{"bundle_type":"pith_open_graph_bundle","bundle_version":"1.0","pith_number":"pith:2025:K2VOK7PAVBR7KDLDETAL7TQGLQ","short_pith_number":"pith:K2VOK7PA","canonical_record":{"source":{"id":"2511.21056","kind":"arxiv","version":2},"metadata":{"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.LG","submitted_at":"2025-11-26T04:48:33Z","cross_cats_sorted":["cs.CL","math.OC"],"title_canon_sha256":"30d30e720c23f5b900aca00c015983c29457b22f29d68938c82e46dc6033be71","abstract_canon_sha256":"d7b7bdd418c973af09021b839c31bf16da496d78b754751d71a560b7c08274b3"},"schema_version":"1.0"},"canonical_sha256":"56aae57de0a863f50d6324c0bfce065c156dab7bc365ac9420ce04724cfab40a","source":{"kind":"arxiv","id":"2511.21056","version":2},"source_aliases":[{"alias_kind":"arxiv","alias_value":"2511.21056","created_at":"2026-06-24T00:14:21Z"},{"alias_kind":"arxiv_version","alias_value":"2511.21056v2","created_at":"2026-06-24T00:14:21Z"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2511.21056","created_at":"2026-06-24T00:14:21Z"},{"alias_kind":"pith_short_12","alias_value":"K2VOK7PAVBR7","created_at":"2026-06-24T00:14:21Z"},{"alias_kind":"pith_short_16","alias_value":"K2VOK7PAVBR7KDLD","created_at":"2026-06-24T00:14:21Z"},{"alias_kind":"pith_short_8","alias_value":"K2VOK7PA","created_at":"2026-06-24T00:14:21Z"}],"events":[{"event_type":"record_created","subject_pith_number":"pith:2025:K2VOK7PAVBR7KDLDETAL7TQGLQ","target":"record","payload":{"canonical_record":{"source":{"id":"2511.21056","kind":"arxiv","version":2},"metadata":{"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.LG","submitted_at":"2025-11-26T04:48:33Z","cross_cats_sorted":["cs.CL","math.OC"],"title_canon_sha256":"30d30e720c23f5b900aca00c015983c29457b22f29d68938c82e46dc6033be71","abstract_canon_sha256":"d7b7bdd418c973af09021b839c31bf16da496d78b754751d71a560b7c08274b3"},"schema_version":"1.0"},"canonical_sha256":"56aae57de0a863f50d6324c0bfce065c156dab7bc365ac9420ce04724cfab40a","receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-06-24T00:14:21.866208Z","signature_b64":"vVwApqSVXs5L5AZL8uVyC1EeByZtUw1e724n9B3xmCvEKl8Fh3Z7AoI2+E2vWdzy3OLtJgEcVXTxNnIaY8VECQ==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"56aae57de0a863f50d6324c0bfce065c156dab7bc365ac9420ce04724cfab40a","last_reissued_at":"2026-06-24T00:14:21.865729Z","signature_status":"signed_v1","first_computed_at":"2026-06-24T00:14:21.865729Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"source_kind":"arxiv","source_id":"2511.21056","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":"2026-06-24T00:14:21Z","supersedes":[],"prev_event":null,"signature":{"signature_status":"signed_v1","algorithm":"ed25519","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","signature_b64":"v6Prz28r73szwafuGkQlFHWC2l06Sd3wcdBfGGR60v2+TbeeWSThsVj+6b5siAd0RMJS0O8grwB/UBVqeEXcCw==","signed_message":"open_graph_event_sha256_bytes","signed_at":"2026-08-05T11:40:28.597221Z"},"content_sha256":"6149f3cbf0a933778e20d3110c93606bb927b618b0cb7432a479706524bb5dbd","schema_version":"1.0","event_id":"sha256:6149f3cbf0a933778e20d3110c93606bb927b618b0cb7432a479706524bb5dbd"},{"event_type":"graph_snapshot","subject_pith_number":"pith:2025:K2VOK7PAVBR7KDLDETAL7TQGLQ","target":"graph","payload":{"graph_snapshot":{"paper":{"title":"Bilevel Data Curation for LLM Fine-tuning: Offline Selection and Online Self-Refining Generation","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":["cs.CL","math.OC"],"primary_cat":"cs.LG","authors_text":"Gaowen Liu, Quan Xiao, Ramana