{"bundle_type":"pith_open_graph_bundle","bundle_version":"1.0","pith_number":"pith:2026:ZWDREX3OKTCODS7LHJHCN6SCJ3","short_pith_number":"pith:ZWDREX3O","canonical_record":{"source":{"id":"2605.17036","kind":"arxiv","version":1},"metadata":{"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.AI","submitted_at":"2026-05-16T15:11:35Z","cross_cats_sorted":["cs.LG","cs.MA","cs.SY","eess.SY"],"title_canon_sha256":"1eb7a4f4d7f8c595979ef1accca31a12fea55af2dce6d104b8957465a2b09f9a","abstract_canon_sha256":"6f28bfdd88d2fc0f3a3dee7e4b6715b21c75c6554a7758d40bb370656070c6be"},"schema_version":"1.0"},"canonical_sha256":"cd87125f6e54c4e1cbeb3a4e26fa424ec408c40096bf3bd2486d3afcba3cf7fd","source":{"kind":"arxiv","id":"2605.17036","version":1},"source_aliases":[{"alias_kind":"arxiv","alias_value":"2605.17036","created_at":"2026-05-20T00:03:37Z"},{"alias_kind":"arxiv_version","alias_value":"2605.17036v1","created_at":"2026-05-20T00:03:37Z"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2605.17036","created_at":"2026-05-20T00:03:37Z"},{"alias_kind":"pith_short_12","alias_value":"ZWDREX3OKTCO","created_at":"2026-05-20T00:03:37Z"},{"alias_kind":"pith_short_16","alias_value":"ZWDREX3OKTCODS7L","created_at":"2026-05-20T00:03:37Z"},{"alias_kind":"pith_short_8","alias_value":"ZWDREX3O","created_at":"2026-05-20T00:03:37Z"}],"events":[{"event_type":"record_created","subject_pith_number":"pith:2026:ZWDREX3OKTCODS7LHJHCN6SCJ3","target":"record","payload":{"canonical_record":{"source":{"id":"2605.17036","kind":"arxiv","version":1},"metadata":{"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.AI","submitted_at":"2026-05-16T15:11:35Z","cross_cats_sorted":["cs.LG","cs.MA","cs.SY","eess.SY"],"title_canon_sha256":"1eb7a4f4d7f8c595979ef1accca31a12fea55af2dce6d104b8957465a2b09f9a","abstract_canon_sha256":"6f28bfdd88d2fc0f3a3dee7e4b6715b21c75c6554a7758d40bb370656070c6be"},"schema_version":"1.0"},"canonical_sha256":"cd87125f6e54c4e1cbeb3a4e26fa424ec408c40096bf3bd2486d3afcba3cf7fd","receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-05-20T00:03:37.102116Z","signature_b64":"JibAOdWObkQBxMETflsXIoRjx950bHm4/u4llmcn4PpXzVjpFGFx9570X7OgPgCzpZoua9dJgik3XEjGFb4tCQ==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"cd87125f6e54c4e1cbeb3a4e26fa424ec408c40096bf3bd2486d3afcba3cf7fd","last_reissued_at":"2026-05-20T00:03:37.101429Z","signature_status":"signed_v1","first_computed_at":"2026-05-20T00:03:37.101429Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"source_kind":"arxiv","source_id":"2605.17036","source_version":1,"attestation_state":"computed"},"signer":{"signer_id":"pith.science","signer_type":"pith_registry","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"created_at":"2026-05-20T00:03:37Z","supersedes":[],"prev_event":null,"signature":{"signature_status":"signed_v1","algorithm":"ed25519","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","signature_b64":"JcSx+It81/V6QpgsqCRg9KO1JLsktRd3EpoqW8cEcL4L/pT1cWpp6IAECGN/KS/tkgI+IHPPckpSxsj9rlWiAw==","signed_message":"open_graph_event_sha256_bytes","signed_at":"2026-05-21T11:08:16.448913Z"},"content_sha256":"57aac237465a0dfb36bbce62c87e169ebae87a7cecb79263ed56f231c2c98fe6","schema_version":"1.0","event_id":"sha256:57aac237465a0dfb36bbce62c87e169ebae87a7cecb79263ed56f231c2c98fe6"},{"event_type":"graph_snapshot","subject_pith_number":"pith:2026:ZWDREX3OKTCODS7LHJHCN6SCJ3","target":"graph","payload":{"graph_snapshot":{"paper":{"title":"Reliability