{"bundle_type":"pith_open_graph_bundle","bundle_version":"1.0","pith_number":"pith:2026:XXKSVFJ2J7UCFHST2TXIEVX7YA","short_pith_number":"pith:XXKSVFJ2","canonical_record":{"source":{"id":"2605.24202","kind":"arxiv","version":1},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.AI","submitted_at":"2026-05-22T20:43:30Z","cross_cats_sorted":["cs.LG"],"title_canon_sha256":"86c05f403a05cb6001c3dc555d0c585830b0a1be68f8f9e2adcce882935f269c","abstract_canon_sha256":"a4e598bf07f2dbb30cc185555e60976e151d497eb49acca96cabc8b20ce8432e"},"schema_version":"1.0"},"canonical_sha256":"bdd52a953a4fe8229e53d4ee8256ffc034fc9912761564d257b34266626bd091","source":{"kind":"arxiv","id":"2605.24202","version":1},"source_aliases":[{"alias_kind":"arxiv","alias_value":"2605.24202","created_at":"2026-05-26T01:02:52Z"},{"alias_kind":"arxiv_version","alias_value":"2605.24202v1","created_at":"2026-05-26T01:02:52Z"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2605.24202","created_at":"2026-05-26T01:02:52Z"},{"alias_kind":"pith_short_12","alias_value":"XXKSVFJ2J7UC","created_at":"2026-05-26T01:02:52Z"},{"alias_kind":"pith_short_16","alias_value":"XXKSVFJ2J7UCFHST","created_at":"2026-05-26T01:02:52Z"},{"alias_kind":"pith_short_8","alias_value":"XXKSVFJ2","created_at":"2026-05-26T01:02:52Z"}],"events":[{"event_type":"record_created","subject_pith_number":"pith:2026:XXKSVFJ2J7UCFHST2TXIEVX7YA","target":"record","payload":{"canonical_record":{"source":{"id":"2605.24202","kind":"arxiv","version":1},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.AI","submitted_at":"2026-05-22T20:43:30Z","cross_cats_sorted":["cs.LG"],"title_canon_sha256":"86c05f403a05cb6001c3dc555d0c585830b0a1be68f8f9e2adcce882935f269c","abstract_canon_sha256":"a4e598bf07f2dbb30cc185555e60976e151d497eb49acca96cabc8b20ce8432e"},"schema_version":"1.0"},"canonical_sha256":"bdd52a953a4fe8229e53d4ee8256ffc034fc9912761564d257b34266626bd091","receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-05-26T01:02:52.201996Z","signature_b64":"Khwb9QPYUuaFuUGPm03TCMma+L8VTOdEXmAP7FBbxUfWf8NJK8Of6JeFBVAKjASx8nuiAtOQStoln5S9rzJeCg==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"bdd52a953a4fe8229e53d4ee8256ffc034fc9912761564d257b34266626bd091","last_reissued_at":"2026-05-26T01:02:52.201166Z","signature_status":"signed_v1","first_computed_at":"2026-05-26T01:02:52.201166Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"source_kind":"arxiv","source_id":"2605.24202","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":"2026-05-26T01:02:52Z","supersedes":[],"prev_event":null,"signature":{"signature_status":"signed_v1","algorithm":"ed25519","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","signature_b64":"WjDWJ090op2qGmG3i+U0kPWYz/Vwr6rdwwymhRk0Xd6WOXb1h/IIkj7XwkL1ZU8MoBS4xQo7FQ9z6WC+SR83Ag==","signed_message":"open_graph_event_sha256_bytes","signed_at":"2026-05-31T01:33:01.246603Z"},"content_sha256":"ca48644667f7099b79de3464f1a356835c1719696d321b3408ae3fc9b8eb9cc5","schema_version":"1.0","event_id":"sha256:ca48644667f7099b79de3464f1a356835c1719696d321b3408ae3fc9b8eb9cc5"},{"event_type":"graph_snapshot","subject_pith_number":"pith:2026:XXKSVFJ2J7UCFHST2TXIEVX7YA","target":"graph","payload":{"graph_snapshot":{"paper":{"title":"When Does Multi-Agent RL Improve LLM Workflows? Workflow, Scale, and Policy-Sharing Tradeoffs","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.LG"],"primary_cat":"cs.AI","authors_text":"Huazheng Wang, Kun Wan, Qingyun Wu, Wentian Zhao, Yaolun Zhang, Yifan Zeng, Yiran Wu","submitted_at":"2026-05-22T20:43:30Z","abstract_excerpt":"Multi-agent LLM workflows route inference through specialized roles to lift end-task accuracy, but jointly training those roles with reinforcement learning is unstable in ways that are poorly understood. We study when end-to-end RL training of multi-agent LLM workflows improves over their base models, comparing Shared-Policy training, where all roles update one policy, with Isolated-Policy training, where each role has its own parameters. Our