Rao Kompella, Tianyi Chen, Yutong Xuan","submitted_at":"2025-11-26T04:48:33Z","abstract_excerpt":"Supervised fine-tuning (SFT) datasets are critical to the downstream performance of large language models, yet they often contain low-quality or harmful question-response pairs. To improve SFT data quality, we develop a unified bilevel framework that combines offline data selection with the online self-refining generation. In the offline setting, bilevel data selection (BDS) selects question-response pairs from the offline SFT dataset to maximize the validation performance. We theoretically show that the optimal model given by BDS outperforms direct data mixing approach in useful data coverage"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2511.21056","kind":"arxiv","version":2},"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/2511.21056/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":"2026-06-24T00:14:21Z","supersedes":[],"prev_event":null,"signature":{"signature_status":"signed_v1","algorithm":"ed25519","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","signature_b64":"edIzNExHx3iquqOBvXVE6LoApLkT7t0NOs3bdfS//8U0E43uokLrcMEum5I6EwtuyIttagh1vyFg4+1++Ce+Bw==","signed_message":"open_graph_event_sha256_bytes","signed_at":"2026-08-05T11:40:28.597717Z"},"content_sha256":"5a14c456e90b6e3bda86afeab0594a382d759b33f3735b7fa1ceaac20ca2d3b0","schema_version":"1.0","event_id":"sha256:5a14c456e90b6e3bda86afeab0594a382d759b33f3735b7fa1ceaac20ca2d3b0"}],"timestamp_proofs":[],"mirror_hints":[{"mirror_type":"https","name":"Pith Resolver","base_url":"https://pith.science","bundle_url":"https://pith.science/pith/K2VOK7PAVBR7KDLDETAL7TQGLQ/bundle.json","state_url":"https://pith.science/pith/K2VOK7PAVBR7KDLDETAL7TQGLQ/state.json","well_known_bundle_url":"https://pith.science/.well-known/pith/K2VOK7PAVBR7KDLDETAL7TQGLQ/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":"2026-08-05T11:40:28Z","links":{"resolver":"https://pith.science/pith/K2VOK7PAVBR7KDLDETAL7TQGLQ","bundle":"https://pith.science/pith/K2VOK7PAVBR7KDLDETAL7TQGLQ/bundle.json","state":"https://pith.science/pith/K2VOK7PAVBR7KDLDETAL7TQGLQ/state.json","well_known_bundle":"https://pith.science/.well-known/pith/K2VOK7PAVBR7KDLDETAL7TQGLQ/bundle.json"},"state":{"state_type":"pith_open_graph_state","state_version":"1.0","pith_number":"pith:2025:K2VOK7PAVBR7KDLDETAL7TQGLQ","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":"d7b7bdd418c973af09021b839c31bf16da496d78b754751d71a560b7c08274b3","cross_cats_sorted":["cs.CL","math.OC"],"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.LG","submitted_at":"2025-11-26T04:48:33Z","title_canon_sha256":"30d30e720c23f5b900aca00c015983c29457b22f29d68938c82e46dc6033be71"},"schema_version":"1.0","source":{"id":"2511.21056","kind":"arxiv","version":2}},"source_aliases":[{"alias_kind":"arxiv","alias_value":"2511.21056","created_at":"2026-06-24T00:14:21Z"},{"alias_kind":"arxiv_version","alias_value":"2511.21056v2","created_at":"2026-06-24T00:14:21Z"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2511.21056","created_at":"2026-06-24T00:14:21Z"},{"alias_kind":"pith_short_12","alias_value":"K2VOK7PAVBR7","created_at":"2026-06-24T00:14:21Z"},{"alias_kind":"pith_short_16","alias_value":"K2VOK7PAVBR7KDLD","created_at":"2026-06-24T00:14:21Z"},{"alias_kind":"pith_short_8","alias_value":"K2VOK7PA","created_at":"2026-06-24T00:14:21Z"}],"graph_snapshots":[{"event_id":"sha256:5a14c456e90b6e3bda86afeab0594a382d759b33f3735b7fa1ceaac20ca2d3b0","target":"graph","created_at":"2026-06-24T00:14:21Z","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/2511.21056/integrity.json","findings":[],"snapshot_sha256":"c28c3603d3b5d939e8dc4c7e95fa8dfce3d595e45f758748cecf8e644a296938","summary":{"advisory":0,"by_detector":{},"critical":0,"informational":0}},"paper":{"abstract_excerpt":"Supervised