INDEX

Explanations

reasons why

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

upro

-0.09

ensa

-0.09

Enlarge

-0.08

aylight

-0.08

 "{\"

-0.08

styleType

-0.08

antha

-0.08

äºŃ

-0.08

chez

-0.08

icter

-0.08

POSITIVE LOGITS

 writ

0.13

 Adjustment

0.10

 coun

0.10

 Aggregate

0.10

 somewhat

0.10

 official

0.10

 policy

0.10

 given

0.10

 dimensions

0.10

 actors

0.10

Activations Density 0.043%