and Effectiveness of Autonomous AI Agents in Supply Chain Management","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"Autonomous AI agents with optimized reasoning models outperform human teams in supply chain management by reducing costs up to 67 percent, but require post-training to control decision unreliability.","cross_cats":["cs.LG","cs.MA","cs.SY","eess.SY"],"primary_cat":"cs.AI","authors_text":"Andre P. Calmon, Carol Xuan Long, David Simchi-Levi, Feng Zhu, Flavio P. Calmon, Huangyuan Su","submitted_at":"2026-05-16T15:11:35Z","abstract_excerpt":"This paper studies autonomous generative AI agents in multi-echelon supply chains using the MIT Beer Game. We identify four inference-time levers that shape performance: model selection, policies and guardrails, centralized data sharing, and prompt engineering. Model capability is the dominant factor: an out-of-the-box reasoning model exceeds human-level performance, and optimized reasoning models reduce costs by up to 67% relative to human teams. However, strong average performance masks substantial reliability risks. We introduce the agent bullwhip effect, the amplification of decision unrel"},"claims":{"count":4,"items":[{"kind":"strongest_claim","text":"Model capability is the dominant factor: an out-of-the-box reasoning model exceeds human-level performance, and optimized reasoning models reduce costs by up to 67% relative to human teams. GRPO post-training substantially reduces tail events, curtails agent bullwhip, and improves the reliability of autonomous supply-chain agents.","source":"verdict.strongest_claim","status":"machine_extracted","claim_id":"C1","attestation":"unclaimed"},{"kind":"weakest_assumption","text":"The MIT Beer Game simulation with its ordering rules and information delays sufficiently captures the coordination dynamics of real multi-echelon supply chains (the entire experimental and theoretical framework is constructed on this testbed).","source":"verdict.weakest_assumption","status":"machine_extracted","claim_id":"C2","attestation":"unclaimed"},{"kind":"one_line_summary","text":"Autonomous AI agents outperform humans in supply chain simulations but exhibit an inherent agent bullwhip effect of amplified decision unreliability, mitigated by GRPO reinforcement learning post-training.","source":"verdict.one_line_summary","status":"machine_extracted","claim_id":"C3","attestation":"unclaimed"},{"kind":"headline","text":"Autonomous AI agents with optimized reasoning models outperform human teams in supply chain management by reducing costs up to 67 percent, but require post-training to control decision unreliability.","source":"verdict.pith_extraction.headline","status":"machine_extracted","claim_id":"C4","attestation":"unclaimed"}],"snapshot_sha256":"d42ede4a3fbdc781deb3268be42be4ab6b4f5e795789a128a31eea46c23110b4"},"source":{"id":"2605.17036","kind":"arxiv","version":1},"verdict":{"id":"16948f34-9179-43f3-bd71-b961b42af81c","model_set":{"reader":"grok-4.3"},"created_at":"2026-05-19T20:09:12.106368Z","strongest_claim":"Model capability is the dominant factor: an out-of-the-box reasoning model exceeds human-level performance, and optimized reasoning models reduce costs by up to 67% relative to human teams. GRPO post-training substantially