{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2023:XFQLQ7NPPJ7MV4QT64R6V2CRQS","short_pith_number":"pith:XFQLQ7NP","schema_version":"1.0","canonical_sha256":"b960b87daf7a7ecaf213f723eae85184a55bf2b2dad04808ae2317df4768d80b","source":{"kind":"arxiv","id":"2310.09983","version":1},"attestation_state":"computed","paper":{"title":"Farzi Data: Autoregressive Data Distillation","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.AI","cs.CL","cs.IR"],"primary_cat":"cs.LG","authors_text":"Derek Zhiyuan Cheng, Jianmo Ni, Julian McAuley, Noveen Sachdeva, Wang-Cheng Kang, Zexue He","submitted_at":"2023-10-15T23:23:27Z","abstract_excerpt":"We study data distillation for auto-regressive machine learning tasks, where the input and output have a strict left-to-right causal structure. More specifically, we propose Farzi, which summarizes an event sequence dataset into a small number of synthetic sequences -- Farzi Data -- which are optimized to maintain (if not improve) model performance compared to training on the full dataset. Under the hood, Farzi conducts memory-efficient data distillation by (i) deriving efficient reverse-mode differentiation of the Adam optimizer by leveraging Hessian-Vector Products; and (ii) factorizing the "},"verification_status":{"content_addressed":true,"pith_receipt":true,"author_attested":false,"weak_author_claims":0,"strong_author_claims":0,"externally_anchored":false,"storage_verified":false,"citation_signatures":0,"replication_records":0,"graph_snapshot":true,"references_resolved":false,"formal_links_present":false},"canonical_record":{"source":{"id":"2310.09983","kind":"arxiv","version":1},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.LG","submitted_at":"2023-10-15T23:23:27Z","cross_cats_sorted":["cs.AI","cs.CL","cs.IR"],"title_canon_sha256":"1f96aeb1c9af94eeec31d735cd9aa41c85a3992a0bf1b3b9e200227c39239a44","abstract_canon_sha256":"b8ea47cb5144209052ff0f3ce488e268ac054ccffcfad6e4d55f2bd984a848f4"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T07:01:13.345375Z","signature_b64":"FdRVeHJVZvcx5eYmIEad+0f4DRCsI1HVFNRPMuxad9BSFE5inOokyLgQ5ILlw3za8FpCsTPWRn42SnL4xp5HCQ==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"b960b87daf7a7ecaf213f723eae85184a55bf2b2dad04808ae2317df4768d80b","last_reissued_at":"2026-07-05T07:01:13.344912Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T07:01:13.344912Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Farzi Data: Autoregressive Data Distillation","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.AI","cs.CL","cs.IR"],"primary_cat":"cs.LG","authors_text":"Derek Zhiyuan Cheng, Jianmo Ni, Julian McAuley, Noveen Sachdeva, Wang-Cheng Kang, Zexue He","submitted_at":"2023-10-15T23:23:27Z","abstract_excerpt":"We study data distillation for auto-regressive machine learning tasks, where the input and output have a strict left-to-right causal structure. More specifically, we propose Farzi, which summarizes an event sequence dataset into a small number of synthetic sequences -- Farzi Data -- which are optimized to maintain (if not improve) model performance compared to training on the full dataset. Under the hood, Farzi conducts memory-efficient data distillation by (i) deriving efficient reverse-mode differentiation of the Adam optimizer by leveraging Hessian-Vector Products; and (ii) factorizing the "},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2310.09983","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/2310.09983/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"},"aliases":[{"alias_kind":"arxiv","alias_value":"2310.09983","created_at":"2026-07-05T07:01:13.344967+00:00"},{"alias_kind":"arxiv_version","alias_value":"2310.09983v1","created_at":"2026-07-05T07:01:13.344967+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2310.09983","created_at":"2026-07-05T07:01:13.344967+00:00"},{"alias_kind":"pith_short_12","alias_value":"XFQLQ7NPPJ7M","created_at":"2026-07-05T07:01:13.344967+00:00"},{"alias_kind":"pith_short_16","alias_value":"XFQLQ7NPPJ7MV4QT","created_at":"2026-07-05T07:01:13.344967+00:00"},{"alias_kind":"pith_short_8","alias_value":"XFQLQ7NP","created_at":"2026-07-05T07:01:13.344967+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":1,"internal_anchor_count":1,"sample":[{"citing_arxiv_id":"2502.02854","citing_title":"TD3: Tucker Decomposition Based Dataset Distillation Method for Sequential Recommendation","ref_index":42,"is_internal_anchor":true}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/XFQLQ7NPPJ7MV4QT64R6V2CRQS","json":"https://pith.science/pith/XFQLQ7NPPJ7MV4QT64R6V2CRQS.json","graph_json":"https://pith.science/api/pith-number/XFQLQ7NPPJ7MV4QT64R6V2CRQS/graph.json","events_json":"https://pith.science/api/pith-number/XFQLQ7NPPJ7MV4QT64R6V2CRQS/events.json","paper":"https://pith.science/paper/XFQLQ7NP"},"agent_actions":{"view_html":"https://pith.science/pith/XFQLQ7NPPJ7MV4QT64R6V2CRQS","download_json":"https://pith.science/pith/XFQLQ7NPPJ7MV4QT64R6V2CRQS.json","view_paper":"https://pith.science/paper/XFQLQ7NP","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2310.09983&json=true","fetch_graph":"https://pith.science/api/pith-number/XFQLQ7NPPJ7MV4QT64R6V2CRQS/graph.json","fetch_events":"https://pith.science/api/pith-number/XFQLQ7NPPJ7MV4QT64R6V2CRQS/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/XFQLQ7NPPJ7MV4QT64R6V2CRQS/action/timestamp_anchor","attest_storage":"https://pith.science/pith/XFQLQ7NPPJ7MV4QT64R6V2CRQS/action/storage_attestation","attest_author":"https://pith.science/pith/XFQLQ7NPPJ7MV4QT64R6V2CRQS/action/author_attestation","sign_citation":"https://pith.science/pith/XFQLQ7NPPJ7MV4QT64R6V2CRQS/action/citation_signature","submit_replication":"https://pith.science/pith/XFQLQ7NPPJ7MV4QT64R6V2CRQS/action/replication_record"}},"created_at":"2026-07-05T07:01:13.344967+00:00","updated_at":"2026-07-05T07:01:13.344967+00:00"}