INDEX

Explanations

introduces clauses

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

ograd

-0.09

apel

-0.09

 ï¼¼:

-0.09

Ä±s

-0.08

 tÃ´i

-0.08

stalk

-0.08

ixin

-0.08

 this

-0.08

Regs

-0.08

 indeed

-0.08

POSITIVE LOGITS

 since

0.20

we

0.18

Since

0.17

 Since

0.17

since

0.17

 because

0.16

æĹ¢çĦ¶

0.15

 Ð¿Ð¾ÑģÐºÐ¾Ð»ÑĮÐºÑĥ

0.15

æĪĳä»¬

0.15

 knows

0.14

Activations Density 0.081%