{"id":"554532f9461d8d312b0e9d3812d083d3f7082efc56f032e7eb53c7ae86e251f3","title":"arXiv paper tests compliance controls for autonomous trading agents","argument":"The authors introduce TradeGauntlet, a replayable limit-order-book simulator that separates detection, prevention and execution cost to test controls for autonomous trading agents. In simulated sessions with Claude Haiku 5.5 and Sonnet 5.5, every conduct alert was a self-match. Gates that rejected self-crossing orders eliminated those alerts, but an aggregate-gate condition reduced Sonnet’s two-sided fills by about 15%. A conduct reminder reduced Sonnet’s alert rate, while reducing Haiku’s fills without a detectable alert benefit. The authors argue that zero alerts alone do not establish an agent’s compliance. Scripted probes and two search methods tested further weaknesses; the searches found split-order schedules that evaded the implemented conduct rules. A learned baseline caught all 2,400 tested searched episodes but also flagged 15% of a large maker’s sessions. Aggregate caps reduced response-dependent gains unevenly, with smaller reductions when only near-touch depth affected prices. A response fitted to five Nasdaq stocks and a separate ABIDES experiment expose how estimates depend on market assumptions. The authors say the experiments do not establish legal compliance or real-market detection accuracy.","topicIds":["intelligence-and-discovery"],"domain":"science","claimType":"attributed reading","limits":"The reported results come from simulated sessions and tests, not demonstrated real-market detection accuracy or established legal compliance. The authors state that estimates depend on market assumptions; the abstract describes a response fitted to five Nasdaq stocks and a separate ABIDES experiment.","speaker":{"name":"Nandi, Sandip, Bose, Aishwarya, Bose, Apurba, Bakshi, Soovadeep","roleAtSource":"Article author"},"source":{"title":"TradeGauntlet: Evaluating Compliance Controls for Autonomous Trading Agents","publisher":"arXiv cs.AI","format":"Article","url":"https://arxiv.org/abs/2611.18493","publishedAt":"2026-11-13","dateBasis":"Date basis not provided."},"evidence":[{"text":"The authors present TradeGauntlet as a replayable limit-order-book simulator for studying compliance controls for autonomous trading agents.","excerpt":"a replayable limit-order-book simulator that separates detection, prevention, and execution cost","url":"https://arxiv.org/abs/2611.18493","relationship":"Source-stated"},{"text":"In the simulated sessions, all conduct alerts were self-matches, and gates rejecting self-crossing orders eliminated those alerts.","excerpt":"all conduct alerts were self-matches. Gates eliminated those alerts by rejecting self-crossing orders","url":"https://arxiv.org/abs/2611.18493","relationship":"Source-stated"},{"text":"The aggregate-gate condition reduced Sonnet's two-sided fills by about 15%.","excerpt":"the aggregate-gate condition reduced Sonnet's two-sided fills by about 15%","url":"https://arxiv.org/abs/2611.18493","relationship":"Source-stated"},{"text":"The authors caution that zero alerts alone do not establish an agent's compliance.","excerpt":"These results show why zero alerts alone do not establish an agent's compliance.","url":"https://arxiv.org/abs/2611.18493","relationship":"Source-stated"},{"text":"Scripted probes and two search methods tested further weaknesses; the searches found split-order schedules that evaded the implemented conduct rules. A learned baseline caught all searched episodes tested but also flagged 15% of a large maker's sessions.","excerpt":"The searches find split-order schedules that evade the implemented conduct rules; a learned baseline catches all 2,400 tested searched episodes but also flags 15% of a large maker's sessions.","url":"https://arxiv.org/abs/2611.18493","relationship":"Source-stated"},{"text":"The paper says its experiments do not establish legal compliance or real-market detection accuracy.","excerpt":"the experiments do not establish legal compliance or real-market detection accuracy.","url":"https://arxiv.org/abs/2611.18493","relationship":"Source-stated"}],"sourceMedia":{"status":"not_found","sourceUrl":"https://arxiv.org/abs/2611.18493","checkedAt":"2026-11-16T06:36:07.250Z","items":[],"note":"No suitable media was found by the supported page extraction."},"assumptions":[],"openQuestion":{"attribution":"Open question","text":"How would the estimated effects change under market assumptions beyond the response fitted to five Nasdaq stocks and the separate ABIDES experiment?"