INDEX

Explanations

failure or prerequisite

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

 lust

-0.09

 Ding

-0.09

Esc

-0.09

 ØµØ§ØŃØ¨

-0.09

esc

-0.09

unch

-0.09

unik

-0.09

utes

-0.09

Kou

-0.08

 eagle

-0.08

POSITIVE LOGITS

 trying

0.19

 somewhere

0.19

 either

0.17

 attempted

0.16

 attempting

0.15

trying

0.14

 Trying

0.14

 tries

0.14

 attempt

0.14

either

0.13

Activations Density 0.086%