INDEX

Explanations

respect

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

å¿ł

-0.11

 Replay

-0.10

IGO

-0.09

ickle

-0.09

 plac

-0.09

esa

-0.09

affer

-0.09

 Robertson

-0.08

uffy

-0.08

 reconc

-0.08

POSITIVE LOGITS

 respect

0.57

respect

0.44

 Respect

0.43

å°Ĭ

0.42

 respects

0.35

 respected

0.33

 ÑĥÐ²

0.31

RES

0.28

-res

0.27

Res

0.26

Activations Density 0.092%