Stage-Audit: Auditable Source-Frontier Discovery for Cross-Wiki Tables
For users of LLM-generated structured data, this work provides a method to audit and improve source traceability, though the evaluation is limited to a small curated set.
Stage-Audit addresses the problem of LLM-curated tables containing unsupported rows with false page-level citations by introducing a disjoint curator-auditor system with row-level source verification. It improves source-frontier precision from 0.356 to 0.505 (+42%) and F1 from 0.334 to 0.451 (+35%) on a 51-instance evaluation set.
LLM-curated tables can appear source-grounded while containing unsupported rows: the curator may recall entries from parametric memory and retroactively attach page-level citations that are not the actual source. We study this hazard in Seed2Frontier discovery: the task of finding complement Wikipedia pages from a seed page to assemble a structured table. Stage-Audit addresses it with disjoint curator-auditor write rights, a row-level source-citation gate, and a 12-check audit taxonomy over keys, schema, source roles, cardinality, and scope. On a curated 51-instance Seed2Frontier evaluation set spanning 15 top-level domains, Stage-Audit improves source-frontier precision over a vanilla LLM curator from 0.356 to 0.505 (+42% relative) and F1 from 0.334 to 0.451 (+35%), while maintaining explicit per-row source traceability. The vanilla-LLM-vs-Stage-Audit comparison isolates the policy contribution rather than LLM-based discovery in general.