INDEX

Explanations

explaining the purpose

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

elah

-0.10

ver

-0.10

 Burk

-0.10

ritz

-0.10

-0.09

uzz

-0.09

upy

-0.09

ãĢĢV

-0.08

CHA

-0.08

POSITIVE LOGITS

 purpose

0.24

Purpose

0.20

 Purpose

0.19

purpose

0.18

-purpose

0.16

 PURPOSE

0.15

urpose

0.15

 intent

0.15

 Intent

0.14

 functionality

0.14

Activations Density 0.088%