},"whatWouldChangeView":{"attribution":"Change condition not recorded","text":""},"editorialComparisons":[],"live":{"title":"arXiv paper tests compliance controls for autonomous trading agents","summary":"The authors introduce TradeGauntlet, a replayable limit-order-book simulator that separates detection, prevention and execution cost to test controls for autonomous trading agents. In simulated sessions with Claude Haiku 5.5 and Sonnet 5.5, every conduct alert was a self-match. Gates that rejected self-crossing orders eliminated those alerts, but an aggregate-gate condition reduced Sonnet’s two-sided fills by about 15%. A conduct reminder reduced Sonnet’s alert rate, while reducing Haiku’s fills without a detectable alert benefit. The authors argue that zero alerts alone do not establish an agent’s compliance. Scripted probes and two search methods tested further weaknesses; the searches found split-order schedules that evaded the implemented conduct rules. A learned baseline caught all 2,400 tested searched episodes but also flagged 15% of a large maker’s sessions. Aggregate caps reduced response-dependent gains unevenly, with smaller reductions when only near-touch depth affected prices. A response fitted to five Nasdaq stocks and a separate ABIDES experiment expose how estimates depend on market assumptions. The authors say the experiments do not establish legal compliance or real-market detection accuracy.","question":"How would the estimated effects change under market assumptions beyond the response fitted to five Nasdaq stocks and the separate ABIDES experiment?","claims":[{"text":"The authors present TradeGauntlet as a replayable limit-order-book simulator for studying compliance controls for autonomous trading agents.","excerpt":"a replayable limit-order-book simulator that separates detection, prevention, and execution cost","support":"source-stated"},{"text":"In the simulated sessions, all conduct alerts were self-matches, and gates rejecting self-crossing orders eliminated those alerts.","excerpt":"all conduct alerts were self-matches. Gates eliminated those alerts by rejecting self-crossing orders","support":"source-stated"},{"text":"The aggregate-gate condition reduced Sonnet's two-sided fills by about 15%.","excerpt":"the aggregate-gate condition reduced Sonnet's two-sided fills by about 15%","support":"source-stated"},{"text":"The authors caution that zero alerts alone do not establish an agent's compliance.","excerpt":"These results show why zero alerts alone do not establish an agent's compliance.","support":"source-stated"},{"text":"Scripted probes and two search methods tested further weaknesses; the searches found split-order schedules that evaded the implemented conduct rules. A learned baseline caught all searched episodes tested but also flagged 15% of a large maker's sessions.","excerpt":"The searches find split-order schedules that evade the implemented conduct rules; a learned baseline catches all 2,400 tested searched episodes but also flags 15% of a large maker's sessions.","support":"source-stated"},{"text":"The paper says its experiments do not establish legal compliance or real-market detection accuracy.","excerpt":"the experiments do not establish legal compliance or real-market detection accuracy.","support":"source-stated"}],"limitations":["The reported results come from simulated sessions and tests, not demonstrated real-market detection accuracy or established legal compliance.","The authors state that estimates depend on market assumptions; the abstract describes a response fitted to five Nasdaq stocks and a separate ABIDES experiment."],"mediaId":"01c24557b7160a8a54e5513b70aaf376373865e1b9e482bb1df333d85c4fe570","correctionReason":null,"id":"554532f9461d8d312b0e9d3812d083d3f7082efc56f032e7eb53c7ae86e251f3","publicationId":"d2c7c446-479f-4f0e-8177-295cc538cc70","version":1,"sourceUrl":"https://arxiv.org/abs/2611.18493","sourceTitle":"TradeGauntlet: Evaluating Compliance Controls for Autonomous Trading Agents","publisher":"arXiv cs.AI","publishedAt":null,"originalPublished":{"value":"2026-11-13","precision":"day","basis":"publisher_citation_date"},"firstObservedAt":"2026-11-16T05:06:31.992Z","createdAt":"2026-11-16T06:04:20.036Z","historical":true,"entityIds":["arxiv"],"topicIds":["intelligence-and-discovery"],"media":[{"id":"01c24557b7160a8a54e5513b70aaf376373865e1b9e482bb1df333d85c4fe570","kind":"article","url":"https://arxiv.org/abs/2611.18493","sourceUrl":"https://arxiv.org/abs/2611.18493","label":"Read original","provenance":"source_link","verification":{"status":"verified","checkedAt":"2026-11-16T06:02:52.699Z","method":"acquired_source","contentType":"text/html"},"recommendationReason":"Recommended by Jev for understanding this source; checked by Luna."