INDEX

Explanations

asking for context or

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

_ctx

-0.10

axon

-0.09

andex

-0.09

ÃŃrk

-0.09

contents

-0.09

 Kristen

-0.09

_COOKIE

-0.09

áº£nh

-0.08

aspect

-0.08

à¤ķà¤¨

-0.08

POSITIVE LOGITS

ually

0.23

ual

0.21

uality

0.17

UAL

0.16

 surrounding

0.14

MenuStrip

0.13

 information

0.13

uele

0.11

 clues

0.11

uale

0.11

Activations Density 0.027%