INDEX

Explanations

with a presence or focus

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

ughter

-0.10

 Dome

-0.09

å¯¾

-0.09

 whopping

-0.09

itud

-0.08

ue

-0.08

 SPDX

-0.08

arez

-0.08

hci

-0.08

eker

-0.08

POSITIVE LOGITS

 roots

0.25

roots

0.19

 whom

0.16

 Roots

0.16

 ties

0.15

standing

0.14

 interests

0.13

stood

0.13

olding

0.12

 regard

0.12

Activations Density 0.069%