INDEX

Explanations

asking 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

licer

-0.10

antz

-0.10

HOW

-0.09

iaz

-0.08

abra

-0.08

uentes

-0.08

vg

-0.08

 WHAT

-0.08

reesome

-0.08

ifiable

-0.08

POSITIVE LOGITS

why

0.82

why

0.67

ä¸ºä»Ģä¹Ī

0.63

Why

0.63

 reasons

0.59

 reason

0.59

Why

0.58

 pourquoi

0.57

 Ð¿Ð¾ÑĩÐµÐ¼Ñĥ

0.48

WHY

0.47

Activations Density 0.381%