Baldi–Ullmo: source adjudication of the completed panel
The supplied proof needs substantive correction. The subgroup-countability assertion in §6.2.1 is false, with a checked counterexample inside an actual nonarithmetic lattice. This does not refute the main geometric finiteness theorem.
The latter has an independent proof in Bader–Fisher–Miller–Stover's complex hyperbolic paper, arXiv:2006.03008, Theorem 1.1 and Corollary 1.2. We checked the precise statements in v1, available on 4 June 2020, before the recorded 17 July 2020 received-date cutoff. We have not independently audited that paper's entire proof. The independent result does not validate the faulty steps of Baldi–Ullmo.
This is coordinating AI source adjudication, with human grades pending, not an accepted native arbitration or formal verification. All native arbitration attempts were incomplete or rejected by validation. Agreement between reports is supporting context; the arguments below supply the evidence.
What has been established in this audit
| Passage in frozen arXiv v3/v4 body | Adjudication | Scope |
|---|---|---|
| §6.2.1, subgroup parameter set Sigma, in proof of Theorem 6.2.2 | Countability assertion false | Uncountably many graph subgroups present the same diagonal correspondence |
| Corollary 1.3.3, repeated as 5.5.4 | Proof officially withdrawn; statement open in the erratum | No counterexample or independent resolution is supplied here |
| Theorem 4.2.2, unitary therefore self-dual | False auxiliary assertion | The adjoint trace calculation has a short conditional correction |
| Corollary 5.2.5, finite centralizer from rank one | False auxiliary assertion | Compact factors may preserve geometric orbits without preserving group equalities |
| Theorem 5.1.4, monodromy | Substantive proof defects | Intersection/projection, simplicity and trace-field identifications need correction |
| Proposition 5.3.3, special-locus comparison | Unresolved essential proof step | The full orbit/preimage identity is not established |
| Proposition 6.1.4 and downward induction | Substantive proof defects | Strict dimension drop, effective parameters and containment exclusion are required |
| Theorem 5.4.1, Ax–Schanuel reduction | Global covering assertion incorrect; local dimension calculation works | Componentwise correction does not resolve its earlier comparison inputs |
| Weakly-special definition and special-point comparison | Literal definitions conflict | Trivial normal subgroups make every point weakly special; the intended Hodge-special notion needs correction |
| Theorem 6.0.1 | Missing maximality condition | Use the intended maximal statement in the headline assessment |
The strongest new checked finding
The counterexample note verifies all membership conditions in the printed Sigma, rather than merely pointing to a possible compact-factor problem. For the DPP p=5 lattice, exact field computations give four real places with signatures (2,1), (2,1), (2,1), (3,0). The compact adjoint factor acts trivially on its period factor. Holding the first and other coordinates fixed while varying this compact coordinate changes the graph subgroup but leaves its orbit, Hodge incidence and intersection dimension intact. Thus every such graph lies in Sigma, despite there being uncountably many of them. This tests the assertion within the paper’s claimed period-map setup; it does not independently certify the construction of that setup.
The calculation imports DPP's lattice and nonarithmeticity theorems from the 2014 preprint and reconstructs the compact conjugate directly from its matrices. It checks all 48 automorphisms of an exact cyclotomic computation field, not just one approximate eigenvalue test. No later Stover computation is used as an original proof input. A torsion-free finite-index passage preserves the relevant adjoint trace field and real forms.
This definitively invalidates the inference that specialness of the geometric correspondence forces its entire real presenting subgroup to be rational. Counting effective first-factor graphs could remove this defect, but the remaining product and Hodge-comparison arguments still need to be supplied.
What is repairable locally, and what remains unresolved
Local corrections give exact witnesses and the corrected adjoint trace identity. They also distinguish the valid global-case Goursat strategy from the unproved arbitrary-subvariety version. The false inference from equal image fields to equal embeddings cannot be fixed merely by citing Goursat: the equality must concern the same individual adjoint traces, and those traces must generate the claimed field for the actual monodromy group.
The special-locus comparison still requires the reverse inclusion between a period-map preimage and a full real subgroup orbit. Its availability does not follow from the official erratum, from generic monodromy alone, or from the existence of a plausible repair. Restriction from a smaller trace field and restriction from the ambient field must also be distinguished.
The final countable-definable-implies-finite principle is sound. Its premises are where the proof fails: the wrong subgroup parameters are counted, a strict nondistinguished-factor dimension drop is not established for all allowed parameters, and the downward induction excludes equality instead of containment. The numerical inequality itself and a componentwise Ax–Schanuel dimension shift are checked in the accompanying notes. Neither is a completed replacement proof.
The integral-VHS construction is not certified by this audit. Some external hypotheses and the coefficient-field/descent conventions remain unverified. We therefore do not upgrade it to “checked” based on the supplied Astra review, nor label every new Sol concern a demonstrated counterexample.
The official erratum is separate
The Annals erratum, 199 (2024), p. 481, acknowledges the incomplete proof of Corollary 1.3.3/5.5.4 and leaves its truth undetermined. The appended v4 explanation identifies a period-preimage identity needing fixed part on an algebraic base; the required finite-volume algebraic base was the desired conclusion. The source-reference audit finds no use of that corollary as an input to the main finiteness proof. The additional issues in this report are independently checked and should not be attributed to the authors' erratum.
Headline and literature status
The intended statement concerns maximal positive-dimensional complex totally geodesic subvarieties of a nonarithmetic finite-volume ball quotient. It is covered by the independent complex-hyperbolic BFMS theorem. A maximal proper complex geodesic subspace cannot lie in a proper totally real geodesic subspace, because a totally real tangent space contains no complex line. Hence maximality in the complex category suffices for this application of their finiteness result; finite covers handle orbifold/torsion conventions. This is a theorem match, not a new proof audit of BFMS.
The independent theorem does not settle the withdrawn lattice-intersection corollary, certify every Hodge-theoretic comparison, or prove the paper's full Ax–Schanuel claims. These remain separately scoped. We found no additional correction resolving the compact-factor parameter assertion in the sources checked; this is not an exhaustive priority search or a claim of discovery.
Editions, coverage and costs
The native panel reviewed arXiv:2005.03524v3 (19 May 2022). Its mathematical body is identical to v4 (22 September 2023) before the added erratum; bibliography and appendix differences are frozen separately. The complete typeset journal PDF was not obtained, so this report does not silently transfer every additional finding to identical journal numbering. The official erratum's journal location is authoritative. Native labels, source lines and hashes anchor all new decisions.
There are 17 terminal attempts: one accepted full-source Sol assessment, seven accepted local assessments, six provider-incomplete attempts and three schema-invalid attempts. There are zero accepted native arbitration reports. The larger-cap reruns added useful Sol, Opus and Gemini reports; they did not produce a usable arbiter verdict. No rejected response or private reasoning stream is treated as a verdict or included in this package.
Conservative usage accounting, including failed work, totals $7.074864 of $10. No new API calls were made for adjudication. This is accounting from provider usage/status records, not an invoice. The coverage files and all-attempt ledger preserve that distinction.
The source inventory covers the complete supplied manuscript. Its graph contains
a curated, explicitly partial set of quoted proof-use edges. Source-bound decisions
color the supplied proof/auxiliary assertion, without making an unresolved
consumer statement false. All other new full-report allegations remain attributed
in allegation-inventory.json; pending claims are not promoted to confirmed errors.
This publication records the scoped source adjudication. It contains no complete repair of §§5–6 and makes no headline-refutation claim. The independent reading view and its roadmap remain separate from these panel conclusions.
Additional supporting archive items are identified in the original report. View the manuscript hashes and repository snapshot.