INDEX

Explanations

accepting them unconditionally

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

 Sanity

-0.09

ucker

-0.09

 Playground

-0.09

 lives

-0.08

eggies

-0.08

aby

-0.08

 McCoy

-0.08

imo

-0.08

imp

-0.08

quam

-0.08

POSITIVE LOGITS

 exactly

0.16

warts

0.13

 accepts

0.13

 acceptance

0.13

accept

0.13

 Accept

0.13

 Exactly

0.12

Exactly

0.12

tal

0.11

actly

0.11

Activations Density 0.039%