Sadovskii Evidence Bridge (SEB)

Sadovskii Evidence Bridge (SEB) is a standalone foundational translation tool that converts front-end observations into a single admissible tokens-only recognition row under fixed accounting. The core contribution is architectural rather than mechanistic: SEB separates provisional writability screening (SEB-0) from committed recognition (SEB) so that evidence cannot be silently repaired before enforcement. Under SEB, ambiguity is written as explicit zeros, missingness is charged as explicit yield loss, and refusal conditions are named and preserved rather than hidden by smoothing, window travel, or selective discards. The output row is replayable from identifiers and token bundles alone, enabling independent audit and preventing backflow in which later decisions redefine earlier evidence. SEB is designed to sit between raw traces and any downstream enforcement layer, so that certify/downgrade/refuse actions are driven by admissible evidence rather than narrative pressure. The tool is intended for signed claims and direction-dependent effects where lane mixing, domain weighting, and extraction flexibility can manufacture apparent robustness, and where clean failure is more valuable than repair.