INDEX

Explanations

definition of

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

izza

-0.10

chie

-0.10

uba

-0.09

dbContext

-0.09

 ÙħØ¶

-0.09

@nate

-0.08

[Byte

-0.08

choices

-0.08

annis

-0.08

POSITIVE LOGITS

 success

0.16

0.14

 notions

0.13

ä»Ģä¹Ī

0.12

 terms

0.12

 what

0.12

 boundaries

0.12

 notion

0.11

 borders

0.11

success

0.10

Activations Density 0.155%