{"bundle_type":"pith_open_graph_bundle","bundle_version":"1.0","pith_number":"pith:2026:G3Z2L34L5RXPLEEOLSFCOQNYHI","short_pith_number":"pith:G3Z2L34L","canonical_record":{"source":{"id":"2605.09855","kind":"arxiv","version":2},"metadata":{"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.LG","submitted_at":"2026-05-11T01:17:58Z","cross_cats_sorted":[],"title_canon_sha256":"bf115b40a20a80e26a8ac69f2869de93b40e84bd0f6c4760bca27275228a5ea4","abstract_canon_sha256":"ed6f18e0ba658cf859a39bb5ba3cf603ec501301d3f21911b73116cc0ec4bfa6"},"schema_version":"1.0"},"canonical_sha256":"36f3a5ef8bec6ef5908e5c8a2741b83a32b31f48b913e0fdb6d220bde85ead82","source":{"kind":"arxiv","id":"2605.09855","version":2},"source_aliases":[{"alias_kind":"arxiv","alias_value":"2605.09855","created_at":"2026-05-20T00:04:35Z"},{"alias_kind":"arxiv_version","alias_value":"2605.09855v2","created_at":"2026-05-20T00:04:35Z"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2605.09855","created_at":"2026-05-20T00:04:35Z"},{"alias_kind":"pith_short_12","alias_value":"G3Z2L34L5RXP","created_at":"2026-05-20T00:04:35Z"},{"alias_kind":"pith_short_16","alias_value":"G3Z2L34L5RXPLEEO","created_at":"2026-05-20T00:04:35Z"},{"alias_kind":"pith_short_8","alias_value":"G3Z2L34L","created_at":"2026-05-20T00:04:35Z"}],"events":[{"event_type":"record_created","subject_pith_number":"pith:2026:G3Z2L34L5RXPLEEOLSFCOQNYHI","target":"record","payload":{"canonical_record":{"source":{"id":"2605.09855","kind":"arxiv","version":2},"metadata":{"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.LG","submitted_at":"2026-05-11T01:17:58Z","cross_cats_sorted":[],"title_canon_sha256":"bf115b40a20a80e26a8ac69f2869de93b40e84bd0f6c4760bca27275228a5ea4","abstract_canon_sha256":"ed6f18e0ba658cf859a39bb5ba3cf603ec501301d3f21911b73116cc0ec4bfa6"},"schema_version":"1.0"},"canonical_sha256":"36f3a5ef8bec6ef5908e5c8a2741b83a32b31f48b913e0fdb6d220bde85ead82","receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-05-20T00:04:35.326989Z","signature_b64":"P1R+LJ+6M25yej0kr4+L4nL0jzutv0YXelaH+7YbUqD5/fDlc0+tqz3QzJcGGrIjvrMk3MRUWgwED9h21INCAw==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"36f3a5ef8bec6ef5908e5c8a2741b83a32b31f48b913e0fdb6d220bde85ead82","last_reissued_at":"2026-05-20T00:04:35.325923Z","signature_status":"signed_v1","first_computed_at":"2026-05-20T00:04:35.325923Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"source_kind":"arxiv","source_id":"2605.09855","source_version":2,"attestation_state":"computed"},"signer":{"signer_id":"pith.science","signer_type":"pith_registry","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"created_at":"2026-05-20T00:04:35Z","supersedes":[],"prev_event":null,"signature":{"signature_status":"signed_v1","algorithm":"ed25519","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","signature_b64":"OG9n58Fc7ufc9nQQkXIJh275lof3p6IcmsEUEmJTuqBB7PIh5pEpW8s5iHFFTG/7CW6BhgmDXGO/JI9gcuYdDw==","signed_message":"open_graph_event_sha256_bytes","signed_at":"2026-08-04T17:33:39.041604Z"},"content_sha256":"62c5cb15624f3573cb231d58cd92697aa39d544e9f3b638d0e93ebaa41827359","schema_version":"1.0","event_id":"sha256:62c5cb15624f3573cb231d58cd92697aa39d544e9f3b638d0e93ebaa41827359"},{"event_type":"graph_snapshot","subject_pith_number":"pith:2026:G3Z2L34L5RXPLEEOLSFCOQNYHI","target":"graph","payload":{"graph_snapshot":{"paper":{"title":"Concordia: Self-Improving Synthetic Tables for Federated LLMs","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"A tri-level loop lets federated clients refine their own synthetic