}],"sourceMedia":{"status":"not_found","sourceUrl":"https://arxiv.org/abs/2611.18493","checkedAt":"2026-11-16T06:36:07.250Z","items":[],"note":"No suitable media was found by the supported page extraction."},"editorialDecision":{"outcome":"accept","reasonCodes":["material_development"],"rubricVersion":"stream-admission-v9","mediaId":"01c24557b7160a8a54e5513b70aaf376373865e1b9e482bb1df333d85c4fe570","confidence":0.6799999999999999,"provider":"typesafe","model":"jev-1.13.0","inputHash":"a80b5daac57b5768ff66219c616f61ce4a3e798bfa4c3993efac08bd764cd4c9","decidedAt":"2026-11-16T06:02:53.391Z"},"provenance":{"textHash":"0f8f682376301e94d0cd1fde08d1cb2fc9455bc3719cd74d9257223c138f4b23","observationId":"e5a7c8c551bdd0cda0f9380c87f23abcb49b542cab5ebfda5cce52e0d9906dea","retrievedAt":"2026-11-16T06:02:52.612Z","kind":"publisher-page","preparer":{"model":"gpt-6-luna","runId":"8fff4288-a9b6-45ee-855a-00d3427591f9"},"reviewer":{"model":"gpt-6-luna","runId":"b8bbb490-c5dc-473a-858b-5dac0874b8d2"}},"audit":{"status":"uncertain","reviewedAt":"2026-11-16T06:04:20.036Z","model":"anthropic/claude-opus-5.5"},"sourceAttribution":{"authors":[{"name":"Nandi, Sandip","url":null,"kind":"unknown"},{"name":"Bose, Aishwarya","url":null,"kind":"unknown"},{"name":"Bose, Apurba","url":null,"kind":"unknown"},{"name":"Bakshi, Soovadeep","url":null,"kind":"unknown"}],"authorRole":"article-author","hosts":[],"status":"stated","authorRoster":{"truncated":false,"omitted":0,"assertionGroupsOmitted":0},"participation":[{"actor":{"name":"Nandi, Sandip","url":null,"kind":"unknown","actorId":null},"role":"author","scope":{"kind":"work-credit","locator":null},"literalCredit":"Nandi, Sandip","status":"stated","provenance":{"basis":"head-meta","sourceUrl":"https://arxiv.org/abs/2611.18493","locator":"head/meta[name=author|citation_author]"},"assertionId":"b7e18c0874c541060059e342e5f07447926f21c97b66b0b2f03e32a710e6b162:participation:0","identity":{"version":"entity-identity-v1","status":"unresolved","actorId":null,"canonicalActorId":null,"candidateActorIds":[],"assertions":[],"sourceObservedAt":"2026-11-16T06:36:07.250Z","knowledgeAsOf":null},"sourceVersion":{"snapshotId":"b7e18c0874c541060059e342e5f07447926f21c97b66b0b2f03e32a710e6b162","sourceUrl":"https://arxiv.org/abs/2611.18493","contentSha256":"836631e872b72a7de754bbaae7dda654b88ebe6251c07ff08c49a63e3b7e6e92","textHash":"0f8f682376301e94d0cd1fde08d1cb2fc9455bc3719cd74d9257223c138f4b23","observedAt":"2026-11-16T06:36:07.250Z"}},{"actor":{"name":"Bose, Aishwarya","url":null,"kind":"unknown","actorId":null},"role":"author","scope":{"kind":"work-credit","locator":null},"literalCredit":"Bose, Aishwarya","status":"stated","provenance":{"basis":"head-meta","sourceUrl":"https://arxiv.org/abs/2611.18493","locator":"head/meta[name=author|citation_author]"},"assertionId":"b7e18c0874c541060059e342e5f07447926f21c97b66b0b2f03e32a710e6b162:participation:1","identity":{"version":"entity-identity-v1","status":"unresolved","actorId":null,"canonicalActorId":null,"candidateActorIds":[],"assertions":[],"sourceObservedAt":"2026-11-16T06:36:07.250Z","knowledgeAsOf":null},"sourceVersion":{"snapshotId":"b7e18c0874c541060059e342e5f07447926f21c97b66b0b2f03e32a710e6b162","sourceUrl":"https://arxiv.org/abs/2611.18493","contentSha256":"836631e872b72a7de754bbaae7dda654b88ebe6251c07ff08c49a63e3b7e6e92","textHash":"0f8f682376301e94d0cd1fde08d1cb2fc9455bc3719cd74d9257223c138f4b23","observedAt":"2026-11-16T06:36:07.250Z"}},{"actor":{"name":"Bose, Apurba","url":null,"kind":"unknown","actorId":null},"role":"author","scope":{"kind":"work-credit","locator":null},"literalCredit":"Bose, Apurba","status":"stated","provenance":{"basis":"head-meta","sourceUrl":"https://arxiv.org/abs/2611.18493","locator":"head/meta[name=author|citation_author]"},"assertionId":"b7e18c0874c541060059e342e5f07447926f21c97b66b0b2f03e32a710e6b162:participation:2","identity":{"version":"entity-identity-v1","status":"unresolved","actorId":null,"canonicalActorId":null,"candidateActorIds":[],"assertions":[],"sourceObservedAt":"2026-11-16T06:36:07.250Z","