reduces tail events, curtails agent bullwhip, and improves the reliability of autonomous supply-chain agents.","one_line_summary":"Autonomous AI agents outperform humans in supply chain simulations but exhibit an inherent agent bullwhip effect of amplified decision unreliability, mitigated by GRPO reinforcement learning post-training.","pipeline_version":"pith-pipeline@v0.9.0","weakest_assumption":"The MIT Beer Game simulation with its ordering rules and information delays sufficiently captures the coordination dynamics of real multi-echelon supply chains (the entire experimental and theoretical framework is constructed on this testbed).","pith_extraction_headline":"Autonomous AI agents with optimized reasoning models outperform human teams in supply chain management by reducing costs up to 67 percent, but require post-training to control decision unreliability."},"integrity":{"clean":false,"summary":{"advisory":1,"critical":0,"by_detector":{"doi_compliance":{"total":1,"advisory":1,"critical":0,"informational":0}},"informational":0},"endpoint":"/pith/2605.17036/integrity.json","findings":[{"note":"DOI in the printed bibliography is fragmented by whitespace or line breaks. A longer candidate (10.1080/24725854.2023.2217248.Niki) was visible in the surrounding text but could not be confirmed against doi.org as printed.","detector":"doi_compliance","severity":"advisory","ref_index":7,"audited_at":"2026-05-19T20:20:45.983094Z","detected_doi":"10.1080/24725854.2023.2217248.Niki","finding_type":"recoverable_identifier","verdict_class":"incontrovertible","detected_arxiv_id":null}],"available":true,"detectors_run":[{"name":"doi_title_agreement","ran_at":"2026-05-19T20:31:18.944638Z","status":"completed","version":"1.0.0","findings_count":0},{"name":"doi_compliance","ran_at":"2026-05-19T20:20:45.983094Z","status":"completed","version":"1.0.0","findings_count":1},{"name":"citation_quote_validity","ran_at":"2026-05-19T19:49:42.513061Z","status":"skipped","version":"0.1.0","findings_count":0},{"name":"cited_work_retraction","ran_at":"2026-05-19T18:51:56.618092Z","status":"completed","version":"1.0.0","findings_count":0},{"name":"claim_evidence","ran_at":"2026-05-19T18:41:56.167281Z","status":"completed","version":"1.0.0","findings_count":0},{"name":"ai_meta_artifact","ran_at":"2026-05-19T18:33:23.003121Z","status":"skipped","version":"1.0.0","findings_count":0}],"snapshot_sha256":"65d34662547aec51f79cacf44c54a0e7be83bedab9899589be08edd501c0ea95"},"references":{"count":18,"sample":[{"doi":"10.1287/opre.1050.0238","year":null,"title":"Leonard Boussioux, Andrew Chen, Ming Fan, and Apurva Jain","work_id":"a4216f2f-82cb-4c8b-a9e0-5d3d2a3ab49f","ref_index":1,"cited_arxiv_id":"","is_internal_anchor":false},{"doi":"","year":null,"title":"Large Language Monkeys: Scaling Inference Compute with Repeated Sampling","work_id":"b124064d-5a56-42ad-86f5-3cc349b86a3a","ref_index":2,"cited_arxiv_id":"2407.21787","is_internal_anchor":true},{"doi":"10.1023/a:1008195614074","year":null,"title":"Guillermo Gallego and Ilkyeong Moon","work_id":"fc272828-a718-469d-b2d1-3680f2d0c570","ref_index":3,"cited_arxiv_id":"","is_internal_anchor":false},{"doi":"10.1057/jors.1993.141","year":1993,"title":"Javier García and Fernando Fernández","work_id":"07253e14-faec-4b44-8a93-06fa7e6ae69b","ref_index":4,"cited_arxiv_id":"","is_internal_anchor":false},{"doi":"10.1038/s41586-025-09422-z","year":null,"title":"Nature645(8081), 