{"schema_version":1,"research_id":"eoa_inverse_innovation_exp05_external_evaluation_20260803","source_assessment_id":"computability_boundary_mapping__economics_finance:P3:v0","cell_id":"computability_boundary_mapping__economics_finance","search_queries":["site:bis.org intraday liquidity collateral management monitoring tools bank collateral","site:ibm.com docs CPLEX solution status time limit infeasible","site:gurobi.com documentation optimization status codes TIME_LIMIT INFEASIBLE","collateral optimization allocation financial institutions constraints academic paper","collateral optimization bank operational challenges intraday settlement constraints report","site:ecb.europa.eu collateral management harmonisation eligibility valuation operational deadline","site:isda.org collateral optimization technology constraints allocation","site:bankofengland.co.uk intraday liquidity collateral management settlement obligations","site:federalreserve.gov SR 11-7 model risk management independent validation outcomes limitations","site:developers.google.com optimization cp-sat solver status UNKNOWN INFEASIBLE","\"A mixed-integer linear programming approach for collateral management\"","collateral allocation optimization mixed integer programming eligibility haircuts concentration","collateral optimization product automated allocation eligibility haircuts concentration limits bank first party","site:broadridge.com collateral optimization allocation eligibility rules","site:clearstream.com collateral optimization allocation service","site:euroclear.com collateral optimization service automated allocation"],"sources":[{"source_id":"S1","title":"Monitoring tools for intraday liquidity management","publisher":"Basel Committee on Banking Supervision, Bank for International Settlements","url":"https://www.bis.org/publ/bcbs248.htm","source_class":"STANDARD","publication_date":"2013-04-11","accessed_at":"2026-08-03","claims_supported":["Intraday-liquidity management is a material part of bank liquidity-risk management.","Internationally active banks must apply monitoring tools addressing timely payment and settlement obligations.","Treasury and liquidity-risk functions are identifiable institutional stakeholders."]},{"source_id":"S2","title":"Single Collateral Management Rulebook for Europe (SCoRE)","publisher":"European Central Bank","url":"https://www.ecb.europa.eu/paym/collateral/score/html/index.es.html","source_class":"OFFICIAL_GUIDANCE","publication_date":"n.d.","accessed_at":"2026-08-03","claims_supported":["Collateral workflows remain fragmented enough to motivate common European rules.","Operational scope includes margin calls, cut-off times, dynamic and static data, collateral sourcing, bilateral and triparty workflows, and ISO 20022 messaging.","Workflow and data integration, not merely solver construction, constrain deployment."]},{"source_id":"S3","title":"Collateral optimization: capabilities that drive financial resource efficiency","publisher":"Ernst & Young","url":"https://www.ey.com/content/dam/ey-unified-site/ey-com/en-us/insights/banking-capital-markets/documents/ey-collateral-optimization.pdf","source_class":"AUTHORITATIVE_SECONDARY","publication_date":"2026","accessed_at":"2026-08-03","claims_supported":["Financial institutions still use fragmented, manual, or offline collateral-allocation processes with incomplete information.","Operational infrastructure and settlement-timing constraints can prevent execution of identified allocations.","Institutions express a need for enterprise data, infrastructure, and algorithms that represent relevant constraints."]},{"source_id":"S4","title":"Euroclear adds new client to its Collateral Optimisation Service","publisher":"Euroclear","url":"https://www.euroclear.com/newsandinsights/en/press/2026/mr-13-euroclear-adds-new-client-to-its-collateral-optimisation-service.html","source_class":"COMMERCIAL_FIRST_PARTY","publication_date":"2026-05-12","accessed_at":"2026-08-03","claims_supported":["MUFG is an identifiable adopter of a live collateral-optimization service.","The service already provides automated, transparent allocation, multiple scenarios, binding-constraint handling, and settlement integration.","The service operates within infrastructure reporting more than EUR2 trillion of collateral under management, showing institutional-scale prior practice."]},{"source_id":"S5","title":"Accessing solution status","publisher":"IBM","url":"https://www.ibm.com/docs/en/icos/22.1.0?topic=information-accessing-solution-status","source_class":"OFFICIAL_PRODUCT_DOCUMENTATION","publication_date":"n.d.","accessed_at":"2026-08-03","claims_supported":["CPLEX already distinguishes proven infeasibility from unknown status.","IBM explicitly identifies a time limit as a common cause of UNKNOWN.","Feasible, infeasible, unknown, and error are established solver-interface semantics."]},{"source_id":"S6","title":"CP-SAT Solver","publisher":"Google for Developers","url":"https://developers.google.com/optimization/cp/cp_solver","source_class":"OFFICIAL_PRODUCT_DOCUMENTATION","publication_date":"n.d.","accessed_at":"2026-08-03","claims_supported":["CP-SAT supports finite integer constraint models and enumeration of all solutions.","CP-SAT distinguishes FEASIBLE, INFEASIBLE, MODEL_INVALID, and UNKNOWN.","A time, memory, or custom limit without a solution or infeasibility proof produces UNKNOWN rather than INFEASIBLE."]