A tokenized U.S. Treasury product lives on two ledgers: an off-chain portfolio of government securities and an on-chain wrapper that claims to represent it. The foundational question is whether the on-chain series carries recoverable information about the underlying -- whether a mapping from token to fiat exists, and with what structure. We proceed in three steps. First, fixed-income measurement conventions are reconciled between the ledgers; these corrections are signed and jointly. Second, on the reconciled series we introduce a price-discovery admissibility criterion: a falsifiable test, based on serial dependence and idiosyncratic dispersion, for whether a product's series is market-informative or administratively generated. Of four products with sufficient history it admits exactly one; the rest are dominated by how net asset value is computed and republished, and treating them as spreads fits artifacts. That most of the universe is inadmissible is our principal finding. Third, for the admitted mapping we give a minimal two-factor affine characterization in which the basis enters additively and orthogonally to rates, recovering a quarterly reversion, a small positive long-run basis, and a sharp March 2026 regime change toward parity. Persistence is weakly identified; we propagate it through a profile likelihood into one consequence, a collateral haircut. The contribution is measurement and identification infrastructure for when on-chain fixed income may be treated as a quantitative object.

翻译：代币化的美国国债产品存在于两个账本上：链下的政府证券投资组合和链上的声称代表该组合的封装资产。其根本问题在于链上序列是否包含关于基础资产的可恢复信息——即代币到法币的映射是否存在，以及其结构如何。我们分三步进行。首先，在两个账本之间协调固定收益计量惯例；这些修正带有符号与联合效应。其次，基于协调后的序列，我们引入价格发现可采纳性标准：一个基于序列依赖性和异质性离散度的可证伪检验，用以判断产品的序列是否包含市场信息或由行政生成。在四个拥有足够历史数据的产品中，该标准仅采纳一个；其余产品均受净资产价值计算与重新发布方式主导，将其视为利差处理可拟合现有样本特征。整个集合中大多数产品不可采纳是我们的主要发现。第三，对于被采纳的映射，我们给出一个最小双因子仿射特征描述，其中基差以可加且与利率正交的方式进入，恢复出季度均值回复、较小的正长期基差，以及2026年3月向平价回归的剧烈制度变迁。持久性识别较弱；我们通过轮廓似然将其传导至一个结果——抵押品扣减率。本文的贡献在于为链上固定收益何时可被视为定量对象提供测量与识别基础设施。