{"status":"ok","message-type":"work","message-version":"1.0.0","message":{"indexed":{"date-parts":[[2026,10,9]],"date-time":"2026-10-09T06:51:21Z","timestamp":1791528681202,"version":"4.3.4"},"posted":{"date-parts":[[2026]]},"group-title":"SSRN","reference-count":0,"publisher":"Elsevier BV","license":[{"start":{"date-parts":[[2026,1,1]],"date-time":"2026-01-01T00:00:00Z","timestamp":1767225600000},"content-version":"unspecified","delay-in-days":0,"URL":"https:\/\/creativecommons.org\/licenses\/by\/4.0\/"}],"content-domain":{"domain":[],"crossmark-restriction":false},"short-container-title":[],"abstract":"<jats:p>&lt;p&gt;&amp;nbsp; &amp;nbsp; &amp;nbsp;We study NVIDIA cuOpt for continuous corporate-bond portfolio construction under spread and default risk, rating and issuer limits, duration matching, duration-times-spread bounds, and turnover constraints. The contribution is an open application benchmark, not a new optimization algorithm. We compare four endpoints: a raw first-order point, a feasibility-repaired point, a basic solution, and a directly modeled conic solution. A central case is a standard-deviation tracking-error target. A Gaussian-calibrated mean-absolute-deviation (MAD) proxy set to the same nominal budget delivers 28.5 bp against a ten-basis-point target, an overshoot of 2.85x. Recalibration is possible: a 2.25 bp MAD budget delivers 9.47 bp in-model and remains below target on 3 fixed validation draws. It requires a 4.44x nominal haircut, reduces the scaled objective by 13.6 bp, and increases CVaR loss by 4.43 bp relative to the exact SOCP portfolio. Thus the defect is not unavailability of the mandate but failure of naive calibration, with an observable cost-accuracy trade-off.&lt;\/p&gt;\n&lt;p&gt;&amp;nbsp; &amp;nbsp; &amp;nbsp;Both conic solvers deliver the reference-size sigma mandate; working-size Clarabel remains feasible while the native cuOpt path records a crash. Contended SOCP timings are excluded. On the MAD-form LP, tight raw PDLP passes the proxy's numerical constraints without crossover, but this is not sigma compliance. A paired turnover leave-one-out and a matched-feasibility QP comparison show that constraint structure, representation, and solver endpoint matter alongside device choice. Results use a controlled credit scenario generator, not an empirical asset-pricing test or an executable bond-trading system. Supplemental pilots examine single-request timing and hypothetical bond-lot repair, retaining failed quality gates rather than reporting new speed advantages. Reported measurements are generated from archived result files.&lt;\/p&gt;<\/jats:p>","DOI":"10.2139\/ssrn.7587239","type":"posted-content","created":{"date-parts":[[2026,10,9]],"date-time":"2026-10-09T06:29:48Z","timestamp":1791527388000},"source":"Crossref","is-referenced-by-count":0,"title":["Credit Portfolio Optimization with cuOpt:&amp;nbsp;A Reproducible Benchmark of Risk, Constraints, and Solver Endpoints"],"prefix":"10.2139","author":[{"ORCID":"https:\/\/orcid.org\/0009-0003-7637-9395","authenticated-orcid":true,"given":"Fred","family":"Hu","sequence":"first","affiliation":[],"role":[{"vocabulary":"crossref","role":"author"}]}],"member":"78","container-title":[],"original-title":[],"deposited":{"date-parts":[[2026,10,9]],"date-time":"2026-10-09T06:29:48Z","timestamp":1791527388000},"score":1,"resource":{"primary":{"URL":"https:\/\/www.ssrn.com\/abstract=7587239"}},"subtitle":[],"short-title":[],"issued":{"date-parts":[[2026]]},"references-count":0,"URL":"https:\/\/doi.org\/10.2139\/ssrn.7587239","relation":{},"subject":[],"published":{"date-parts":[[2026]]},"subtype":"preprint"}}