{"schema_version":1,"experiment_id":"eoa_inverse_innovation_exp09_archetype_breadth150_20260804","research_id":"eoa_inverse_innovation_exp09_light_prior_art_20260804","cell_id":"preimage_set_characterization__accounting_auditing","search_lanes":{"direct_problem_and_intervention":{"queries":["bank reconciliation zero balance offsetting errors false assurance audit","site:gao.gov bank reconciliations adjustments errors audit","constraint solver bank reconciliation alternative transaction configurations zero residual"],"source_ids":["SRC1","SRC3"],"no_result_note":null},"synonyms_and_historical_terms":{"queries":["bank reconciliation compensating errors proof of cash audit","site:pcaobus.org bank reconciliation audit balance transaction testing","accounting reconciliation subset sum algorithm bank transactions paper"],"source_ids":["SRC1","SRC2","SRC4"],"no_result_note":null},"products_practices_and_standards":{"queries":["site:blackline.com transaction matching many-to-many reconciliation rules audit certification","site:docs.oracle.com account reconciliation transaction matching one-to-many many-to-one tolerance","site:pcaobus.org bank reconciliation audit balance transaction testing"],"source_ids":["SRC2","SRC3"],"no_result_note":null},"component_combination":{"queries":["constraint programming financial reconciliation transactions matching alternative solutions","transaction reconciliation optimization many-to-many matching research paper accounting","accounting reconciliation subset sum algorithm bank transactions paper"],"source_ids":["SRC3","SRC4"],"no_result_note":null}},"sources":[{"source_id":"SRC1","title":"Department of Defense: Actions Needed to Reduce Accounting Adjustments","publisher":"U.S. Government Accountability Office","url":"https://www.gao.gov/products/gao-20-96","source_type":"PRIMARY_RESEARCH","claims_supported":["GAO found that forced-balance adjustments produced agreement with Treasury balances without reconciling and researching causes of differences.","Some adjustments lacked adequate supporting documentation, creating risk of inaccurate, invalid, or unapproved entries and unreliable financial reporting.","The report's evidence included policy review, interviews, and a random sample of 242 adjustments."]},{"source_id":"SRC2","title":"Auditing Standard No. 15: Audit Evidence","publisher":"Public Company Accounting Oversight Board","url":"https://pcaobus.org/oversight/standards/archived-standards/pre-reorganized-auditing-standards-interpretations/details/auditing-standard-no-15_1787","source_type":"OFFICIAL_STANDARD","claims_supported":["Audit evidence must be sufficient, relevant, and reliable rather than inferred from a single internally produced result.","Company-produced information used as evidence must be tested for accuracy and completeness and evaluated for sufficient precision and detail.","Recognized audit approaches include reperformance, selection of all items, specific-item testing, and sampling; specific-item results cannot be projected to the entire population."]},{"source_id":"SRC3","title":"Setting Up and Configuring Account Reconciliation","publisher":"Oracle","url":"https://docs.oracle.com/en/cloud/saas/account-reconcile-cloud/suarc/GUID-52A3F1EF-7D35-41F5-A284-861268631470.pdf","source_type":"FIRST_PARTY_PRODUCT","claims_supported":["Oracle Account Reconciliation supports one-to-many, many-to-one, many-to-many, subset, tolerance, and balancing-attribute matching.","The engine explicitly encounters ambiguity: multiple transactions can qualify, with a lowest-transaction-ID rule selecting one, while a No Ambiguous setting leaves qualifying transactions unmatched.","Many-to-many matching compares summed transaction amounts, demonstrating that distinct transaction groupings can collapse to the same aggregate match result.","Suggested one-sided adjustments require preparer confirmation or rejection."]},{"source_id":"SRC4","title":"The Subset Sum Matching Problem","publisher":"arXiv (paper accepted at ECAI 2025)","url":"https://arxiv.org/abs/2508.19218","source_type":"PRIMARY_RESEARCH","claims_supported":["The paper formalizes subset-sum matching as a combinatorial-optimization abstraction of financial reconciliation applications.","It defines valid matches over subsets and presents two suboptimal algorithms and one optimal algorithm with benchmark evaluation.","The work establishes that financial reconciliation can involve multiple combinatorial subset configurations rather than only one-to-one matching."]}],"problem_evidence":{"status":"PARTLY_SUPPORTED","finding":"The underlying assurance problem is visible: GAO documents forced agreement produced by adjustments without adequate cause research or support, while Oracle documents ambiguous qualifying matches and many-to-many equality based on summed amounts. These sources support the possibility that aggregate agreement can conceal materially different transaction or adjustment states. They do not establish the prevalence of reviewers treating every zero bank-reconciliation residual as a unique transaction-level explanation, nor do they directly measure resulting undetected misstatements in ordinary monthly bank reconciliations.","source_ids":["SRC1","SRC2","SRC3"]},"closest_prior_art":[{"name":"Oracle Account Reconciliation transaction-matching engine","source_ids":["SRC3"],"overlap":"Freezes configurable matching conditions and tolerances; supports subset and many-to-many sum matching; recognizes ambiguous alternatives; and provides reviewable suggested adjustments.","remaining_difference":"The documentation describes producing or withholding operational matches, including deterministic tie-breaking, rather than enumerating or symbolically characterizing every admissible transaction-level configuration yielding a specified zero residual with a coverage and completeness certificate."