633–638 (2025) https://doi.org/10.1038/s41586-025-09422-z","work_id":"9835b482-5032-4135-93dd-82a066677569","ref_index":5,"cited_arxiv_id":"","is_internal_anchor":false}],"resolved_work":18,"snapshot_sha256":"8bb015b5f488f3e1ef6399198ba749f4208f0afd0d7c6ac730f7fc52bbda782f","internal_anchors":4},"formal_canon":{"evidence_count":2,"snapshot_sha256":"89005928ddb4780468bdef62de7f289bfc1a6c229f0068ffea4b63eaafa79bf1"},"author_claims":{"count":0,"strong_count":0,"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"builder_version":"pith-number-builder-2026-05-17-v1"},"verdict_id":"16948f34-9179-43f3-bd71-b961b42af81c"},"signer":{"signer_id":"pith.science","signer_type":"pith_registry","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"created_at":"2026-05-20T00:03:37Z","supersedes":[],"prev_event":null,"signature":{"signature_status":"signed_v1","algorithm":"ed25519","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","signature_b64":"urFZRpdgAH5hx3BpL/3Fbv9/l3ZhosVGBapH/zVEC1pPGDXjmMwJ+SiZ/3izNSndOtnEKsXnLa2JQAEnHO9BAw==","signed_message":"open_graph_event_sha256_bytes","signed_at":"2026-05-21T11:08:16.449789Z"},"content_sha256":"2aea3d424d8ded9cd147c33dcb3e88f83d195d0176c91279c49ff13f5f09c2e8","schema_version":"1.0","event_id":"sha256:2aea3d424d8ded9cd147c33dcb3e88f83d195d0176c91279c49ff13f5f09c2e8"},{"event_type":"integrity_finding","subject_pith_number":"pith:2026:ZWDREX3OKTCODS7LHJHCN6SCJ3","target":"integrity","payload":{"note":"DOI in the printed bibliography is fragmented by whitespace or line breaks. A longer candidate (10.1080/24725854.2023.2217248.Niki) was visible in the surrounding text but could not be confirmed against doi.org as printed.","snippet":"doi: 10.1080/24725854. 2023.2217248. Niki Kotecha and Antonio del Rio Chanona. Leveraging graph neural networks and multi-agent reinforcement learning for inventory control in supply chains.Computers & Chemical Engineering, 199:109111,","arxiv_id":"2605.17036","detector":"doi_compliance","evidence":{"ref_index":7,"verdict_class":"incontrovertible","resolved_title":null,"printed_excerpt":"doi: 10.1080/24725854. 2023.2217248. Niki Kotecha and Antonio del Rio Chanona. Leveraging graph neural networks and multi-agent reinforcement learning for inventory control in supply chains.Computers & Chemical Engineering, 199:109111,","reconstructed_doi":"10.1080/24725854.2023.2217248.Niki"},"severity":"advisory","ref_index":7,"audited_at":"2026-05-19T20:20:45.983094Z","event_type":"pith.integrity.v1","detected_doi":"10.1080/24725854.2023.2217248.Niki","detector_url":"https://pith.science/pith-integrity-protocol#doi_compliance","external_url":null,"finding_type":"recoverable_identifier","evidence_hash":"c3567abaf18445f5e3068dd53ea11a1a3c60b98f67ee8d47cf2b69e5aafa9d42","paper_version":1,"verdict_class":"incontrovertible","resolved_title":null,"detector_version":"1.0.0","detected_arxiv_id":null,"integrity_event_id":2878,"payload_sha256":"d847833efb1a04945d33801e49fc045fdfaa6a9dc9630b310f62bd8be4b16d31","signature_b64":"XwnRAcje21D9D0OgzP/tzCJCh0uIbjyFm5eApO3meDahjlZczedpeKaTTGltPWZhz92L1Enc3lJvRvQlKy1ICw==","signing_key_id":"pith-v1-2026-05"},"signer":{"signer_id":"pith.science","signer_type":"pith_registry","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"created_at