INDEX

Explanations

replying to user questions

end-of-user-turn markers, with stronger activation when the preceding user prompt contains sensitive or safety-risk content.

New Auto-Interp

Configuration

Prompts (Dashboard)

238,145 prompts, 512 tokens each

Dataset (Dashboard)

lmsys + oasst1

Embeds

IFrame

Link

Not in Any Lists

Negative Logits

ob

0.39

kem

0.38

 pós

0.38

 agré

0.38

）（

0.38

fæ

0.37

 tief

0.36

 surf

0.36

 base

0.36

 pass

0.36

POSITIVE LOGITS

 replied

0.82

 답변

0.75

 answered

0.71

 reply

0.70

answered

0.70

回答

0.69

 Answer

0.68

ANSWER

0.67

 REPLY

0.65

answer

0.64

Activations Density 0.174%