INDEX

Explanations

asking for sum or value

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

 estim

-0.10

InputElement

-0.09

 Bender

-0.09

 plus

-0.09

acks

-0.08

Limits

-0.08

 ï½°

-0.08

limits

-0.08

 because

-0.08

Sud

-0.08

POSITIVE LOGITS

 highest

0.16

 majority

0.12

 recorded

0.11

 almost

0.11

 Highest

0.11

 recording

0.11

 except

0.11

highest

0.11

 Majority

0.10

 values

0.10

Activations Density 0.069%