knowledgeAsOf":null},"sourceVersion":{"snapshotId":"b7e18c0874c541060059e342e5f07447926f21c97b66b0b2f03e32a710e6b162","sourceUrl":"https://arxiv.org/abs/2611.18493","contentSha256":"836631e872b72a7de754bbaae7dda654b88ebe6251c07ff08c49a63e3b7e6e92","textHash":"0f8f682376301e94d0cd1fde08d1cb2fc9455bc3719cd74d9257223c138f4b23","observedAt":"2026-11-16T06:36:07.250Z"}},{"actor":{"name":"Bakshi, Soovadeep","url":null,"kind":"unknown","actorId":null},"role":"author","scope":{"kind":"work-credit","locator":null},"literalCredit":"Bakshi, Soovadeep","status":"stated","provenance":{"basis":"head-meta","sourceUrl":"https://arxiv.org/abs/2611.18493","locator":"head/meta[name=author|citation_author]"},"assertionId":"b7e18c0874c541060059e342e5f07447926f21c97b66b0b2f03e32a710e6b162:participation:3","identity":{"version":"entity-identity-v1","status":"unresolved","actorId":null,"canonicalActorId":null,"candidateActorIds":[],"assertions":[],"sourceObservedAt":"2026-11-16T06:36:07.250Z","knowledgeAsOf":null},"sourceVersion":{"snapshotId":"b7e18c0874c541060059e342e5f07447926f21c97b66b0b2f03e32a710e6b162","sourceUrl":"https://arxiv.org/abs/2611.18493","contentSha256":"836631e872b72a7de754bbaae7dda654b88ebe6251c07ff08c49a63e3b7e6e92","textHash":"0f8f682376301e94d0cd1fde08d1cb2fc9455bc3719cd74d9257223c138f4b23","observedAt":"2026-11-16T06:36:07.250Z"}}],"participationCoverage":{"total":4,"returned":4,"complete":true,"fullAssertions":"source-metadata-endpoint"},"distribution":null,"publisher":null,"sourceUrl":"https://arxiv.org/abs/2611.18493","snapshotId":"b7e18c0874c541060059e342e5f07447926f21c97b66b0b2f03e32a710e6b162","contentSha256":"836631e872b72a7de754bbaae7dda654b88ebe6251c07ff08c49a63e3b7e6e92","textHash":"0f8f682376301e94d0cd1fde08d1cb2fc9455bc3719cd74d9257223c138f4b23","observedAt":"2026-11-16T06:36:07.250Z","basis":"publisher-page-metadata","sourceLinks":44,"linksTruncated":false,"metadataUrl":"/api/pipeline/stream/554532f9461d8d312b0e9d3812d083d3f7082efc56f032e7eb53c7ae86e251f3/source"}},"canonicalUrl":"https://escapemeridian.com/readings/arxiv-paper-tests-compliance-controls-for-autonomous-trading-554532f9","formats":{"html":"/readings/arxiv-paper-tests-compliance-controls-for-autonomous-trading-554532f9","markdown":"/readings/arxiv-paper-tests-compliance-controls-for-autonomous-trading-554532f9.md","json":"/readings/arxiv-paper-tests-compliance-controls-for-autonomous-trading-554532f9.json"},"attribution":"Headline and summary by Escape Meridian, prepared by Luna from the attributed source. A source check is not independent confirmation of the source’s claims.","maintainedGuideLinks":[],"conditions":[],"readerContext":{"version":"reader-context-v1","attribution":"Reference context selected by Escape Meridian; separate from source evidence. Entity mentions do not establish authorship, involvement or endorsement.","terms":[],"entities":[{"id":"arxiv","name":"arXiv","kind":"organization","context":"arXiv is a research manuscript repository where authors share papers that may not yet have been peer reviewed.","referenceUrls":["https://arxiv.org/","https://info.arxiv.org/help/api/index.html"],"path":"/entities/arxiv","json":"/entities/arxiv.json","markdown":"/entities/arxiv.md"},{"id":"claude","name":"Claude","kind":"model","context":"Claude is Anthropic’s family of AI models and assistants for tasks such as writing, analysis and coding.","referenceUrls":["https://claude.com/product/overview","https://www.anthropic.com/news","https://www.anthropic.com/research/team/economics"],"path":"/entities/claude","json":"/entities/claude.json","markdown":"/entities/claude.md"}]},"researchQuestions":[],"withdrawn":false,"references":{"schemaVersion":1,"target":{"type":"reading","id":"554532f9461d8d312b0e9d3812d083d3f7082efc56f032e7eb53c7ae86e251f3","revision":"d2c7c446-479f-4f0e-8177-295cc538cc70","revisionKind":"publication-id","title":"arXiv paper tests compliance controls for autonomous trading agents","url":"https://escapemeridian.com/readings/arxiv-paper-tests-compliance-controls-for-autonomous-trading-554532f9"},"source":{"type":"source","id":"https://arxiv.org/abs/2611.18493","revision":"0f8f682376301e94d0cd1fde08d1cb2fc9455bc3719cd74d9257223c138f4b23","revisionKind":"sha256-source-text","title":"TradeGauntlet: Evaluating Compliance Controls for Autonomous Trading Agents","url":"https://arxiv.org/abs/2611.18493"},"status":"available","items":[],"next":null,"scope":"Explicit citations in public readings to this reading. Shared topics are not citations. Other record types need an explicit reference adapter."