sentences that state technical explanations or factual/descriptive information (i.e., salient content words in expository sentences).

It detects descriptions of a person’s clothing and physical appearance.

The neuron detects numeric expressions—numbers, measurements, and decimal/percentage-like tokens.

the neuron responds to content-bearing or topical words (important nouns, verbs, pronouns and discourse markers) rather than function or filler tokens.

sentences or phrases expressing future claims, promises, or predictions (marked by modal/future constructions like "would," "going to," "will").

legal discussion of standards of review—phrases contrasting questions of law and fact (de novo, standard of review, jurisdiction, etc.).

signals that a section heading or paragraph-level label (e.g., a titled or colon-ended section start) is beginning.

It detects numeric tokens and digit-heavy sequences (numbers, figure/section/table indices and other multi-digit numeric strings).

tokens carrying strong semantic content or topical importance (salient content words).

mentions of "viruses" (references to viruses).

the neuron highlights salient, information-dense tokens—important content words (main verbs, nouns, numbers) and emphatic punctuation that carry the core facts or claims.

This neuron detects the definite article "the" (the token " the", especially in phrases like "What is the ...").

tokens that introduce or express an evaluative/opinionative stance (judgments, endorsements, or assessments).

This neuron detects conversation turn boundaries — it activates on the end_of_turn token (end of a user turn).

tokens that mark sentence boundaries or strong punctuation (sentence-ending periods, commas, quotes, exclamations and similar separators).

sentences or phrases containing copular verbs (forms of "be") that state existence or identity (e.g., "is/are ...", "is not ...").

the neuron detects URL and web-domain fragments (parts of web addresses and links).

mentions of the speaker (first-person "I"/"I'm" and similar self-references).

sentences that ask questions, especially "how"‑style interrogative phrases (question words and the following verbs).

tokens that occur at the start of a sentence or turn (sentence/turn-initial words and markers).