fine-tuning (SFT) datasets are critical to the downstream performance of large language models, yet they often contain low-quality or harmful question-response pairs. To improve SFT data quality, we develop a unified bilevel framework that combines offline data selection with the online self-refining generation. In the offline setting, bilevel data selection (BDS) selects question-response pairs from the offline SFT dataset to maximize the validation performance. We theoretically show that the optimal model given by BDS outperforms direct data mixing approach in useful data coverage","authors_text":"Gaowen Liu, Quan Xiao, Ramana Rao Kompella, Tianyi Chen, Yutong Xuan","cross_cats":["cs.CL","math.OC"],"headline":"","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.LG","submitted_at":"2025-11-26T04:48:33Z","title":"Bilevel Data Curation for LLM Fine-tuning: Offline Selection and Online Self-Refining Generation"},"references":{"count":0,"internal_anchors":0,"resolved_work":0,"sample":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2511.21056","kind":"arxiv","version":2},"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":"sha256:6149f3cbf0a933778e20d3110c93606bb927b618b0cb7432a479706524bb5dbd","target":"record","created_at":"2026-06-24T00:14:21Z","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":"d7b7bdd418c973af09021b839c31bf16da496d78b754751d71a560b7c08274b3","cross_cats_sorted":["cs.CL","math.OC"],"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.LG","submitted_at":"2025-11-26T04:48:33Z","title_canon_sha256":"30d30e720c23f5b900aca00c015983c29457b22f29d68938c82e46dc6033be71"},"schema_version":"1.0","source":{"id":"2511.21056","kind":"arxiv","version":2}},"canonical_sha256":"56aae57de0a863f50d6324c0bfce065c156dab7bc365ac9420ce04724cfab40a","receipt":{"algorithm":"ed25519","builder_version":"pith-number-builder-2026-05-17-v1","canonical_sha256":"56aae57de0a863f50d6324c0bfce065c156dab7bc365ac9420ce04724cfab40a","first_computed_at":"2026-06-24T00:14:21.865729Z","key_id":"pith-v1-2026-05","kind":"pith_receipt","last_reissued_at":"2026-06-24T00:14:21.865729Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","receipt_version":"0.3","signature_b64":"vVwApqSVXs5L5AZL8uVyC1EeByZtUw1e724n9B3xmCvEKl8Fh3Z7AoI2+E2vWdzy3OLtJgEcVXTxNnIaY8VECQ==","signature_status":"signed_v1","signed_at":"2026-06-24T00:14:21.866208Z","signed_message":"canonical_sha256_bytes"},"source_id":"2511.21056","source_kind":"arxiv","source_version":2}}},"equivocations":[],"invalid_events":[],"applied_event_ids":["sha256:6149f3cbf0a933778e20d3110c93606bb927b618b0cb7432a479706524bb5dbd","sha256:5a14c456e90b6e3bda86afeab0594a382d759b33f3735b7fa1ceaac20ca2d3b0"],"state_sha256":"2a68e5427c5793a0628ce869ac5c1f3d10c494e7c6e6c6c13f83aeadbb2704ba"},"bundle_signature":{"signature_status":"signed_v1","algorithm":"ed25519","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","signature_b64":"ynFZHtdMZB7H//ORdoon5WW+3LmRPvEF6eE/pRKEwDcXolfvtnwIF2onQHqCQb+GuXhqZ8cE72V2K7cI0pTZAQ==","signed_message":"bundle_sha256_bytes","signed_at":"2026-08-05T11:40:28.601718Z","bundle_sha256":"94265ae79997f6f1bdebc815af39abffe68d07e84b9595ef15f03352eae801ed"}}