INDEX

Explanations

consistent with previous findings

The neuron detects mentions of prior research findings or citations (e.g. references to “results,” “consistent with,” “previous findings,” “others’ work,” etc.).

New Auto-Interp

Configuration

Prompts (Dashboard)

24,576 prompts, 128 tokens each

Dataset (Dashboard)

monology/pile-uncopyrighted

Embeds

IFrame

Link

Not in Any Lists

Negative Logits

 también

-2.05

for

-1.85

in

-1.72

not

-1.61

</em>

-1.55

</h2>

-1.50

 tohto

-1.47

 berbeda

-1.42

</strong>

-1.41

 này

-1.38

POSITIVE LOGITS

 other

1.66

 sors

1.52

kic

1.50

 resten

1.43

 klokken

1.43

 aguj

1.41

ślę

1.40

 bén

1.36

aucun

1.35

 synes

1.34

Activations Density 0.088%