INDEX

Explanations

process or steps

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

Negative Logits

peri

-0.11

zo

-0.11

ihan

-0.10

iro

-0.10

uat

-0.09

ician

-0.08

osit

-0.08

;;;;

-0.08

 Creed

-0.08

ily

-0.08

POSITIVE LOGITS

 starting

0.22

 beginning

0.17

starting

0.17

 Starting

0.17

Starting

0.16

 each

0.15

each

0.13

 Beginning

0.13

å¼Ģå§ĭ

0.12

 starts

0.12

Activations Density 0.024%

No Known Activations