Answers are grounded in this paper and cite their sources.
Core-First Language Theory (CFLT) is a normative discourse-level linearization protocol that fixes the relative order of two constituent tiers as [Core] → [Reason] → [Space] → [Time]: the speaker's salience anchor (the event, identity, state, or request being committed to) occupies sequence-initial position, and the circumstantial ground frame of reason, place, and time follows. CFLT operates above the morphosyntactic level — each language assembles the event nucleus (the Core) in its native syntax, while the protocol governs only the order of the ground frame and the boundary between the two. It addresses a structural bottleneck shared by second-language acquisition and Large Language Model prompting: the cost of linearizing a multidimensional semantic representation into a single ordered string. Adult L2 learners pay it as working-memory overhead; LLMs pay it as order-sensitivity and Lost-in-the-Middle attention degradation. A pilot study (720 trials: 24 cases × 2 conditions × 3 runs × 5 frontier models, English + Mandarin) gives preliminary evidence on the LLM-prompting side — Core-first prompts raised extraction accuracy on a distractor-heavy condition from a mean of 65.6% to 100% across all five models and reduced completion tokens by up to 38% on reasoning models, with one model showing no benefit on a buried-decision variant; results are framed as suggestive, not confirmatory. The human-side predictions are reserved for a falsifiable research agenda. Reproducible data, prompts, and evaluation scripts are included in the repository.
No takes yet. Share an insight, caveat, or question.
Weifeng Yi (2026) studied this question.
Synapse has enriched 8 closely related papers on similar clinical questions. Consider them for comparative context: