INDEX

Explanations

fill followed by a word

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

 Kral

-0.11

ÂŃt

-0.10

athan

-0.10

achuset

-0.09

ephir

-0.09

å¾Ĵ

-0.09

 pitch

-0.09

athon

-0.09

osphere

-0.09

aveled

-0.08

POSITIVE LOGITS

ings

0.18

fill

0.16

up

0.15

ter

0.15

-in

0.14

(fill

0.14

åħħ

0.14

ip

0.13

tering

0.13

ers

0.12

Activations Density 0.028%