experimental matrix spans Eval-Opt, Voting, and Orch-Workers workflows, math and code tasks, and three model scales (0.6B, 1.7B, 4B). We find that multi-"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2605.24202","kind":"arxiv","version":1},"verdict":{"id":null,"model_set":{},"created_at":null,"strongest_claim":"","one_line_summary":"","pipeline_version":null,"weakest_assumption":"","pith_extraction_headline":""},"integrity":{"clean":true,"summary":{"advisory":0,"critical":0,"by_detector":{},"informational":0},"endpoint":"/pith/2605.24202/integrity.json","findings":[],"available":true,"detectors_run":[],"snapshot_sha256":"c28c3603d3b5d939e8dc4c7e95fa8dfce3d595e45f758748cecf8e644a296938"},"references":{"count":0,"sample":[],"resolved_work":0,"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57","internal_anchors":0},"formal_canon":{"evidence_count":0,"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"author_claims":{"count":0,"strong_count":0,"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"builder_version":"pith-number-builder-2026-05-17-v1"},"verdict_id":null},"signer":{"signer_id":"pith.science","signer_type":"pith_registry","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"created_at":"2026-05-26T01:02:52Z","supersedes":[],"prev_event":null,"signature":{"signature_status":"signed_v1","algorithm":"ed25519","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","signature_b64":"Acz0edNA4P4TVY0rZxsZBDMFwMaSKERSvWSCzcal42Jo6stFNAnpI5mDsEMKTyh/ay3js45JqB2XsHrD9PxvCw==","signed_message":"open_graph_event_sha256_bytes","signed_at":"2026-05-31T01:33:01.247368Z"},"content_sha256":"686e9b02b718229aa2d74e407ddf0e0730530cf0bc5d47feda857c4798c94d99","schema_version":"1.0","event_id":"sha256:686e9b02b718229aa2d74e407ddf0e0730530cf0bc5d47feda857c4798c94d99"}],"timestamp_proofs":[],"mirror_hints":[{"mirror_type":"https","name":"Pith Resolver","base_url":"https://pith.science","bundle_url":"https://pith.science/pith/XXKSVFJ2J7UCFHST2TXIEVX7YA/bundle.json","state_url":"https://pith.science/pith/XXKSVFJ2J7UCFHST2TXIEVX7YA/state.json","well_known_bundle_url":"https://pith.science/.well-known/pith/XXKSVFJ2J7UCFHST2TXIEVX7YA/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":"2026-05-31T01:33:01Z","links":{"resolver":"https://pith.science/pith/XXKSVFJ2J7UCFHST2TXIEVX7YA","bundle":"https://pith.science/pith/XXKSVFJ2J7UCFHST2TXIEVX7YA/bundle.json","state":"https://pith.science/pith/XXKSVFJ2J7UCFHST2TXIEVX7YA/state.json","well_known_bundle":"https://pith.science/.well-known/pith/XXKSVFJ2J7UCFHST2TXIEVX7YA/bundle.json"},"state":{"state_type":"pith_open_graph_state","state_version":"1.0","pith_number":"pith:2026:XXKSVFJ2J7UCFHST2TXIEVX7YA","merge_version":"pith-open-graph-merge-v1","event_count":2,"valid_event_count":2,"invalid_event_count":0,"equivocation_count":0,"current":{"canonical_record":{"metadata":{"abstract_canon_sha256":"a4e598bf07f2dbb30cc185555e60976e151d497eb49acca96cabc8b20ce8432e","cross_cats_sorted":["cs.LG"],"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.AI","submitted_at":"2026-05-22T20:43:30Z","title_canon_sha256":"86c05f403a05cb6001c3dc555d0c585830b0a1be68f8f9e2adcce882935f269c"},"schema_version":"1.0","source":{"id":"2605.24202","kind":"arxiv","version":1}},"source_aliases":[{"alias_kind":"arxiv","alias_value":"2605.24202","created_at":"2026-05-26T01:02:52Z"},{"alias_kind":"arxiv_version","alias_value":"2605.24202v1","created_at":"2026-05-26T01:02:52Z"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2605.24202","created_at":"2026-05-26T01:02:52Z"},{"alias_kind":"pith_short_12","alias_value":"XXKSVFJ2J7UC","created_at":"2026-05-26T01:02:52Z"},{"alias_kind":"pith_short_16","alias_value":"XXKSVFJ2J7UCFHST","created_at":"2026-05-26T01:02:52Z"},{"alias_kind":"pith_short_8","alias_value":"XXKSVFJ2","created_at":"2026-05-26T01:02:52Z"}],"graph_snapshots":[{"event_id":"sha256:686e9b02b718229aa2d74e407ddf0e0730530cf0bc5d47feda857c4798c94d99","target":"graph","created_at":"2026-05-26T01:02:52Z","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":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"formal_canon":{"evidence_count":0,"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"integrity":{"available":true,"clean":true,"detectors_run":[],"endpoint":"/pith/2605.24202/integrity.json","findings":[],"snapshot_sha256":"c28c3603d3b5d939e8dc4c7e95fa8dfce3d595e45f758748cecf8e644a296938","summary":{"advisory":0,"by_detector":{},"critical":0,"informational":0}},"paper":{"abstract_excerpt":"Multi-agent