INDEX

Explanations

holding and restraining

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

//{{

-0.12

 uncont

-0.10

Deferred

-0.09

 Shield

-0.09

Pis

-0.09

EditMode

-0.09

 ramp

-0.09

ãĥ¬ãĤ¤

-0.09

 detach

-0.09

visa

-0.08

POSITIVE LOGITS

 secure

0.20

 hold

0.18

 immobil

0.17

çī

0.16

 HOLD

0.16

 holds

0.16

hold

0.15

 Holds

0.15

imm

0.15

 held

0.15

Activations Density 0.053%