INDEX

Explanations

assuming you

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

ehr

-0.10

inspace

-0.10

holm

-0.09

 depending

-0.09

å«

-0.09

 subur

-0.09

 Depending

-0.09

expects

-0.09

urovision

-0.09

Dil

-0.08

POSITIVE LOGITS

 perfect

0.17

 ideal

0.14

perfect

0.13

 Perfect

0.13

 simpl

0.13

 assume

0.13

 PERF

0.12

Perfect

0.12

 linear

0.12

 constant

0.12

Activations Density 0.181%