INDEX

Explanations

plans, purposes, pleasure

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

 Hayward

-0.11

 Guth

-0.09

aby

-0.09

èī¯

-0.09

 Perf

-0.09

 Harm

-0.09

 pinch

-0.09

 Leonard

-0.09

rances

-0.09

á»ĭ

-0.08

POSITIVE LOGITS

 benefit

0.45

 sake

0.36

 convenience

0.30

 Benefit

0.30

 purposes

0.28

enefit

0.25

 enjoyment

0.24

 pleasure

0.23

 amusement

0.23

benef

0.21

Activations Density 0.086%