INDEX

Explanations

concluding

The neuron fires on occurrences of the root “conclud-” (e.g. conclude, concluded, concluding, conclusion).

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

 Перейти

-2.91

</th>

-2.19

-2.08

 uncovering

-2.08

𖥧

-1.96

蟥

-1.95

 Ähn

-1.91

</h1>

-1.89

geſ

-1.85

pielen

-1.83

POSITIVE LOGITS

</em>

2.55

2.05

给大家

2.02

许久

2.02

を注文

1.97

 godk

1.92

～。

1.91

～～

1.88

の良い

1.84

Activations Density 0.005%