},"revisions":[{"id":"554532f9461d8d312b0e9d3812d083d3f7082efc56f032e7eb53c7ae86e251f3","title":"arXiv paper tests compliance controls for autonomous trading agents","argument":"The authors introduce TradeGauntlet, a replayable limit-order-book simulator that separates detection, prevention and execution cost to test controls for autonomous trading agents. In simulated sessions with Claude Haiku 5.5 and Sonnet 5.5, every conduct alert was a self-match. Gates that rejected self-crossing orders eliminated those alerts, but an aggregate-gate condition reduced Sonnet’s two-sided fills by about 15%. A conduct reminder reduced Sonnet’s alert rate, while reducing Haiku’s fills without a detectable alert benefit. The authors argue that zero alerts alone do not establish an agent’s compliance. Scripted probes and two search methods tested further weaknesses; the searches found split-order schedules that evaded the implemented conduct rules. A learned baseline caught all 2,400 tested searched episodes but also flagged 15% of a large maker’s sessions. Aggregate caps reduced response-dependent gains unevenly, with smaller reductions when only near-touch depth affected prices. A response fitted to five Nasdaq stocks and a separate ABIDES experiment expose how estimates depend on market assumptions. The authors say the experiments do not establish legal compliance or real-market detection accuracy.","topicIds":["intelligence-and-discovery"],"domain":"science","claimType":"attributed reading","limits":"The reported results come from simulated sessions and tests, not demonstrated real-market detection accuracy or established legal compliance. The authors state that estimates depend on market assumptions; the abstract describes a response fitted to five Nasdaq stocks and a separate ABIDES experiment.","speaker":{"name":"Nandi, Sandip, Bose, Aishwarya, Bose, Apurba, Bakshi, Soovadeep","roleAtSource":"Article author"},"source":{"title":"TradeGauntlet: Evaluating Compliance Controls for Autonomous Trading Agents","publisher":"arXiv cs.AI","format":"Article","url":"https://arxiv.org/abs/2611.18493","publishedAt":"2026-11-13","dateBasis":"Date basis not provided."},"evidence":[{"text":"The authors present TradeGauntlet as a replayable limit-order-book simulator for studying compliance controls for autonomous trading agents.","excerpt":"a replayable limit-order-book simulator that separates detection, prevention, and execution cost","url":"https://arxiv.org/abs/2611.18493","relationship":"Source-stated"},{"text":"In the simulated sessions, all conduct alerts were self-matches, and gates rejecting self-crossing orders eliminated those alerts.","excerpt":"all conduct alerts were self-matches. Gates eliminated those alerts by rejecting self-crossing orders","url":"https://arxiv.org/abs/2611.18493","relationship":"Source-stated"},{"text":"The aggregate-gate condition reduced Sonnet's two-sided fills by about 15%.","excerpt":"the aggregate-gate condition reduced Sonnet's two-sided fills by about 15%","url":"https://arxiv.org/abs/2611.18493","relationship":"Source-stated"},{"text":"The authors caution that zero alerts alone do not establish an agent's compliance.","excerpt":"These results show why zero alerts alone do not establish an agent's compliance.","url":"https://arxiv.org/abs/2611.18493","relationship":"Source-stated"},{"text":"Scripted probes and two search methods tested further weaknesses; the searches found split-order schedules that evaded the implemented conduct rules. A learned baseline caught all searched episodes tested but also flagged 15% of a large maker's sessions.","excerpt":"The searches find split-order schedules that evade the implemented conduct rules; a learned baseline catches all 2,400 tested searched episodes but also flags 15% of a large maker's sessions.","url":"https://arxiv.org/abs/2611.18493","relationship":"Source-stated"},{"text":"The paper says its experiments do not establish legal compliance or real-market detection accuracy.","excerpt":"the experiments do not establish legal compliance or real-market detection accuracy.","url":"https://arxiv.org/abs/2611.18493","relationship":"Source-stated"}],"sourceMedia":{"status":"not_found","sourceUrl":"https://arxiv.org/abs/2611.18493","checkedAt":"2026-11-16T06:36:07.250Z","items":[],"note":"No suitable media was found by the supported page extraction."