INDEX

Explanations

machines or equipment failures

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

 leak

-0.11

Interceptor

-0.09

 pist

-0.09

Gam

-0.09

leanup

-0.09

ecial

-0.09

olle

-0.09

 rapes

-0.09

ptron

-0.09

 Leak

-0.09

POSITIVE LOGITS

 failure

0.29

 failures

0.24

 Failure

0.23

failure

0.21

Failure

0.20

 fail

0.16

 FAILURE

0.16

fail

0.16

 Fail

0.16

æķħ

0.16

Activations Density 0.069%