INDEX

Explanations

entity action verbs

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

 sweeps

-0.12

 undercut

-0.12

 conform

-0.11

 undermining

-0.10

Raises

-0.10

 stutter

-0.10

Adds

-0.10

 augment

-0.10

Defines

-0.09

 strengthening

-0.09

POSITIVE LOGITS

Ð»ÑİÑĶ

0.11

Ð»ÑıÑĶ

0.11

Ð»ÑıÐµÑĤ

0.11

 cleans

0.11

Ð³Ð°ÑĶ

0.10

 empt

0.10

 à¤ķà¤°à¤¤

0.10

isman

0.10

errs

0.10

à¤¸à¤¤

0.09

Activations Density 0.239%