INDEX

Explanations

unique show phenomenon possible interpretations

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

<|begin_of_text|>

-0.17

.printStackTrace

-0.14

 Ø¨ÙĪØ§Ø¨Ø©

-0.12

Ø§ÙĦØ¥ÙĨØ¬ÙĦÙĬØ²ÙĬØ©

-0.11

 crossorigin

-0.09

gorit

-0.09

oreferrer

-0.09

olumn

-0.09

ropolis

-0.08

POSITIVE LOGITS

UserCode

0.09

Inlining

0.09

StackNavigator

0.09

;;;;;;;;;;;;;;;;

0.09

SetUp

0.09

Â

0.09

dealloc

0.09

AFX

0.08

NAL

0.08

mayan

0.08

Activations Density 0.022%