},{"source_id":"S7","title":"Approaching Collateral Optimization for NISQ and Quantum-Inspired Computing","publisher":"arXiv","url":"https://arxiv.org/abs/2305.16395","source_class":"PRIMARY_RESEARCH","publication_date":"2023-05-25","accessed_at":"2026-08-03","claims_supported":["Collateral allocation is already formulated as MILP and binary optimization.","Large institutional collateral pools create combinatorial, time-consuming allocation problems.","Classical CPLEX, Gurobi, and related solvers are established approaches, while heuristic formulations trade accuracy and resources."]},{"source_id":"S8","title":"Supervisory Guidance on Model Risk Management","publisher":"Board of Governors of the Federal Reserve System","url":"https://www.federalreserve.gov/frrs/guidance/supervisory-guidance-on-model-risk-management.htm","source_class":"GOVERNMENT_OR_REGULATOR","publication_date":"2026-04-17","accessed_at":"2026-08-03","claims_supported":["Bank model use should reflect purpose, limitations, materiality, testing, validation, monitoring, documentation, and clear accountability.","Validation should assess assumptions, methods, data, reliability, limitations, and outcomes before use where practicable.","Vendor and internally developed quantitative systems remain subject to institution-level governance and oversight."]}],"problem_evidence":{"support":"MODERATE","rationale":"Collateral allocation and timely settlement visibly matter, and current sources document fragmented data, manual decisions, tight operational constraints, and combinatorial difficulty. Solver documentation also confirms that a cutoff is semantically different from proven infeasibility. However, none of the eight sources documents the proposal's decisive incident pattern: a financial institution actually translating a cutoff or UNKNOWN result into NO_FEASIBLE_ALLOCATION. The broad risk is supported; its asserted prevalence and realized loss are not.","source_ids":["S1","S2","S3","S5","S6","S7"]},"stakeholder_evidence":{"support":"MODERATE","rationale":"Treasury, collateral, liquidity-risk, settlement, and model-risk functions are credible adopters or authorizers. MUFG is a named adopter of collateral optimization, Basel imposes monitoring expectations, and 2026 Federal Reserve guidance supports validation and limitation controls. These sources express pull for optimization and governance, but not specifically for an exhaustive reference enumerator or the proposed status taxonomy.","source_ids":["S1","S3","S4","S8"]},"prior_art":{"proximity":"ESTABLISHED_PRACTICE","closest_analogues":[{"name":"Commercial solver status semantics","similarity":"IBM CPLEX and Google CP-SAT already separate feasible results, proven infeasibility, invalid models, errors, and cutoff-driven unknown outcomes—the intervention's central semantic rule.","remaining_difference":"The proposal adds an institution-specific downstream contract, a versioned snapshot specification, and independent checking of collateral-specific witnesses and exhaustive-coverage claims.","source_ids":["S5","S6"]},{"name":"MILP-based collateral optimization","similarity":"Published research already models collateral allocation as a finite MILP or binary combinatorial problem and uses classical solvers.","remaining_difference":"The research optimizes allocation and studies computational performance; it does not establish that a bank's operational interface preserves UNKNOWN instead of mapping it to infeasible.","source_ids":["S7"]},{"name":"Euroclear Collateral Optimisation Service","similarity":"A deployed service already performs automated, transparent, constraint-aware scenario allocation integrated with settlement infrastructure, with MUFG as a named adopter.","remaining_difference":"Public product information does not disclose its timeout-to-status mapping, infeasibility evidence, independent checker, or exhaustive reference implementation.","source_ids":["S4"]},{"name":"Bank model-risk governance and collateral workflow standards","similarity":"Existing official practice requires validation, documented limitations, accountability, standardized data, cut-off handling, and ongoing monitoring.","remaining_difference":"Neither source prescribes this proposal's exact five-state interface or brute-force finite-snapshot reference decider.","source_ids":["S2","S8"]}],"distinctive_claim_remaining":"For an institution whose current workflow collapses solver cutoff into economic infeasibility, adding a separately checked finite-snapshot reference decider and enforcing FEASIBLE/PROVEN_INFEASIBLE/UNKNOWN_AT_CUTOFF/INVALID_OR_STALE/SYSTEM_FAILURE end to end will eliminate false infeasibility records without materially increasing missed settlement deadlines. This is contrastive and falsifiable, but the mathematical and solver-status components are already established practice.","confidence":"HIGH"},"implementation_evidence":{"support":"STRONG","rationale":"For finitely many indivisible or integer-unit lots, calls, destinations, fixed-point values, and mechanically evaluable predicates, exhaustive enumeration is technically total, and CP-SAT supplies directly relevant