INDEX

Explanations

access to actions

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

adro

-0.11

 Allow

-0.10

åħģ

-0.10

 consec

-0.10

appable

-0.09

 allow

-0.09

 enduring

-0.09

utherford

-0.09

 binds

-0.09

POSITIVE LOGITS

 access

0.18

 accesses

0.15

 exercise

0.14

 accessing

0.14

 express

0.13

 Exercise

0.12

access

0.12

Exercise

0.12

Access

0.12

exercise

0.11

Activations Density 0.076%