":"2026-05-19T20:22:21Z","supersedes":[],"prev_event":null,"signature":{"signature_status":"signed_v1","algorithm":"ed25519","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","signature_b64":"IEI/GQkqzLDMn9XTgIFB95VTlL+dOFnEO6GtC5pkGv3b6q03PGZRIMFNynAB8yrk4jha7PGaV4WmUjgiJQFYCg==","signed_message":"open_graph_event_sha256_bytes","signed_at":"2026-05-21T11:08:16.450679Z"},"content_sha256":"8a68f38171b03d12629d3bb7a66ef697c07902f3e1ca3b2c1bd79e7a7d03689d","schema_version":"1.0","event_id":"sha256:8a68f38171b03d12629d3bb7a66ef697c07902f3e1ca3b2c1bd79e7a7d03689d"}],"timestamp_proofs":[],"mirror_hints":[{"mirror_type":"https","name":"Pith Resolver","base_url":"https://pith.science","bundle_url":"https://pith.science/pith/ZWDREX3OKTCODS7LHJHCN6SCJ3/bundle.json","state_url":"https://pith.science/pith/ZWDREX3OKTCODS7LHJHCN6SCJ3/state.json","well_known_bundle_url":"https://pith.science/.well-known/pith/ZWDREX3OKTCODS7LHJHCN6SCJ3/bundle.json","status":"primary"}],"public_keys":[{"key_id":"pith-v1-2026-05","algorithm":"ed25519","format":"raw","public_key_b64":"stVStoiQhXFxp4s2pdzPNoqVNBMojDU/fJ2db5S3CbM=","public_key_hex":"b2d552b68890857171a78b36a5dccf368a953413288c353f7c9d9d6f94b709b3","fingerprint_sha256_b32_first128bits":"RVFV5Z2OI2J3ZUO7ERDEBCYNKS","fingerprint_sha256_hex":"8d4b5ee74e4693bcd1df2446408b0d54","rotates_at":null,"url":"https://pith.science/pith-signing-key.json","notes":"Pith uses this Ed25519 key to sign canonical record SHA-256 digests. Verify with: ed25519_verify(public_key, message=canonical_sha256_bytes, signature=base64decode(signature_b64))."}],"merge_version":"pith-open-graph-merge-v1","built_at":"2026-05-21T11:08:16Z","links":{"resolver":"https://pith.science/pith/ZWDREX3OKTCODS7LHJHCN6SCJ3","bundle":"https://pith.science/pith/ZWDREX3OKTCODS7LHJHCN6SCJ3/bundle.json","state":"https://pith.science/pith/ZWDREX3OKTCODS7LHJHCN6SCJ3/state.json","well_known_bundle":"https://pith.science/.well-known/pith/ZWDREX3OKTCODS7LHJHCN6SCJ3/bundle.json"},"state":{"state_type":"pith_open_graph_state","state_version":"1.0","pith_number":"pith:2026:ZWDREX3OKTCODS7LHJHCN6SCJ3","merge_version":"pith-open-graph-merge-v1","event_count":3,"valid_event_count":3,"invalid_event_count":0,"equivocation_count":0,"current":{"canonical_record":{"metadata":{"abstract_canon_sha256":"6f28bfdd88d2fc0f3a3dee7e4b6715b21c75c6554a7758d40bb370656070c6be","cross_cats_sorted":["cs.LG","cs.MA","cs.SY","eess.SY"],"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.AI","submitted_at":"2026-05-16T15:11:35Z","title_canon_sha256":"1eb7a4f4d7f8c595979ef1accca31a12fea55af2dce6d104b8957465a2b09f9a"},"schema_version":"1.0","source":{"id":"2605.17036","kind":"arxiv","version":1}},"source_aliases":[{"alias_kind":"arxiv","alias_value":"2605.17036","created_at":"2026-05-20T00:03:37Z"},{"alias_kind":"arxiv_version","alias_value":"2605.17036v1","created_at":"2026-05-20T00:03:37Z"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2605.17036","created_at":"2026-05-20T00:03:37Z"},{"alias_kind":"pith_short_12","alias_value":"ZWDREX3OKTCO","created_at":"2026-05-20T00:03:37Z"},{"alias_kind":"pith_short_16","alias_value":"ZWDREX3OKTCODS7L","created_at":"2026-05-20T00:03:37Z"},{"alias_kind":"pith_short_8","alias_value":"ZWDREX3O","created_at":"2026-05-20T00:03:37Z"}],"graph_snapshots":[{"event_id":"sha256:2aea3d424d8ded9cd147c33dcb3e88f83d