{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2023:YVFSOUNMIXAAQ5ARO4DNGUVMX5","short_pith_number":"pith:YVFSOUNM","schema_version":"1.0","canonical_sha256":"c54b2751ac45c00874117706d352acbf5e210e8f59e1647727777dc6bae5cc0c","source":{"kind":"arxiv","id":"2305.18341","version":2},"attestation_state":"computed","paper":{"title":"Coarse-Tuning Models of Code with Reinforcement Learning Feedback","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":["cs.AI","cs.LG"],"primary_cat":"cs.PL","authors_text":"(2) UT Austin, (3) University of Wisconsin), Abhinav Jain (1), Chima Adiole (1), Chris Jermaine (1) ((1) Rice University, Swarat Chaudhuri (2), Thomas Reps (3)","submitted_at":"2023-05-25T22:09:08Z","abstract_excerpt":"Large Language Models (LLMs) pre-trained on code have recently emerged as the dominant approach to program synthesis. However, these models are trained using next-token prediction, which ignores the syntax and semantics of code. We propose RLCF, that further trains a pre-trained LLM via reinforcement learning, using feedback from a grounding function that scores the quality of the code. The grounding function uses (i) compiler-derived feedback on whether the code it generates passes a set of correctness checks; and (ii) feedback from a different LLM that compares the generated code to a refere"},"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":"2305.18341","kind":"arxiv","version":2},"metadata":{"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.PL","submitted_at":"2023-05-25T22:09:08Z","cross_cats_sorted":["cs.AI","cs.LG"],"title_canon_sha256":"d600772fdbe43fa439e7e5aac6ff5055640fe2f1c75758395e0fda24ff90b439","abstract_canon_sha256":"838d2109150f60fdb640b456794af01a62fd86a774125e416c47d1298206e20e"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T07:27:25.314749Z","signature_b64":"bp8T4kIv20QXFfI0qYiSMxQQo7RLHliJuGDeXErsnn3G9u84AL+E5Vb2rLls3TI4NnEc3SB7HDycpt0p136vAQ==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"c54b2751ac45c00874117706d352acbf5e210e8f59e1647727777dc6bae5cc0c","last_reissued_at":"2026-07-05T07:27:25.314109Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T07:27:25.314109Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Coarse-Tuning Models of Code with Reinforcement Learning Feedback","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":["cs.AI","cs.LG"],"primary_cat":"cs.PL","authors_text":"(2) UT Austin, (3) University of Wisconsin), Abhinav Jain (1), Chima Adiole (1), Chris Jermaine (1) ((1) Rice University, Swarat Chaudhuri (2), Thomas Reps (3)","submitted_at":"2023-05-25T22:09:08Z","abstract_excerpt":"Large Language Models (LLMs) pre-trained on code have recently emerged as the dominant approach to program synthesis. However, these models are trained using next-token prediction, which ignores the syntax and semantics of code. We propose RLCF, that further trains a pre-trained LLM via reinforcement learning, using feedback from a grounding function that scores the quality of the code. The grounding function uses (i) compiler-derived feedback on whether the code it generates passes a set of correctness checks; and (ii) feedback from a different LLM that compares the generated code to a refere"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2305.18341","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/2305.18341/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":"2305.18341","created_at":"2026-07-05T07:27:25.314187+00:00"},{"alias_kind":"arxiv_version","alias_value":"2305.18341v2","created_at":"2026-07-05T07:27:25.314187+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2305.18341","created_at":"2026-07-05T07:27:25.314187+00:00"},{"alias_kind":"pith_short_12","alias_value":"YVFSOUNMIXAA","created_at":"2026-07-05T07:27:25.314187+00:00"},{"alias_kind":"pith_short_16","alias_value":"YVFSOUNMIXAAQ5AR","created_at":"2026-07-05T07:27:25.314187+00:00"},{"alias_kind":"pith_short_8","alias_value":"YVFSOUNM","created_at":"2026-07-05T07:27:25.314187+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":1,"internal_anchor_count":1,"sample":[{"citing_arxiv_id":"2501.09025","citing_title":"Cyber Shadows: Neutralizing Security Threats with AI and Targeted Policy Measures","ref_index":27,"is_internal_anchor":true}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/YVFSOUNMIXAAQ5ARO4DNGUVMX5","json":"https://pith.science/pith/YVFSOUNMIXAAQ5ARO4DNGUVMX5.json","graph_json":"https://pith.science/api/pith-number/YVFSOUNMIXAAQ5ARO4DNGUVMX5/graph.json","events_json":"https://pith.science/api/pith-number/YVFSOUNMIXAAQ5ARO4DNGUVMX5/events.json","paper":"https://pith.science/paper/YVFSOUNM"},"agent_actions":{"view_html":"https://pith.science/pith/YVFSOUNMIXAAQ5ARO4DNGUVMX5","download_json":"https://pith.science/pith/YVFSOUNMIXAAQ5ARO4DNGUVMX5.json","view_paper":"https://pith.science/paper/YVFSOUNM","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2305.18341&json=true","fetch_graph":"https://pith.science/api/pith-number/YVFSOUNMIXAAQ5ARO4DNGUVMX5/graph.json","fetch_events":"https://pith.science/api/pith-number/YVFSOUNMIXAAQ5ARO4DNGUVMX5/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/YVFSOUNMIXAAQ5ARO4DNGUVMX5/action/timestamp_anchor","attest_storage":"https://pith.science/pith/YVFSOUNMIXAAQ5ARO4DNGUVMX5/action/storage_attestation","attest_author":"https://pith.science/pith/YVFSOUNMIXAAQ5ARO4DNGUVMX5/action/author_attestation","sign_citation":"https://pith.science/pith/YVFSOUNMIXAAQ5ARO4DNGUVMX5/action/citation_signature","submit_replication":"https://pith.science/pith/YVFSOUNMIXAAQ5ARO4DNGUVMX5/action/replication_record"}},"created_at":"2026-07-05T07:27:25.314187+00:00","updated_at":"2026-07-05T07:27:25.314187+00:00"}