LLM workflows route inference through specialized roles to lift end-task accuracy, but jointly training those roles with reinforcement learning is unstable in ways that are poorly understood. We study when end-to-end RL training of multi-agent LLM workflows improves over their base models, comparing Shared-Policy training, where all roles update one policy, with Isolated-Policy training, where each role has its own parameters. Our experimental matrix spans Eval-Opt, Voting, and Orch-Workers workflows, math and code tasks, and three model scales (0.6B, 1.7B, 4B). We find that multi-","authors_text":"Huazheng Wang, Kun Wan, Qingyun Wu, Wentian Zhao, Yaolun Zhang, Yifan Zeng, Yiran Wu","cross_cats":["cs.LG"],"headline":"","license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.AI","submitted_at":"2026-05-22T20:43:30Z","title":"When Does Multi-Agent RL Improve LLM Workflows? Workflow, Scale, and Policy-Sharing Tradeoffs"},"references":{"count":0,"internal_anchors":0,"resolved_work":0,"sample":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2605.24202","kind":"arxiv","version":1},"verdict":{"created_at":null,"id":null,"model_set":{},"one_line_summary":"","pipeline_version":null,"pith_extraction_headline":"","strongest_claim":"","weakest_assumption":""}},"verdict_id":null}}],"author_attestations":[],"timestamp_anchors":[],"storage_attestations":[],"citation_signatures":[],"replication_records":[],"corrections":[],"mirror_hints":[],"record_created":{"event_id":"sha256:ca48644667f7099b79de3464f1a356835c1719696d321b3408ae3fc9b8eb9cc5","target":"record","created_at":"2026-05-26T01:02:52Z","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":"a4e598bf07f2dbb30cc185555e60976e151d497eb49acca96cabc8b20ce8432e","cross_cats_sorted":["cs.LG"],"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.AI","submitted_at":"2026-05-22T20:43:30Z","title_canon_sha256":"86c05f403a05cb6001c3dc555d0c585830b0a1be68f8f9e2adcce882935f269c"},"schema_version":"1.0","source":{"id":"2605.24202","kind":"arxiv","version":1}},"canonical_sha256":"bdd52a953a4fe8229e53d4ee8256ffc034fc9912761564d257b34266626bd091","receipt":{"algorithm":"ed25519","builder_version":"pith-number-builder-2026-05-17-v1","canonical_sha256":"bdd52a953a4fe8229e53d4ee8256ffc034fc9912761564d257b34266626bd091","first_computed_at":"2026-05-26T01:02:52.201166Z","key_id":"pith-v1-2026-05","kind":"pith_receipt","last_reissued_at":"2026-05-26T01:02:52.201166Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","receipt_version":"0.3","signature_b64":"Khwb9QPYUuaFuUGPm03TCMma+L8VTOdEXmAP7FBbxUfWf8NJK8Of6JeFBVAKjASx8nuiAtOQStoln5S9rzJeCg==","signature_status":"signed_v1","signed_at":"2026-05-26T01:02:52.201996Z","signed_message":"canonical_sha256_bytes"},"source_id":"2605.24202","source_kind":"arxiv","source_version":1}}},"equivocations":[],"invalid_events":[],"applied_event_ids":["sha256:ca48644667f7099b79de3464f1a356835c1719696d321b3408ae3fc9b8eb9cc5","sha256:686e9b02b718229aa2d74e407ddf0e0730530cf0bc5d47feda857c4798c94d99"],"state_sha256":"af193c06a77f7646d405be3ca94f44d50a286c40d723bd1eb45ee07b4ebb6d70"},"bundle_signature":{"signature_status":"signed_v1","algorithm":"ed25519","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","signature_b64":"uV30BToI5gsMz6IVRZYnksNLSO5fWVAAb2HL5RT20vej0g2RgKl20nrMagokit/R4NLhnMTaJzXJXOOFbzyzAA==","signed_message":"bundle_sha256_bytes","signed_at":"2026-05-31T01:33:01.251080Z","bundle_sha256":"aec158cbb88410122d8e2d318b18e00f184cec5b5f2c8ca11ade350ed224be3f"}}