INDEX

Explanations

comma with

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

 potentially

-0.10

opp

-0.09

ipe

-0.09

çĶ±äºİ

-0.09

éĥ½æĺ¯

-0.08

oge

-0.08

haven

-0.08

 Ð¿ÑĢÐ°Ð²Ð´Ð°

-0.08

cur

-0.08

 fres

-0.08

POSITIVE LOGITS

 able

0.21

 enough

0.19

è¶³

0.17

èĥ½å¤Ł

0.17

 shouldn

0.17

èĥ½

0.16

 should

0.15

 Ø¨ØªÙĪØ§ÙĨ

0.15

 Able

0.15

 easy

0.15

Activations Density 0.155%