integer modeling and status semantics. IBM independently confirms the same cutoff distinction. The primary barriers are semantic fidelity, numeric consistency, data freshness, integration, certificate practicality, and governance—not computability. ECB workflow standards and Federal Reserve model-risk guidance show that those barriers require formal controls and accountable approval. An offline read-only pilot creates no direct authority or settlement hazard; live use must remain under treasury, operations, and risk control.","source_ids":["S2","S5","S6","S7","S8"]},"scores":{"meaningful_impact":{"score":4,"rationale":"Incorrect collateral or liquidity decisions can affect timely settlement, and external sources document material operational and economic inefficiencies. Realized impact from the specific status error remains unmeasured.","source_ids":["S1","S3","S4"]},"stakeholder_pull":{"score":4,"rationale":"Banks, market infrastructures, supervisors, and named adopter MUFG visibly invest in collateral optimization and governance, although no source requests this exact control.","source_ids":["S1","S4","S8"]},"incremental_advantage":{"score":2,"rationale":"End-to-end status preservation and independent reference checking could improve auditability, but standard solvers already implement the core distinction between timeout, unknown, and proven infeasibility.","source_ids":["S5","S6"]},"distinctiveness_plausibility":{"score":1,"rationale":"The proposal combines familiar solver semantics, exhaustive finite search, model validation, and collateral optimization. Only institution-specific integration and enforcement remain plausibly distinctive.","source_ids":["S4","S5","S6","S7","S8"]},"technical_implementability":{"score":5,"rationale":"A tiny finite-snapshot reference enumerator, independent witness checker, and forced-cutoff tests are straightforward with integer constraint tooling.","source_ids":["S5","S6","S7"]},"adoption_authority_feasibility":{"score":4,"rationale":"Treasury or collateral operations can authorize an offline exercise, while model-risk and risk-control functions have established roles for validation and use limitations. Production authority is institution-specific.","source_ids":["S4","S8"]},"evidence_readiness":{"score":3,"rationale":"Synthetic implementation evidence is immediately obtainable, but proving the alleged workflow defect and economic benefit requires proprietary solver logs and downstream records.","source_ids":["S3","S5","S6"]},"safety_net_benefit":{"score":4,"rationale":"Preserving UNKNOWN prevents an unsupported negative verdict and supports controlled escalation. It does not itself provide collateral or prevent settlement delay when computation remains unresolved.","source_ids":["S1","S5","S6"]},"scalability":{"score":3,"rationale":"The status contract and checker scale conceptually, but exhaustive enumeration grows combinatorially and enterprise deployment faces fragmented data, settlement, and organizational constraints.","source_ids":["S2","S3","S7"]}},"score_confidence":"MODERATE","costs":{"first_evidence":{"band_2026_usd":"10K_TO_50K","scope":"One partner-bank log audit plus an offline prototype: define the frozen-snapshot schema, implement a checker and tiny enumerator, generate adversarial synthetic cases, and replay a small sample of de-identified cutoff events.","confidence":"MODERATE","assumptions":["Four to eight person-weeks across an optimization engineer, collateral subject-matter expert, data analyst, and independent reviewer.","Existing solver and development infrastructure are available.","No live custody or payment-system connection is included."],"source_ids":["S5","S6","S8"]},"initial_deployment_startup":{"band_2026_usd":"50K_TO_250K","scope":"Build a shadow-mode service with snapshot validation, fixed-point arithmetic, witness checking, status mapping, audit records, regression tests, and one downstream integration.","confidence":"LOW","assumptions":["One business unit and one existing collateral platform.","Existing eligibility and inventory feeds can be reused.","No replacement of the production optimizer or custody platform."],"source_ids":["S2","S3","S8"]},"operational_launch":{"band_2026_usd":"250K_TO_1M","scope":"Production hardening across treasury, collateral operations, model risk, data governance, and settlement interfaces, including parallel testing, approvals, monitoring, training, and rollback controls.","confidence":"LOW","assumptions":["A large financial institution with several legacy interfaces.","Launch is limited to one legal entity or collateral pool.","The project changes status handling but does not rebuild enterprise collateral infrastructure."],"source_ids":["S2","S3","S4","S8"]},"annual_recurring":{"band_2026_usd":"50K_TO_250K","scope":"Ongoing model validation, rule and schema maintenance, regression testing, incident review, monitoring, documentation, and periodic independent challenge.","confidence":"LOW","assumptions":["Quarterly rule releases and annual independent validation.","No dedicated 24x7 engineering team.","Existing operations and model-risk staff absorb routine monitoring."],"source_ids":["S2","S8"]}},"verified_pipeline_gates":{"externally_supported_problem":{"status":"UNCERTAIN","reason":"The importance