},"assumptions":[],"openQuestion":{"attribution":"Open question","text":"How would the estimated effects change under market assumptions beyond the response fitted to five Nasdaq stocks and the separate ABIDES experiment?"},"whatWouldChangeView":{"attribution":"Change condition not recorded","text":""},"editorialComparisons":[],"live":{"title":"arXiv paper tests compliance controls for autonomous trading agents","summary":"The authors introduce TradeGauntlet, a replayable limit-order-book simulator that separates detection, prevention and execution cost to test controls for autonomous trading agents. In simulated sessions with Claude Haiku 5.5 and Sonnet 5.5, every conduct alert was a self-match. Gates that rejected self-crossing orders eliminated those alerts, but an aggregate-gate condition reduced Sonnet’s two-sided fills by about 15%. A conduct reminder reduced Sonnet’s alert rate, while reducing Haiku’s fills without a detectable alert benefit. The authors argue that zero alerts alone do not establish an agent’s compliance. Scripted probes and two search methods tested further weaknesses; the searches found split-order schedules that evaded the implemented conduct rules. A learned baseline caught all 2,400 tested searched episodes but also flagged 15% of a large maker’s sessions. Aggregate caps reduced response-dependent gains unevenly, with smaller reductions when only near-touch depth affected prices. A response fitted to five Nasdaq stocks and a separate ABIDES experiment expose how estimates depend on market assumptions. The authors say the experiments do not establish legal compliance or real-market detection accuracy.","question":"How would the estimated effects change under market assumptions beyond the response fitted to five Nasdaq stocks and the separate ABIDES experiment?","claims":[{"text":"The authors present TradeGauntlet as a replayable limit-order-book simulator for studying compliance controls for autonomous trading agents.","excerpt":"a replayable limit-order-book simulator that separates detection, prevention, and execution cost","support":"source-stated"},{"text":"In the simulated sessions, all conduct alerts were self-matches, and gates rejecting self-crossing orders eliminated those alerts.","excerpt":"all conduct alerts were self-matches. Gates eliminated those alerts by rejecting self-crossing orders","support":"source-stated"},{"text":"The aggregate-gate condition reduced Sonnet's two-sided fills by about 15%.","excerpt":"the aggregate-gate condition reduced Sonnet's two-sided fills by about 15%","support":"source-stated"},{"text":"The authors caution that zero alerts alone do not establish an agent's compliance.","excerpt":"These results show why zero alerts alone do not establish an agent's compliance.","support":"source-stated"},{"text":"Scripted probes and two search methods tested further weaknesses; the searches found split-order schedules that evaded the implemented conduct rules. A learned baseline caught all searched episodes tested but also flagged 15% of a large maker's sessions.","excerpt":"The searches find split-order schedules that evade the implemented conduct rules; a learned baseline catches all 2,400 tested searched episodes but also flags 15% of a large maker's sessions.","support":"source-stated"},{"text":"The paper says its experiments do not establish legal compliance or real-market detection accuracy.","excerpt":"the experiments do not establish legal compliance or real-market detection accuracy.","support":"source-stated"}],"limitations":["The reported results come from simulated sessions and tests, not demonstrated real-market detection accuracy or established legal compliance.","The authors state that estimates depend on market assumptions; the abstract describes a response fitted to five Nasdaq stocks and a separate ABIDES experiment."],"mediaId":"01c24557b7160a8a54e5513b70aaf376373865e1b9e482bb1df333d85c4fe570","correctionReason":null,"id":"554532f9461d8d312b0e9d3812d083d3f7082efc56f032e7eb53c7ae86e251f3","publicationId":"d2c7c446-479f-4f0e-8177-295cc538cc70","version":1,"sourceUrl":"https://arxiv.org/abs/2611.18493","sourceTitle":"TradeGauntlet: Evaluating Compliance Controls for Autonomous Trading Agents","publisher":"arXiv cs.AI","publishedAt":null,"originalPublished":{"value":"2026-11-13","precision":"day","basis":"publisher_citation_date"},"firstObservedAt":"2026-11-16T05:06:31.992Z","createdAt":"2026-11-16T06:04:20.036Z","historical":true,"entityIds":["arxiv"],"topicIds":["intelligence-and-discovery"],"media":[{"id":"01c24557b7160a8a54e5513b70aaf376373865e1b9e482bb1df333d85c4fe570","kind":"article","url":"https://arxiv.org/abs/2611.18493","sourceUrl":"https://arxiv.org/abs/2611.18493","label":"Read original","provenance":"source_link","verification":{"status":"verified","checkedAt":"2026-11-16T06:02:52.699Z","method":"acquired_source","contentType":"text/html"},"recommendationReason":"Recommended by Jev for understanding this source; checked by Luna."}],"sourceMedia":{"status":"not_found","sourceUrl":"https://arxiv.org/abs/2611.18493","checkedAt":"2026-11-16T06:36:07.250Z","items":[],"note":"No suitable media was found by the supported page extraction."},"editorialDecision":{"outcome":"accept","reasonCodes":["material_development"],"rubricVersion":"stream-admission-v9","mediaId":"01c24557b7160a8a54e5513b70aaf376373865e1b9e482bb1df333d85c4fe570","confidence":0.6799999999999999,"provider":"typesafe","model":"jev-1.13.0","inputHash":"a80b5daac57b5768ff66219c616f61ce4a3e798bfa4c3993efac08bd764cd4c9","decidedAt":"2026-11-16T06:02:53.391Z"},"provenance":{"textHash":"0f8f682376301e94d0cd1fde08d1cb2fc9455bc3719cd74d9257223c138f4b23","observationId":"e5a7c8c551bdd0cda0f9380c87f23abcb49b542cab5ebfda5cce52e0d9906dea","retrievedAt":"2026-11-16T06:02:52.612Z","kind":"publisher-page","preparer":{"model":"gpt-6-luna","runId":"8fff4288-a9b6-45ee-855a-00d3427591f9"},"reviewer":{"model":"gpt-6-luna","runId":"b8bbb490-c5dc-473a-858b-5dac0874b8d2"}},"audit":{"status":"uncertain","reviewedAt":"2026-11-16T06:04:20.036Z","model":"anthropic/claude-opus-5.5"},"sourceAttribution":{"authors":[{"name":"Nandi, Sandip","url":null,"kind":"unknown"},{"name":"Bose, Aishwarya","url":null,"kind":"unknown"},{"name":"Bose, Apurba","url":null,"kind":"unknown"},{"name":"Bakshi, Soovadeep","url":null,"kind":"unknown"}],"authorRole":"article-author","hosts":[],"status":"stated","authorRoster":{"truncated":false,"omitted":0,"assertionGroupsOmitted":0},"participation":[{"actor":{"name":"Nandi, Sandip","url":null,"kind":"unknown","actorId":null},"role":"author","scope":{"kind":"work-credit","locator":null},"literalCredit":"Nandi, Sandip","status":"stated","provenance":{"basis":"head-meta","sourceUrl":"https://arxiv.org/abs/2611.18493","locator":"head/meta[name=author|citation_author]"},"assertionId":"b7e18c0874c541060059e342e5f07447926f21c97b66b0b2f03e32a710e6b162:participation:0","identity":{"version":"entity-identity-v1","status":"unresolved","actorId":null,"canonicalActorId":null,"candidateActorIds":[],"assertions":[],"sourceObservedAt":"2026-11-16T06:36:07.250Z","knowledgeAsOf":null},"sourceVersion":{"snapshotId":"b7e18c0874c541060059e342e5f07447926f21c97b66b0b2f03e32a710e6b162","sourceUrl":"https://arxiv.org/abs/2611.18493","contentSha256":"836631e872b72a7de754bbaae7dda654b88ebe6251c07ff08c49a63e3b7e6e92","textHash":"0f8f682376301e94d0cd1fde08d1cb2fc9455bc3719cd74d9257223c138f4b23","observedAt":"2026-11-16T06:36:07.250Z"}},{"actor":{"name":"Bose, Aishwarya","url":null,"kind":"unknown","actorId":null},"role":"author","scope":{"kind":"work-credit","locator":null},"literalCredit":"Bose, Aishwarya","status":"stated","provenance":{"basis":"head-meta","sourceUrl":"https://arxiv.org/abs/2611.18493","locator":"head/meta[name=author|citation_author]"},"assertionId":"b7e18c0874c541060059e342e5f07447926f21c97b66b0b2f03e32a710e6b162:participation:1","identity":{"version":"entity-identity-v1","status":"unresolved","actorId":null,"canonicalActorId":null,"candidateActorIds":[],"assertions":[],"sourceObservedAt":"2026-11-16T06:36:07.250Z","knowledgeAsOf":null},"sourceVersion":{"snapshotId":"b7e18c0874c541060059e342e5f07447926f21c97b66b0b2f03e32a710e6b162","sourceUrl":"https://arxiv.org/abs/2611.18493","contentSha256":"836631e872b72a7de754bbaae7dda654b88ebe6251c07ff08c49a63e3b7e6e92","textHash":"0f8f682376301e94d0cd1fde08d1cb2fc9455bc3719cd74d9257223c138f4b23","observedAt":"2026-11-16T06:36:07.250Z"}},{"actor":{"name":"Bose, Apurba","url":null,"kind":"unknown","actorId":null},"role":"author","scope":{"kind":"work-credit","locator":null},"literalCredit":"Bose, Apurba","status":"stated","provenance":{"basis":"head-meta","sourceUrl":"https://arxiv.org/abs/2611.18493","locator":"head/meta[name=author|citation_author]"},"assertionId":"b7e18c0874c541060059e342e5f07447926f21c97b66b0b2f03e32a710e6b162:participation:2","identity":{"version":"entity-identity-v1","status":"unresolved","actorId":null,"