INDEX
Explanations
structured headings and lists
The neuron detects major section headings and structural document markers (e.g., numbered or bolded section titles and topic headers).
New Auto-Interp
Negative Logits
<unused1865>
0.29
<unused331>
0.29
වන
0.26
전을
0.25
타입을
0.25
<unused356>
0.25
certes
0.24
대를
0.24
ुल्लाह
0.24
硋
0.24
POSITIVE LOGITS
</h2>
0.43
↵↵
0.42
↵
0.41
</h3>
0.40
?
0.35
？
0.35
：
0.34
</h4>
0.33
。
0.33
؟
0.31
Activations Density 2.626%