When Does Better Confidence Improve a Knowledge-Graph Audit?

A study posted to arXiv (2611.05877v1) finds that better confidence estimates in LLM-built knowledge graphs can yield little audit improvement, because an error-rate × fan-out queue leaves 3.98–4.74 percentage points more errors than a joint model with the same actions on held-out cross-document graphs built from Re-DocRED. Opening shared actions to that joint planner cut residual errors by 0.89–1.43 points relative to its own fact-only plans, while generic LLM verification improved three of four model-budget configurations on each held-out set. The authors model non-adaptive auditing as budgeted covering and bound the regret of scalar-preserving first-order planning by a structural mass term Δ^cl, with equivalence between joint and independent-block models remaining inconclusive.

arXiv:2611.05877v1 Announce Type: new Abstract: Confidence estimates guide human review in LLM-built knowledge graphs, but better error prediction can yield little improvement in an audit. We study this gap through repairability, selection headroom and label validity. Modelling non-adaptive auditing as budgeted covering, we characterise the minimal decision-relevant error statistics and bound the regret of scalar-preserving first-order planning by a structural mass $\Delta^{\mathrm{cl}}$. We simulate ideal repair against benchmark-aligned labels for five extraction configurations and two cross-document graphs on Re-DocRED, followed by frozen confirmation on test and replication on 500 hash-selected training documents. On the held-out cross-document graphs, an error-rate $\times$ fan-out queue leaves 3.98-4.74 percentage points more errors than a joint model with the same actions. By comparison, opening shared actions for that joint planner reduces residual errors by 0.89-1.43 points relative to its own fact-only plans, in point estimates. Equivalence between joint and independent-block models remains inconclusive. Generic LLM verification improves three of four model-budget configurations on each held-out set. LLM adjudication identifies possible benchmark omissions.