},{"name":"Subset Sum Matching Problem algorithms","source_ids":["SRC4"],"overlap":"Formulates financial reconciliation as a bounded combinatorial problem over subsets and supplies optimal and suboptimal solution algorithms.","remaining_difference":"Its objective is an optimal collection of disjoint reconciliation matches, not the complete preimage of one approved bank-reconciliation output across omission, duplication, cutoff, adjustment, and amount-error states, with audit-oriented witnesses, exclusions, and interpretation controls."},{"name":"GAO adjustment root-cause and support procedures","source_ids":["SRC1"],"overlap":"Requires causes of reconciliation differences and accounting adjustments to be researched, documented, monitored, and supported instead of merely forcing balances to agree.","remaining_difference":"It is a control and root-cause discipline, not a solver-based enumeration of all bounded configurations compatible with the displayed result."},{"name":"PCAOB reperformance, full-population, specific-item, and sampling procedures","source_ids":["SRC2"],"overlap":"Provides established audit-evidence routes for checking company-produced information, including reperformance and automated examination of an entire population.","remaining_difference":"It governs evidence sufficiency and item selection but does not require inverse-set characterization of every admissible configuration that maps to a reconciliation residual."}],"prior_art_disposition":"ADJACENT_PRIOR_ART","contrastive_claim_remaining":"For a frozen reconciliation mapping and explicitly finite discrepancy domain, make the complete set or symbolic family of distinct transaction-level configurations producing the approved zero or tolerance-band residual the audit object, and accompany it with collision witnesses, rejected counterexamples, searched-versus-excluded state counts, and a prohibition on treating logical compatibility as evidence of occurrence. The retained prior art supports combinatorial matching, ambiguity handling, root-cause investigation, and audit reperformance separately, but does not show this complete bounded-preimage package.","contrastive_claim_falsifier":"The claim is falsified by a pre-existing bank-reconciliation or audit method that, before this proposal, required freezing the account-period mapping, exhaustively enumerating or symbolically characterizing every admissible zero-residual configuration within a declared discrepancy domain, reporting multiplicity and coverage completeness, preserving witnesses and counterexamples, and separating compatibility from proof of occurrence. It is also operationally falsified if the bounded implementation misses designed zero-sum configurations or cannot account for searched and excluded states.","gates":{"adequate_source_search":{"status":"PASS","rationale":"The screen searched direct phrasing, compensating-error and proof-of-cash terminology, official audit material, first-party reconciliation products, subset-sum research, and component combinations. Four opened sources span four publishers and include official, first-party, and primary-research evidence.","source_ids":["SRC1","SRC2","SRC3","SRC4"]},"supported_problem":{"status":"PASS","rationale":"Official findings show that forced balance agreement can coexist with unsupported adjustments and unresearched causes, and first-party documentation shows ambiguous and many-to-many sum-equivalent matching. This partly supports the stated false-assurance mechanism while leaving its prevalence unmeasured.","source_ids":["SRC1","SRC3"]},"distinct_testable_claim":{"status":"PASS","rationale":"The remaining claim is structurally distinct and falsifiable: complete bounded-preimage characterization with multiplicity and coverage evidence, rather than selecting an operational match, optimizing a match set, investigating known adjustments, or sampling records.","source_ids":["SRC1","SRC2","SRC3","SRC4"]},"bounded_next_test":{"status":"PASS","rationale":"A read-only test on one closed account-period, no more than 25 ambiguous or high-risk items, and configurations containing at most two deviations is finite. Recovery of three designed zero-sum witnesses, rejection of counterexamples, exact searched/excluded counts, boundary logging, and reproduction of the approved configuration provide explicit pass/fail observations.","source_ids":["SRC2","SRC3","SRC4"]},"no_obvious_safety_or_authority_stop":{"status":"PASS","rationale":"The proposed first test uses an authorized analyst, a controlled read-only copy, approved records and synthetic perturbations, no posting rights, and no fraud or control conclusion from solver output. Independent evidence remains necessary, and suggested adjustments in established software likewise require human review. Sensitive financial data, precision boundaries, and access scope remain controls to monitor rather than obvious stops under the stated limits.","source_ids":["SRC2","SRC3"]}},"screen_survival":true,"world_novelty_boundary":"This four-source bounded public-web screen establishes only coarse researchability and an adjacent-prior-art disposition. It cannot establish world novelty, patentability, freedom to operate, market size, expert acceptance, prevalence, implementation feasibility beyond the proposed small test, realized audit value, or absence of undiscovered proprietary, patent, standards, product, or non-indexed prior art."}