INDEX

Explanations

achieving goals or objectives

The neuron fires strongly on the final content token just before a document or section break, effectively detecting “end‐of‐chunk” positions.

New Auto-Interp

Configuration

Prompts (Dashboard)

392,802 prompts, 256 tokens each

Dataset (Dashboard)

monology/pile-uncopyrighted

Embeds

IFrame

Link

Not in Any Lists

Negative Logits

ad

0.86

ir

0.83

ar

0.80

akun

0.77

é

0.77

AN

0.75

Sp

0.74

ale

0.73

Play

0.72

POSITIVE LOGITS

在這

0.93

鶉

0.88

曆

0.85

 Deputies

0.82

 funcionários

0.80

選擇

0.78

্স্ট

0.76

據

0.76

oslav

0.76

 সময়ে

0.75

Activations Density 0.000%