INDEX

Explanations

MRT findings and German descriptions

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

 Sass

-0.11

 Samuel

-0.10

 Gund

-0.09

Mig

-0.09

 cigaret

-0.09

 beams

-0.08

Kut

-0.08

 Beam

-0.08

 Electro

-0.08

POSITIVE LOGITS

MR

0.29

 magnetic

0.27

 Magnetic

0.26

MR

0.25

ç£ģ

0.25

MRI

0.24

gad

0.22

MRI

0.22

 magnets

0.21

 magnet

0.20

Activations Density 0.027%