195d0176c91279c49ff13f5f09c2e8","target":"graph","created_at":"2026-05-20T00:03:37Z","signer":{"key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","signer_id":"pith.science","signer_type":"pith_registry"},"payload":{"graph_snapshot":{"author_claims":{"count":0,"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57","strong_count":0},"builder_version":"pith-number-builder-2026-05-17-v1","claims":{"count":4,"items":[{"attestation":"unclaimed","claim_id":"C1","kind":"strongest_claim","source":"verdict.strongest_claim","status":"machine_extracted","text":"Model capability is the dominant factor: an out-of-the-box reasoning model exceeds human-level performance, and optimized reasoning models reduce costs by up to 67% relative to human teams. GRPO post-training substantially reduces tail events, curtails agent bullwhip, and improves the reliability of autonomous supply-chain agents."},{"attestation":"unclaimed","claim_id":"C2","kind":"weakest_assumption","source":"verdict.weakest_assumption","status":"machine_extracted","text":"The MIT Beer Game simulation with its ordering rules and information delays sufficiently captures the coordination dynamics of real multi-echelon supply chains (the entire experimental and theoretical framework is constructed on this testbed)."},{"attestation":"unclaimed","claim_id":"C3","kind":"one_line_summary","source":"verdict.one_line_summary","status":"machine_extracted","text":"Autonomous AI agents outperform humans in supply chain simulations but exhibit an inherent agent bullwhip effect of amplified decision unreliability, mitigated by GRPO reinforcement learning post-training."},{"attestation":"unclaimed","claim_id":"C4","kind":"headline","source":"verdict.pith_extraction.headline","status":"machine_extracted","text":"Autonomous AI agents with optimized reasoning models outperform human teams in supply chain management by reducing costs up to 67 percent, but require post-training to control decision unreliability."}],"snapshot_sha256":"d42ede4a3fbdc781deb3268be42be4ab6b4f5e795789a128a31eea46c23110b4"},"formal_canon":{"evidence_count":2,"snapshot_sha256":"89005928ddb4780468bdef62de7f289bfc1a6c229f0068ffea4b63eaafa79bf1"},"integrity":{"available":true,"clean":false,"detectors_run":[{"findings_count":0,"name":"doi_title_agreement","ran_at":"2026-05-19T20:31:18.944638Z","status":"completed","version":"1.0.0"},{"findings_count":1,"name":"doi_compliance","ran_at":"2026-05-19T20:20:45.983094Z","status":"completed","version":"1.0.0"},{"findings_count":0,"name":"citation_quote_validity","ran_at":"2026-05-19T19:49:42.513061Z","status":"skipped","version":"0.1.0"},{"findings_count":0,"name":"cited_work_retraction","ran_at":"2026-05-19T18:51:56.618092Z","status":"completed","version":"1.0.0"},{"findings_count":0,"name":"claim_evidence","ran_at":"2026-05-19T18:41:56.167281Z","status":"completed","version":"1.0.0"},{"findings_count":0,"name":"ai_meta_artifact","ran_at":"2026-05-19T18:33:23.003121Z","status":"skipped","version":"1.0.0"}],"endpoint":"/pith/2605.17036/integrity.json","findings":[{"audited_at":"2026-05-19T20:20:45.983094Z","detected_arxiv_id":null,"detected_doi":"10.1080/24725854.2023.2217248.Niki","detector":"doi_compliance","finding_type":"recoverable_identifier","note":"DOI in the printed bibliography is fragmented by whitespace or line breaks. A longer candidate (10.1080/24725854.2023.2217248.Niki) was visible in the surrounding text but could not be confirmed against doi.org as printed.","ref_index":7,"severity":"advisory","verdict_class":"incontrovertible"}],"snapshot_sha256":"65d34662547aec51f79cacf44c54a0e7be83bedab9899589be08edd501c0ea95","summary":{"advisory":1,"by_detector":{"doi_compliance":{"advisory":1,"critical":0,"informational":0,"total":1}},"critical":0,"informational":0}},"paper":{"abstract_excerpt":"This paper studies autonomous generative AI agents in multi-echelon supply chains using the MIT Beer Game. We identify four inference-time levers that shape performance: model selection, policies and guardrails, centralized data sharing, and prompt engineering. Model capability is the dominant factor: an out-of-the-box reasoning model exceeds human-level performance, and optimized reasoning models reduce costs by up to 67% relative to human teams. However, strong average performance masks substantial reliability risks. We introduce the agent bullwhip effect, the amplification of decision unrel","authors_text":"Andre P. Calmon, Carol Xuan Long, David Simchi-Levi, Feng Zhu, Flavio P. Calmon, Huangyuan Su","cross_cats":["cs.LG","cs.MA","cs.SY","eess.SY"],"headline":"Autonomous AI agents with optimized reasoning models outperform human teams in supply chain management by reducing costs up to 67 percent, but require post-training to control decision unreliability.","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.AI","submitted_at":"2026-05-16T15:11:35Z","title":"Reliability and Effectiveness of Autonomous AI Agents in Supply Chain Management"},"references":{"count":18,"internal_anchors":4,"resolved_work":18,"sample":[{"cited_arxiv_id":"","doi":"10.1287/opre.1050.0238","is_internal_anchor":false,"ref_index":1,"title":"Leonard Boussioux, Andrew Chen, Ming Fan, and Apurva Jain","work_id":"a4216f2f-82cb-4c8b-a9e0-5d3d2a3ab49f","year":null},{"cited_arxiv_id":"2407.21787","doi":"","is_internal_anchor":true,"ref_index":2,"title":"Large Language Monkeys: Scaling Inference Compute with Repeated Sampling","work_id":"b124064d-5a56-42ad-86f5-3cc349b86a3a","year":null},{"cited_arxiv_id":"","doi":"10.1023/a:1008195614074","is_internal_anchor":false,"ref_index":3,"title":"Guillermo Gallego and Ilkyeong Moon","work_id":"fc272828-a718-469d-b2d1-3680f2d0c570","year":null},{"cited_arxiv_id":"","doi":"10.1057/jors.1993.141","is_internal_anchor":false,"ref_index":4,"title":"Javier García and Fernando Fernández","work_id":"07253e14-faec-4b44-8a93-06fa7e6ae69b","year":1993},{"cited_arxiv_id":"","doi":"10.1038/s41586-025-09422-z","is_internal_anchor":false,"ref_index":5,"title":"Nature645(8081), 633–638 (2025) https://doi.org/10.1038/s41586-025-09422-z","work_id":"9835b482-5032-4135-93dd-82a066677569","year":null}],"snapshot_sha256":"8bb015b5f488f3e1ef6399198ba749f4208f0afd0d7c6ac730f7fc52bbda782f"},"source":{"id":"2605.17036","kind":"arxiv","version":1},"verdict":{"created_at":"2026-05-19T20:09:12.106368Z","id":"16948f34-9179-43f3-bd71-b961b42af81c","model_set":{"reader":"grok-4.3"},"one_line_summary":"Autonomous AI agents outperform humans in supply chain simulations but exhibit an inherent agent bullwhip effect of amplified decision unreliability, mitigated by GRPO reinforcement learning post-training.","pipeline_version":"pith-pipeline@v0.9.0","pith_extraction_headline":"Autonomous AI agents with optimized reasoning models outperform human teams in supply chain management by reducing costs up to 67 percent, but require post-training to control decision unreliability.","strongest_claim":"Model capability is the dominant factor: an out-of-the-box reasoning model exceeds human-level performance, and optimized reasoning models reduce costs by up to 67% relative to human teams. GRPO post-training substantially reduces tail events, curtails agent bullwhip, and improves the reliability of autonomous supply-chain agents.","weakest_assumption":"The MIT Beer Game simulation with its ordering rules and information delays sufficiently captures the coordination dynamics of real multi-echelon supply chains (the entire experimental and theoretical framework is constructed on this testbed)."}},"verdict_id":"16948f34-9179-43f3-bd71-b961b42af81c"}}],"author_attestations":[],"timestamp_anchors":[],"storage_attestations":[],"citation_signatures":[],"replication_records":[],"corrections":[],"mirror_hints":[],"record_created":{"event_id":"sha256:57aac237465a0dfb36bbce62c87e169ebae87a7cecb79263ed56f231c2c98fe6","target":"record","created_at":"2026-05-20T00:03:37Z","signer":{"key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","signer_id":"pith.science","signer_type":"pith_registry"},"payload":{"attestation_state":"computed","canonical_record":{"metadata":{"abstract_canon_sha256":"6f28bfdd88d2fc0f3a3dee7e4b6715b21c75c6554a7758d40bb370656070c6be","cross_cats_sorted":["cs.LG","cs.MA","cs.SY","eess.SY"],"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.AI","submitted_at":"2026-05-16T15:11:35Z","title_canon_sha256":"1eb7a4f4d7f8c595979ef1accca31a12fea55af2dce6d104b8957465a2b09f9a"},"schema_version":"1.0","source":{"id":"2605.17036","kind":"arxiv","version":1}},"canonical_sha256":"cd87125f6e54c4e1cbeb3a4e26fa424ec408c40096bf3bd2486d3afcba3cf7fd","receipt":{"algorithm":"ed25519","builder_version":"pith-number-builder-2026-05-17-v1","canonical_sha256":"cd87125f6e54c4e1cbeb3a4e26fa424ec408c40096bf3bd2486d3afcba3cf7fd","first_computed_at":"2026-05-20T00:03:37.101429Z","key_id":"pith-v1-2026-05","kind":"pith_receipt","last_reissued_at":"2026-05-20T00:03:37.101429Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","receipt_version":"0.3","signature_b64":"JibAOdWObkQBxMETflsXIoRjx950bHm4/u4llmcn4PpXzVjpFGFx9570X7OgPgCzpZoua9dJgik3XEjGFb4tCQ==","signature_status":"signed_v1","signed_at":"2026-05-20T00:03:37.102116Z","signed_message":"canonical_sha256_bytes"},"source_id":"2605.17036","source_kind":"arxiv","source_version":1}}},"equivocations":[],"invalid_events":[],"applied_event_ids":["sha256:8a68f38171b03d12629d3bb7a66ef697c07902f3e1ca3b2c1bd79e7a7d03689d","sha256:57aac237465a0dfb36bbce62c87e169ebae87a7cecb79263ed56f231c2c98fe6","sha256:2aea3d424d8ded9cd147c33dcb3e88f83d195d0176c91279c49ff13f5f09c2e8"],"state_sha256":"80fbd2897e37bc5cf8052474f77cbff6d49748f1e8192c23cb8b5c36341f03e0"},"bundle_signature":{"signature_status":"signed_v1","algorithm":"ed25519","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","signature_b64":"kG0q+VH8crjxFN4hnJwZ2a3eEjMTgh+3+twkPwMIBlq5pcfgK63/vHRbGLa8sXmQ1MlTuq4nR08Ag5RcUvZWAA==","signed_message":"bundle_sha256_bytes","signed_at":"2026-05-21T11:08:16.453523Z","bundle_sha256":"b2cb734520b4f475194b04d4b02ebf7cf251c2fcaedeee260c22162d0c1ddf61"}}