The neuron lights up on single‐subword tokens that invoke numeric or quantitative concepts—terms like density, cost/values (including negative values), “first” in “first attempt,” “as yet,” etc.—i.e. words pointing to amounts, measures, or ordered/quantitative status.