tables using local scorers and shared ensembles without sharing data.","cross_cats":[],"primary_cat":"cs.LG","authors_text":"Alejandro Lopez-Lira, Duanyu Feng, Guojun Xiong, Jimin Huang, Mingquan Lin, Nuo Chen, Prayag Tiwari, Sophia Ananiadou, Xiaoyu Wang, Xueqing Peng, Zhiqiang Zhang","submitted_at":"2026-05-11T01:17:58Z","abstract_excerpt":"Federated learning (FL) enables training large language models (LLMs) without sharing raw data, but adapting LLMs under strict data isolation and non-IID client distributions remains challenging in practice. Synthetic data offers a natural privacy-preserving surrogate for local training, yet existing federated pipelines typically treat synthetic generation as static or loosely coupled with downstream optimization, leading to rapidly diminishing utility under heterogeneous clients. We study federated adaptation of LLMs on tabular tasks where raw records and validation data cannot be shared, and"},"claims":{"count":4,"items":[{"kind":"strongest_claim","text":"Experiments on privacy-sensitive tabular benchmarks from finance and healthcare demonstrate that Concordia consistently improves federated performance, cross-client stability, and robustness to distribution shift compared to static and decoupled synthetic-data baselines.","source":"verdict.strongest_claim","status":"machine_extracted","claim_id":"C1","attestation":"unclaimed"},{"kind":"weakest_assumption","text":"That lightweight utility scorers learned locally from private validation feedback can reliably reweight synthetic samples and that an ensemble of heterogeneous scorers can effectively guide generator refinement via GRPO without exposing validation data or generator parameters.","source":"verdict.weakest_assumption","status":"machine_extracted","claim_id":"C2","attestation":"unclaimed"},{"kind":"one_line_summary","text":"Concordia aligns synthetic table generation with federated validation utility via client-level LoRA training, utility scorers, and outer GRPO refinement to boost performance over static synthetic baselines.","source":"verdict.one_line_summary","status":"machine_extracted","claim_id":"C3","attestation":"unclaimed"},{"kind":"headline","text":"A tri-level loop lets federated clients refine their own synthetic tables using local scorers and shared ensembles without sharing data.","source":"verdict.pith_extraction.headline","status":"machine_extracted","claim_id":"C4","attestation":"unclaimed"}],"snapshot_sha256":"5cfd9d115ab11808b8af7669541520618722adcd312cd65ce44f134d860bb187"},"source":{"id":"2605.09855","kind":"arxiv","version":2},"verdict":{"id":"de59f645-1335-4ac5-80dc-40ff5b4359be","model_set":{"reader":"grok-4.3"},"created_at":"2026-05-12T04:21:38.155361Z","strongest_claim":"Experiments on privacy-sensitive tabular benchmarks from finance and healthcare demonstrate that Concordia consistently improves federated performance, cross-client stability, and robustness to distribution shift compared to static and decoupled synthetic-data baselines.","one_line_summary":"Concordia aligns synthetic table generation with federated validation utility via client-level LoRA training, utility scorers, and outer GRPO refinement to boost performance over static synthetic baselines.","pipeline_version":"pith-pipeline@v0.9.0","weakest_assumption":"That lightweight utility scorers learned locally from private validation feedback can reliably reweight synthetic samples and that an ensemble of heterogeneous scorers can effectively guide generator refinement via GRPO without exposing validation data or generator parameters.","pith_extraction_headline":"A tri-level loop lets federated clients refine their own synthetic tables using local scorers and shared ensembles without sharing data."},"integrity":{"clean":true,"summary":{"advisory":0,"critical":0,"by_detector":{},"informational":0},"endpoint":"/pith/2605.09855/integrity.json","findings":[],"available":true,"detectors_run":[{"name":"ai_meta_artifact","ran_at":"2026-05-19T16:34:50.343534Z","status":"completed","version":"1.0.0","findings_count":0},{"name":"doi_title_agreement","ran_at":"2026-05-19T12:31:17.480393Z","status":"completed","version":"1.0.0","findings_count":0},{"name":"doi_compliance","ran_at":"2026-05-19T09:52:12.106751Z","status":"completed","version":"1.0.0","findings_count":0}],"snapshot_sha256":"06185aa3bbaff3cead62ca2383b7e9a7dd8bc237f8a303bb9cc32df5940a00d9"},"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":"de59f645-1335-4ac5-80dc-40ff5b4359be"},"signer":{"signer_id":"pith.science","signer_type":"pith_registry","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"created_at":"2026-05-20T00:04:35Z","supersedes":[],"prev_event":null,"signature":{"signature_status":"signed_v1","algorithm":"ed25519","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","signature_b64":"ypabGBC5Tt9rptZEXCEgzbMwjl8u7IkdwivkaJkz6DtkrwUSmDxK2EyveUsIxZV6J7GlmlbYomwm3FM3eASQDg==","signed_message":"open_graph_event_sha256_bytes","signed_at":"2026-08-04T17:33:39.042373Z"},"content_sha256":"9a44844c06a7311a311af717280bee2883febeecfea4907e747fac39affca7d0","schema_version":"1.0","event_id":"sha256:9a44844c06a7311a311af717280bee2883febeecfea4907e747fac39affca7d0"}],"timestamp_proofs":[],"mirror_hints":[{"mirror_type":"https","name":"Pith Resolver","base_url":"https://pith.science","bundle_url":"https://pith.science/pith/G3Z2L34L5RXPLEEOLSFCOQNYHI/bundle.json","state_url":"https://pith.science/pith/G3Z2L34L5RXPLEEOLSFCOQNYHI/state.json","well_known_bundle_url":"https://pith.science/.well-known/pith/G3Z2L34L5RXPLEEOLSFCOQNYHI/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":"2026-08-04T17:33:39Z","links":{"resolver":"https://pith.science/pith/G3Z2L34L5RXPLEEOLSFCOQNYHI","bundle":"https://pith.science/pith/G3Z2L34L5RXPLEEOLSFCOQNYHI/bundle.json","state":"https://pith.science/pith/G3Z2L34L5RXPLEEOLSFCOQNYHI/state.json","well_known_bundle":"https://pith.science/.well-known/pith/G3Z2L34L5RXPLEEOLSFCOQNYHI/bundle.json"},"state":{"state_type":"pith_open_graph_state","state_version":"1.0","pith_number":"pith:2026:G3Z2L34L5RXPLEEOLSFCOQNYHI","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":"ed6f18e0ba658cf859a39bb5ba3cf603ec501301d3f21911b73116cc0ec4bfa6","cross_cats_sorted":[],"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.LG","submitted_at":"2026-05-11T01:17:58Z","title_canon_sha256":"bf115b40a20a80e26a8ac69f2869de93b40e84bd0f6c4760bca27275228a5ea4"},"schema_version":"1.0","source":{"id":"2605.09855","kind":"arxiv","version":2}},"source_aliases":[{"alias_kind":"arxiv","alias_value":"2605.09855","created_at":"2026-05-20T00:04:35Z"},{"alias_kind":"arxiv_version","alias_value":"2605.09855v2","created_at":"2026-05-20T00:04:35Z"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2605.09855","created_at":"2026-05-20T00:04:35Z"},{"alias_kind":"pith_short_12","alias_value":"G3Z2L34L5RXP","created_at":"2026-05-20T00:04:35Z"},{"alias_kind":"pith_short_16","alias_value":"G3Z2L34L5RXPLEEO","created_at":"2026-05-20T00:04:35Z"},{"alias_kind":"pith_short_8","alias_value":"G3Z2L34L","created_at":"2026-05-20T00:04:35Z"}],"graph_snapshots":[{"event_id":"sha256:9a44844c06a7311a311af717280bee2883febeecfea4907e747fac39affca7d0","target":"graph","created_at":"2026-05-20T00:04:35Z","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":"Experiments on privacy-sensitive tabular benchmarks from finance and healthcare demonstrate that Concordia consistently improves federated performance, cross-client stability, and robustness to distribution shift compared to static and decoupled synthetic-data baselines."},{"attestation":"unclaimed","claim_id":"C2","kind":"weakest_assumption","source":"verdict.weakest_assumption","status":"machine_extracted","text":"That lightweight utility scorers learned locally from private validation feedback can reliably reweight synthetic samples and that an ensemble of heterogeneous scorers can effectively guide generator refinement via GRPO without exposing validation data or generator parameters."},{"attestation":"unclaimed","claim_id":"C3","kind":"one_line_summary","source":"verdict.one_line_summary","status":"machine_extracted","text":"Concordia aligns synthetic table generation with federated validation utility via client-level LoRA training, utility scorers, and outer GRPO refinement to boost performance over static synthetic baselines."},{"attestation":"unclaimed","claim_id":"C4","kind":"headline","source":"verdict.pith_extraction.headline","status":"machine_extracted","text":"A tri-level loop lets federated clients refine their own synthetic tables using local scorers and shared ensembles without sharing data."}],"snapshot_sha256":"5cfd9d115ab11808b8af7669541520618722adcd312cd65ce44f134d860bb187"},"formal_canon":{"evidence_count":0,"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"integrity":{"available":true,"clean":true,"detectors_run":[{"findings_count":0,"name":"ai_meta_artifact","ran_at":"2026-05-19T16:34:50.343534Z","status":"completed","version":"1.0.0"},{"findings_count":0,"name":"doi_title_agreement","ran_at":"2026-05-19T12:31:17.480393Z","status":"completed","version":"1.0.0"},{"findings_count":0,"name":"doi_compliance","ran_at":"2026-05-19T09:52:12.106751Z","status":"completed","version":"1.0.0"}],"endpoint":"/pith/2605.09855/integrity.json","findings":[],"snapshot_sha256":"06185aa3bbaff3cead62ca2383b7e9a7dd8bc237f8a303bb9cc32df5940a00d9","summary":{"advisory":0,"by_detector":{},"critical":0,"informational":0}},"paper":{"abstract_excerpt":"Federated learning (FL) enables training large language models (LLMs) without sharing raw data, but adapting LLMs under strict data isolation and non-IID client distributions remains challenging in practice. Synthetic data offers a natural privacy-preserving surrogate for local training, yet existing federated pipelines typically treat synthetic generation as static or loosely coupled with downstream optimization, leading to rapidly diminishing utility under heterogeneous clients. We study federated adaptation of LLMs on tabular tasks where raw records and validation data cannot be shared, and","authors_text":"Alejandro Lopez-Lira, Duanyu Feng, Guojun Xiong, Jimin Huang, Mingquan Lin, Nuo Chen, Prayag Tiwari, Sophia Ananiadou, Xiaoyu Wang, Xueqing Peng, Zhiqiang Zhang","cross_cats":[],"headline":"A tri-level loop lets federated clients refine their own synthetic tables using local scorers and shared ensembles without sharing data.","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.LG","submitted_at":"2026-05-11T01:17:58Z","title":"Concordia: Self-Improving Synthetic Tables for Federated LLMs"},"references":{"count":0,"internal_anchors":0,"resolved_work":0,"sample":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2605.09855","kind":"arxiv","version":2},"verdict":{"created_at":"2026-05-12T04:21:38.155361Z","id":"de59f645-1335-4ac5-80dc-40ff5b4359be","model_set":{"reader":"grok-4.3"},"one_line_summary":"Concordia aligns synthetic table generation with federated validation utility via client-level LoRA training, utility scorers, and outer GRPO refinement to boost performance over static synthetic baselines.","pipeline_version":"pith-pipeline@v0.9.0","pith_extraction_headline":"A tri-level loop lets federated clients refine their own synthetic tables using local scorers and shared ensembles without sharing data.","strongest_claim":"Experiments on privacy-sensitive tabular benchmarks from finance and healthcare demonstrate that Concordia consistently improves federated performance, cross-client stability, and robustness to distribution shift compared to static and decoupled synthetic-data baselines.","weakest_assumption":"That lightweight utility scorers learned locally from private validation feedback can reliably reweight synthetic samples and that an ensemble of heterogeneous scorers can effectively guide generator refinement via GRPO without exposing validation data or generator parameters."}},"verdict_id":"de59f645-1335-4ac5-80dc-40ff5b4359be"}}],"author_attestations":[],"timestamp_anchors":[],"storage_attestations":[],"citation_signatures":[],"replication_records":[],"corrections":[],"mirror_hints":[],"record_created":{"event_id":"sha256:62c5cb15624f3573cb231d58cd92697aa39d544e9f3b638d0e93ebaa41827359","target":"record","created_at":"2026-05-20T00:04:35Z","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":"ed6f18e0ba658cf859a39bb5ba3cf603ec501301d3f21911b73116cc0ec4bfa6","cross_cats_sorted":[],"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.LG","submitted_at":"2026-05-11T01:17:58Z","title_canon_sha256":"bf115b40a20a80e26a8ac69f2869de93b40e84bd0f6c4760bca27275228a5ea4"},"schema_version":"1.0","source":{"id":"2605.09855","kind":"arxiv","version":2}},"canonical_sha256":"36f3a5ef8bec6ef5908e5c8a2741b83a32b31f48b913e0fdb6d220bde85ead82","receipt":{"algorithm":"ed25519","builder_version":"pith-number-builder-2026-05-17-v1","canonical_sha256":"36f3a5ef8bec6ef5908e5c8a2741b83a32b31f48b913e0fdb6d220bde85ead82","first_computed_at":"2026-05-20T00:04:35.325923Z","key_id":"pith-v1-2026-05","kind":"pith_receipt","last_reissued_at":"2026-05-20T00:04:35.325923Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","receipt_version":"0.3","signature_b64":"P1R+LJ+6M25yej0kr4+L4nL0jzutv0YXelaH+7YbUqD5/fDlc0+tqz3QzJcGGrIjvrMk3MRUWgwED9h21INCAw==","signature_status":"signed_v1","signed_at":"2026-05-20T00:04:35.326989Z","signed_message":"canonical_sha256_bytes"},"source_id":"2605.09855","source_kind":"arxiv","source_version":2}}},"equivocations":[],"invalid_events":[],"applied_event_ids":["sha256:62c5cb15624f3573cb231d58cd92697aa39d544e9f3b638d0e93ebaa41827359","sha256:9a44844c06a7311a311af717280bee2883febeecfea4907e747fac39affca7d0"],"state_sha256":"acf5fc168716cded930bcb6d3227cbc1602dcd1c3f063efb1e4f93fc5c6c7096"},"bundle_signature":{"signature_status":"signed_v1","algorithm":"ed25519","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","signature_b64":"H2hgG+mAeKcnONke7GyhCqtS0wSJn3XftCcmdtazJS6xoHRBtq9qrJBHslHRmAs1mB6vb2iRb2bL5+PoBkPUAQ==","signed_message":"bundle_sha256_bytes","signed_at":"2026-08-04T17:33:39.047020Z","bundle_sha256":"1ec83eea4cc12283868246dae5f45c0cf2da3fa71f951336ec21e6451217fcf2"}}