{"schema_version":"papers/paper-detail-v1","title":"Efficient Post-training of LLMs for Code Generation With Offline Reinforcement Learning","surface":"papers","opportunity_kernel":{"paper_id":"18d0a5f9-96c3-43b7-b07e-534e63f89769","title":"Efficient Post-training of LLMs for Code Generation With Offline Reinforcement Learning","authors":["Mingze Wu","Abhinav Anand","Shweta Verma","Mira Mezini"],"arxiv_id":"2605.28409v1","doi":null,"published_at":"2026-05-27T12:43:14.000Z","score_object":{"overall":{"value":4,"scale":"0-10","confidence":0.85,"confidence_reason":"Aligned to persisted papers.viability_score for public contract consistency.","model_version":"phase0-backfill-v1","pipeline_version":"phase0-kernel-v1","computed_at":"2026-05-28T16:19:34.182Z","fresh_until":"2026-06-27T16:19:34.182Z","is_stale":false,"source_count":1,"missingness":[]},"technical":{"value":1.4,"scale":"0-10","confidence":0.55,"confidence_reason":"Backfilled from paper_extraction_scorecards.reconstruction_score.","model_version":"phase0-backfill-v1","pipeline_version":"phase0-kernel-v1","computed_at":"2026-05-28T16:20:46.592Z","fresh_until":"2026-06-11T16:20:46.592Z","is_stale":false,"source_count":1,"missingness":["reproducibility_results.reproducibility_score","deployability_scores.score"]},"commercial":{"value":4,"scale":"0-10","confidence":0.75,"confidence_reason":"Backfilled from persisted commercial_flags and repo availability.","model_version":"phase0-backfill-v1","pipeline_version":"phase0-kernel-v1","computed_at":"2026-05-28T16:19:34.182Z","fresh_until":"2026-06-27T16:19:34.182Z","is_stale":false,"source_count":1,"missingness":[]},"market":{"value":5.5,"scale":"0-10","confidence":0.8,"confidence_reason":"Backfilled from distribution_readiness_scores.score.","model_version":"phase0-backfill-v1","pipeline_version":"phase0-kernel-v1","computed_at":"2026-05-28T16:51:21.852Z","fresh_until":"2026-06-11T16:51:21.852Z","is_stale":false,"source_count":1,"missingness":[]},"team":{"value":3.6,"scale":"0-10","confidence":0.28,"confidence_reason":"Heuristic fallback from paper author count.","model_version":"phase0-backfill-v1","pipeline_version":"phase0-kernel-v1","computed_at":"2026-05-28T16:19:34.182Z","fresh_until":"2026-06-11T16:19:34.182Z","is_stale":false,"source_count":1,"missingness":["paper_author_affiliations"]},"methodology":{"value":3.88,"scale":"0-10","confidence":0.82,"confidence_reason":"Backfilled from paper_extraction_scorecards.total_score.","model_version":"phase0-backfill-v1","pipeline_version":"phase0-kernel-v1","computed_at":"2026-05-28T16:20:46.592Z","fresh_until":"2026-06-27T16:20:46.592Z","is_stale":false,"source_count":2,"missingness":[]}},"evidence_receipt":{"freshness":"fresh","proof_status":"unverified","repo_status":"missing","references_count":0,"source_count":3,"coverage":0.5,"missingness":["repo_url","references","proof_status"],"unresolved_unknowns":["proof verification has not been recorded yet"],"last_verification_at":"2026-05-28T16:51:21.852Z"},"lineage_hash":"6f92d1f0263485a94ab193a51454ec17ff680532825d2768d02bc9c9f7825e3c"},"distribution":null,"replication_evidence":[],"author_dna":[]}