09

rost

-0.09

hoo

-0.09

 Ä°mpar

-0.09

POSITIVE LOGITS

 creates

0.16

 create

0.16

 perpet

0.15

 creating

0.14

create

0.13

 ÑģÐ¾Ð·Ð´

0.13

Create

0.12

 Create

0.12

 legit

0.12

 dáº«n

0.12

Activations Density 0.073%