The article revisits the canonical (Dem>Num>A>N) and non-canonical (A>Dem>Num>N & Dem>A>Num>N) prenominal patterns in Mandarin Chinese, from the perspective of the Labeling Algorithm. It shows that the syntactic distribution of adjectives are different, depending on the attributive-predicative sources and the presence or absence of the subordinator -de. The attributive adjective (with or without -de) and the de-less predicative adjective are analyzed as direct modification of the noun phrase, and it derives as a specifier of the noun, then falling into an agreement Agr projection. The non-canonical pattern undergoes AgrP-movement to the nominal left-periphery, which involves focus moving. However, the predicative adjective with -de is associated with a predicate of a reduced relative clause, and it undergoes successive focus-driven movement in the non-canonical pattern. The supporting evidence is illustrated in superlative and exclamatory adjectives as well as the CP-DP splitting hypothesis. The labeling-based explanation provides a specific picture to see the internal structure of adjectives, to look at how to package the informational constituents in the prenominal pattern, and to demonstrate the systematic parallelism between nominal and clausal domains in Mandarin Chinese.