INDEX

Explanations

limitations and struggles

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

 Forced

-0.09

ANGO

-0.09

ective

-0.09

 zaten

-0.09

APS

-0.08

stdarg

-0.08

anka

-0.08

aphore

-0.08

PFN

-0.08

 inde

-0.08

POSITIVE LOGITS

 suffer

0.20

 struggles

0.20

 struggle

0.18

uffers

0.18

 suffers

0.18

 require

0.17

uffer

0.16

 suffered

0.16

 requires

0.15

 struggled

0.15

Activations Density 0.039%