and technical possibility of the error are externally supported, but no direct source verifies that a bank actually maps a cutoff or UNKNOWN to NO_FEASIBLE_ALLOCATION.","source_ids":["S1","S3","S5","S6"]},"externally_credible_adopter_or_authorizer":{"status":"YES","reason":"MUFG is a named collateral-optimization adopter; bank treasury, liquidity, and model-risk functions are recognized stakeholders under Basel and Federal Reserve guidance.","source_ids":["S1","S4","S8"]},"distinct_testable_incremental_claim":{"status":"YES","reason":"A controlled comparison can test whether the guarded status contract eliminates false infeasibility records without unacceptable deadline impact, despite substantial collision with existing solver semantics.","source_ids":["S5","S6"]},"bounded_next_evidence_step":{"status":"YES","reason":"A fixed-window audit and shadow replay can use explicit comparators, a capped snapshot sample, and predetermined falsifiers.","source_ids":["S5","S6","S8"]},"no_unresolved_safety_or_authority_stop":{"status":"YES","reason":"A read-only, offline audit and replay does not pledge assets or issue settlement instructions. Production use remains subject to treasury, operations, and model-risk authority.","source_ids":["S1","S8"]},"credible_cost_scope_and_range":{"status":"YES","reason":"The four ranges distinguish evidence, shadow deployment, production launch, and recurring governance, with scope and staffing assumptions explicit. Confidence remains low for deployment because institution-specific integration is unknown.","source_ids":["S2","S3","S8"]}},"next_evidence_step":"With one partner institution, audit a predeclared 90-day sample of optimizer and downstream workflow logs. Include every run that ended by time, node, memory, numerical, user, or system limit, capped at 200 events, plus 50 randomly sampled completed runs. Reconstruct frozen inputs where available and replay them through (A) the current production status mapping, (B) the proposed guarded mapping using the incumbent solver's native status, and (C) an independent CP-SAT or exhaustive reference for snapshots within a predeclared small envelope. Primary measures are the number of cutoff events recorded as infeasible, later feasible witnesses contradicting negative records, status disagreements, checker failures, and added decision latency. Falsify the problem claim if no cutoff is mapped to infeasible and every negative already has proof-equivalent support; falsify the intervention if it accepts an invalid witness, produces a proved negative contradicted by a valid allocation, or its latency causes more missed deadlines than false negatives prevented. Compare manual review burden and escalation frequency between A and B. Do not connect the study to pledge, custody, payment, or settlement instructions.","blocking_evidence":["No direct external evidence establishes that the proposed timeout-to-infeasible error occurs in a real institution.","The frequency, financial consequence, and settlement consequence of such errors are unknown.","Institution-specific eligibility logic, rounding, custody, legal, and timing constraints may prevent the finite snapshot from faithfully representing executable allocations.","The storage and independent-checking burden for production-scale infeasibility evidence is unmeasured.","Operator and downstream-system responses to UNKNOWN_AT_CUTOFF require live workflow observation.","The cost ranges lack institution-specific staffing, integration, licensing, and control estimates."],"research_disposition":"PARTNERED_RESEARCH_PROGRAM","world_novelty_boundary":"The eight-source review finds established solver-status semantics, collateral MILP research, deployed optimization services, workflow standards, and model-risk governance. It does not establish exhaustive world novelty, patentability, freedom to operate, market size, or realized impact; all remain unmeasured.","arm":"COMPLETE_PROPOSAL_PORTFOLIO","candidate_version":0,"controller_recommendation":{"action":"STOP_EMPIRICAL_RESEARCH_NEEDED","repairable":false,"material_progress_observed":true,"progress_targets":["Obtain proprietary evidence on whether cutoff or UNKNOWN outcomes are mapped to infeasibility and how often.","Measure contradictory later witnesses, funding effects, settlement effects, and manual escalation costs over a fixed observation window.","Demonstrate semantic fidelity between the frozen snapshot and executable collateral constraints.","Compare current and guarded status mappings in shadow mode with predeclared latency and error thresholds.","Validate witness checking, negative-verdict support, numeric consistency, stale-input handling, and downstream preservation of UNKNOWN.","Replace broad cost assumptions with institution-specific integration, governance, licensing, and recurring-support estimates."],"reason":"Bounded web research establishes importance, adopters, technical feasibility, governance authority, and substantial prior-art collision. It cannot verify the proposal's defining operational failure or its incremental benefit. Those questions require proprietary logs, workflow observation, and shadow replay; under the evaluation rule this requires an empirical-research stop rather than further web research."},"proposal_index":3}