ABSTRACT                          

Holocene relative sea-level (RSL) evolution across the northern South China Sea (SCS) remains highly debated, especially regarding the timing, magnitude and duration of the mid-Holocene highstand. Existing disagreements are likely influenced by differences in data quality and the interpretation of sea-level proxies, as previous compilations have rarely applied unified quality screening and comprehensive uncertainty quantification to sealevel index points (SLIPs) and marine- and terrestrial-limiting data. In this work, we systematically compiled 374 published records (including SLIPs and limiting data), standardized their depositional metadata and uncertainty parameters, and retained 146 records for the primary reconstruction after multi-criteria quality filtering. We then reconstructed Holocene RSL evolution using an errors-in-variables integrated Gaussian process (EIV-IGP) Bayesian model, and compared modeled outputs with far-field glacial isostatic adjustment (GIA) predictions from two ICE-6G_C-based model configurations. The reconstruction showed a rapid RSL rise during the early Holocene. RSL then remained above present from 7.0 to 3.4 cal ka BP, with the best-constrained estimates indicating a peak within the range of 1.4–2.6 m. In situ Porites microatolls and intact stratigraphic beachrocks delivered the most robust highstand constraints, while shell and organic muddy deposits only provided low-precision, broad limiting bounds. Mismatches between reconstructed RSL and GIA curves were interpreted as composite residuals that may reflect GIA model uncertainty, regional vertical land motion, sediment compaction, and residual proxy uncertainties. This study provides a standardized and reproducible SLIP database for the northern SCS with explicit uncertainty estimates. The results indicate that differences in proxy reliability contribute substantially to the long-standing regional disagreement, while the quality.