INDEX

Explanations

first or earliest

This neuron detects the publication‐metadata phrase “appeared first on,” i.e. the words marking where and when a post first appeared.

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

 obviously

0.62

 Obviously

0.61

INITIAL

0.59

Obviously

0.59

BEGIN

0.58

 initialize

0.57

 Initialization

0.56

 obvious

0.56

 Initialize

0.56

NOTA

0.55

POSITIVE LOGITS

 ήταν

0.91

 ছিল

0.78

 был

0.77

 estis

0.74

 была

0.72

 были

0.72

 среди

0.70

 იყო

0.70

 među

0.70

 були

0.69

Activations Density 0.067%