INDEX

Explanations

observed effect or superficial aspects

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

 Pier

-0.10

Scenario

-0.10

 Futures

-0.09

 Scenario

-0.09

abet

-0.09

 scenarios

-0.09

 organs

-0.09

æĥħ

-0.09

 scenario

-0.09

-opacity

-0.09

POSITIVE LOGITS

 belief

0.22

 superst

0.21

 beliefs

0.19

belief

0.18

çĽ¸ä¿¡

0.18

 placebo

0.17

 believed

0.16

 efficacy

0.16

 effectiveness

0.16

 believes

0.15

Activations Density 0.177%