INDEX

Explanations

Little followed by specific words

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

iers

-0.11

hoe

-0.10

Ð¾Ð³Ð¾Ð´

-0.10

ieren

-0.09

 Anders

-0.09

asl

-0.09

res

-0.09

 drafts

-0.09

ensch

-0.09

gne

-0.09

POSITIVE LOGITS

-known

0.21

st

0.20

bit

0.19

_endian

0.18

Endian

0.17

_ENDIAN

0.17

 league

0.17

-used

0.15

-bit

0.15

Bits

0.15

Activations Density 0.021%