{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2024:KAN4ECZNLNTIQSO3RPFGFKAMNB","short_pith_number":"pith:KAN4ECZN","schema_version":"1.0","canonical_sha256":"501bc20b2d5b668849db8bca62a80c685d7b2639cc7541ce3ae8b9398ca9511f","source":{"kind":"arxiv","id":"2406.06887","version":4},"attestation_state":"computed","paper":{"title":"$\\textbf{PLUM}$: Improving Code LMs with Execution-Guided On-Policy Preference Learning Driven By Synthetic Test Cases","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.AI","cs.LG","cs.PL","cs.SE"],"primary_cat":"cs.CL","authors_text":"Dylan Zhang, Hao Peng, Shizhe Diao, Xueyan Zou","submitted_at":"2024-06-11T02:07:18Z","abstract_excerpt":"Preference learning provides a promising solution to address the limitations of supervised fine-tuning (SFT) for code language models, where the model is not explicitly trained to differentiate between correct and incorrect code. Recent findings demonstrate that on-policy data is the key to successful preference learning, where the preference data is collected using the same policy LM being trained. Inspired by this, we propose PLUM, an on-policy $\\textbf{P}$reference $\\textbf{L}$earning framework A$\\textbf{u}$gmented with test cases for code L$\\textbf{M}$ s. The framework operates in three ke"},"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":"2406.06887","kind":"arxiv","version":4},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.CL","submitted_at":"2024-06-11T02:07:18Z","cross_cats_sorted":["cs.AI","cs.LG","cs.PL","cs.SE"],"title_canon_sha256":"755ccb21dbda7ebf88624ee3c37c0ec01da2644d7258e08a03ea97287639eea9","abstract_canon_sha256":"edea031b31aa00746d5e4a14f9957031c4f769b38df3f6311758ed59d98e88b8"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T09:19:29.041783Z","signature_b64":"yCe8x4OUsa6Z/ZFwhDcBbrKo3lUNssmZ3pKXstO1IrslJNBUOZzvxbArj1Idg6Ar40ED2GehKpEQAJqpQ4NMBw==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"501bc20b2d5b668849db8bca62a80c685d7b2639cc7541ce3ae8b9398ca9511f","last_reissued_at":"2026-07-05T09:19:29.041269Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T09:19:29.041269Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"$\\textbf{PLUM}$: Improving Code LMs with Execution-Guided On-Policy Preference Learning Driven By Synthetic Test Cases","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.AI","cs.LG","cs.PL","cs.SE"],"primary_cat":"cs.CL","authors_text":"Dylan Zhang, Hao Peng, Shizhe Diao, Xueyan Zou","submitted_at":"2024-06-11T02:07:18Z","abstract_excerpt":"Preference learning provides a promising solution to address the limitations of supervised fine-tuning (SFT) for code language models, where the model is not explicitly trained to differentiate between correct and incorrect code. Recent findings demonstrate that on-policy data is the key to successful preference learning, where the preference data is collected using the same policy LM being trained. Inspired by this, we propose PLUM, an on-policy $\\textbf{P}$reference $\\textbf{L}$earning framework A$\\textbf{u}$gmented with test cases for code L$\\textbf{M}$ s. The framework operates in three ke"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2406.06887","kind":"arxiv","version":4},"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/2406.06887/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":"2406.06887","created_at":"2026-07-05T09:19:29.041329+00:00"},{"alias_kind":"arxiv_version","alias_value":"2406.06887v4","created_at":"2026-07-05T09:19:29.041329+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2406.06887","created_at":"2026-07-05T09:19:29.041329+00:00"},{"alias_kind":"pith_short_12","alias_value":"KAN4ECZNLNTI","created_at":"2026-07-05T09:19:29.041329+00:00"},{"alias_kind":"pith_short_16","alias_value":"KAN4ECZNLNTIQSO3","created_at":"2026-07-05T09:19:29.041329+00:00"},{"alias_kind":"pith_short_8","alias_value":"KAN4ECZN","created_at":"2026-07-05T09:19:29.041329+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":1,"internal_anchor_count":0,"sample":[{"citing_arxiv_id":"2606.06826","citing_title":"SkelDPO: A Skeleton-Guided Direct Preference Optimization Framework for Efficient Code Generation","ref_index":53,"is_internal_anchor":false}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/KAN4ECZNLNTIQSO3RPFGFKAMNB","json":"https://pith.science/pith/KAN4ECZNLNTIQSO3RPFGFKAMNB.json","graph_json":"https://pith.science/api/pith-number/KAN4ECZNLNTIQSO3RPFGFKAMNB/graph.json","events_json":"https://pith.science/api/pith-number/KAN4ECZNLNTIQSO3RPFGFKAMNB/events.json","paper":"https://pith.science/paper/KAN4ECZN"},"agent_actions":{"view_html":"https://pith.science/pith/KAN4ECZNLNTIQSO3RPFGFKAMNB","download_json":"https://pith.science/pith/KAN4ECZNLNTIQSO3RPFGFKAMNB.json","view_paper":"https://pith.science/paper/KAN4ECZN","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2406.06887&json=true","fetch_graph":"https://pith.science/api/pith-number/KAN4ECZNLNTIQSO3RPFGFKAMNB/graph.json","fetch_events":"https://pith.science/api/pith-number/KAN4ECZNLNTIQSO3RPFGFKAMNB/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/KAN4ECZNLNTIQSO3RPFGFKAMNB/action/timestamp_anchor","attest_storage":"https://pith.science/pith/KAN4ECZNLNTIQSO3RPFGFKAMNB/action/storage_attestation","attest_author":"https://pith.science/pith/KAN4ECZNLNTIQSO3RPFGFKAMNB/action/author_attestation","sign_citation":"https://pith.science/pith/KAN4ECZNLNTIQSO3RPFGFKAMNB/action/citation_signature","submit_replication":"https://pith.science/pith/KAN4ECZNLNTIQSO3RPFGFKAMNB/action/replication_record"}},"created_at":"2026-07-05T09:19:29.041329+00:00","updated_at":"2026-07-05T09:19:29.041329+00:00"}