INDEX

Explanations

failure and its consequences

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

Fatal

-0.11

_FATAL

-0.10

ä¹±

-0.10

licted

-0.09

 Interr

-0.09

GNUC

-0.09

Severity

-0.09

orre

-0.09

Absent

-0.09

severity

-0.09

POSITIVE LOGITS

 failure

0.75

 Failure

0.62

 failures

0.59

failure

0.57

å¤±è´¥

0.56

 failed

0.55

 fail

0.54

Failure

0.53

 Fail

0.50

 FAILURE

0.49

Activations Density 0.108%