canonicalActorId":null,"candidateActorIds":[],"assertions":[],"sourceObservedAt":"2026-11-16T06:36:07.250Z","knowledgeAsOf":null},"sourceVersion":{"snapshotId":"b7e18c0874c541060059e342e5f07447926f21c97b66b0b2f03e32a710e6b162","sourceUrl":"https://arxiv.org/abs/2611.18493","contentSha256":"836631e872b72a7de754bbaae7dda654b88ebe6251c07ff08c49a63e3b7e6e92","textHash":"0f8f682376301e94d0cd1fde08d1cb2fc9455bc3719cd74d9257223c138f4b23","observedAt":"2026-11-16T06:36:07.250Z"}},{"actor":{"name":"Bakshi, Soovadeep","url":null,"kind":"unknown","actorId":null},"role":"author","scope":{"kind":"work-credit","locator":null},"literalCredit":"Bakshi, Soovadeep","status":"stated","provenance":{"basis":"head-meta","sourceUrl":"https://arxiv.org/abs/2611.18493","locator":"head/meta[name=author|citation_author]"},"assertionId":"b7e18c0874c541060059e342e5f07447926f21c97b66b0b2f03e32a710e6b162:participation:3","identity":{"version":"entity-identity-v1","status":"unresolved","actorId":null,"canonicalActorId":null,"candidateActorIds":[],"assertions":[],"sourceObservedAt":"2026-11-16T06:36:07.250Z","knowledgeAsOf":null},"sourceVersion":{"snapshotId":"b7e18c0874c541060059e342e5f07447926f21c97b66b0b2f03e32a710e6b162","sourceUrl":"https://arxiv.org/abs/2611.18493","contentSha256":"836631e872b72a7de754bbaae7dda654b88ebe6251c07ff08c49a63e3b7e6e92","textHash":"0f8f682376301e94d0cd1fde08d1cb2fc9455bc3719cd74d9257223c138f4b23","observedAt":"2026-11-16T06:36:07.250Z"}}],"participationCoverage":{"total":4,"returned":4,"complete":true,"fullAssertions":"source-metadata-endpoint"},"distribution":null,"publisher":null,"sourceUrl":"https://arxiv.org/abs/2611.18493","snapshotId":"b7e18c0874c541060059e342e5f07447926f21c97b66b0b2f03e32a710e6b162","contentSha256":"836631e872b72a7de754bbaae7dda654b88ebe6251c07ff08c49a63e3b7e6e92","textHash":"0f8f682376301e94d0cd1fde08d1cb2fc9455bc3719cd74d9257223c138f4b23","observedAt":"2026-11-16T06:36:07.250Z","basis":"publisher-page-metadata","sourceLinks":44,"linksTruncated":false,"metadataUrl":"/api/pipeline/stream/554532f9461d8d312b0e9d3812d083d3f7082efc56f032e7eb53c7ae86e251f3/source"}}}],"audits":[{"publicationId":"d2c7c446-479f-4f0e-8177-295cc538cc70","verdict":"uncertain","findings":[{"dimension":"attribution","status":"pass","explanation":"The reading consistently credits the findings and caveats to the paper's authors on arXiv. The listed authors and arXiv identity match the evidence. Claims stay attributed rather than being presented as established fact.","excerpt":"Authors:Sandip Nandi, Aishwarya Bose, Apurba Bose, Soovadeep Bakshi"},{"dimension":"summary","status":"uncertain","explanation":"The summary is faithful but hard to follow without specialist knowledge. Central terms are left unexplained: limit-order-book, self-match, self-crossing orders, two-sided fills, aggregate-gate, near-touch depth, response-dependent gain, large maker and ABIDES. A general reader cannot tell why self-matching matters or what the fill-reduction trade-off means. Smallest repair: add plain explanations or tooltips for self-match and fills, plus one sentence on why blocking bad orders can cost legitimate trading. The summary also omits the 1,680-session scale and that Sonnet refused the adversarial prompt in all 240 rounds; restoring the refusal result would help.","excerpt":"A distinct adversarial prompt elicited refusals from Sonnet in all 240 rounds, yielding no candidate strategies to evaluate."},{"dimension":"date","status":"pass","explanation":"The originalPublished date of 2026-11-13 matches the arXiv v1 submission date. It is not the later feed date of 2026-11-16.","excerpt":"[Submitted on 13 Nov 2026]"},{"dimension":"support","status":"pass","explanation":"Each claim and summary sentence closely tracks the abstract, including the 15% figures, the 2,400 episodes and the dependence on market assumptions. The limitations repeat the authors' own caveats. The question follows from the stated dependence on market assumptions rather than leading the reader. No benefits or harms are invented.","excerpt":"the experiments do not establish legal compliance or real-market detection accuracy."},{"dimension":"media","status":"pass","explanation":"The selected media is the verified direct arXiv abstract page for this paper, which is the most direct access to the source.","excerpt":null}],"createdAt":"2026-11-16T06:04:20.036Z","model":"anthropic/claude-opus-5.5"}],"correctionUrl":"https://escapemeridian.com/corrections/"}