{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2024:ZMMNP36SSQ4XYFGSYXLHFYE2LG","short_pith_number":"pith:ZMMNP36S","schema_version":"1.0","canonical_sha256":"cb18d7efd294397c14d2c5d672e09a59a5cea078a2802c450e4074ee254c93cf","source":{"kind":"arxiv","id":"2410.02089","version":2},"attestation_state":"computed","paper":{"title":"RLEF: Grounding Code LLMs in Execution Feedback with Reinforcement Learning","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.AI"],"primary_cat":"cs.CL","authors_text":"Gabriel Synnaeve, Jade Copet, Jonas Gehring, Kunhao Zheng, Quentin Carbonneaux, Taco Cohen, Vegard Mella","submitted_at":"2024-10-02T23:25:17Z","abstract_excerpt":"Large language models (LLMs) deployed as agents solve user-specified tasks over multiple steps while keeping the required manual engagement to a minimum. Crucially, such LLMs need to ground their generations in any feedback obtained to reliably achieve the desired outcomes. We propose an end-to-end reinforcement learning method for teaching models to leverage execution feedback in the realm of code synthesis, where state-of-the-art LLMs struggle to improve code iteratively compared to independent sampling. We benchmark on competitive programming tasks, where we achieve new state-of-the art res"},"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":"2410.02089","kind":"arxiv","version":2},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.CL","submitted_at":"2024-10-02T23:25:17Z","cross_cats_sorted":["cs.AI"],"title_canon_sha256":"c50d4ca167ea926f2ce1754c14aa653203c62ca9daccd440508ac00dff7e6d4b","abstract_canon_sha256":"4c76b6a694bcc41a13d04ccf82a78ba89c5b17dc4edfa01ef9414304e1effa2e"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T10:16:13.316625Z","signature_b64":"DA/OR+GD52Vm89An/64OaiRYuSz2GYG9lbfvB6R8OQ4QIJTAY6ZJYuta/cVaw1I2b0luySMkpus7lB5JOR7VDw==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"cb18d7efd294397c14d2c5d672e09a59a5cea078a2802c450e4074ee254c93cf","last_reissued_at":"2026-07-05T10:16:13.316085Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T10:16:13.316085Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"RLEF: Grounding Code LLMs in Execution Feedback with Reinforcement Learning","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.AI"],"primary_cat":"cs.CL","authors_text":"Gabriel Synnaeve, Jade Copet, Jonas Gehring, Kunhao Zheng, Quentin Carbonneaux, Taco Cohen, Vegard Mella","submitted_at":"2024-10-02T23:25:17Z","abstract_excerpt":"Large language models (LLMs) deployed as agents solve user-specified tasks over multiple steps while keeping the required manual engagement to a minimum. Crucially, such LLMs need to ground their generations in any feedback obtained to reliably achieve the desired outcomes. We propose an end-to-end reinforcement learning method for teaching models to leverage execution feedback in the realm of code synthesis, where state-of-the-art LLMs struggle to improve code iteratively compared to independent sampling. We benchmark on competitive programming tasks, where we achieve new state-of-the art res"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2410.02089","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/2410.02089/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":"2410.02089","created_at":"2026-07-05T10:16:13.316146+00:00"},{"alias_kind":"arxiv_version","alias_value":"2410.02089v2","created_at":"2026-07-05T10:16:13.316146+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2410.02089","created_at":"2026-07-05T10:16:13.316146+00:00"},{"alias_kind":"pith_short_12","alias_value":"ZMMNP36SSQ4X","created_at":"2026-07-05T10:16:13.316146+00:00"},{"alias_kind":"pith_short_16","alias_value":"ZMMNP36SSQ4XYFGS","created_at":"2026-07-05T10:16:13.316146+00:00"},{"alias_kind":"pith_short_8","alias_value":"ZMMNP36S","created_at":"2026-07-05T10:16:13.316146+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":26,"internal_anchor_count":0,"sample":[{"citing_arxiv_id":"2606.21228","citing_title":"Sakana Fugu Technical Report","ref_index":298,"is_internal_anchor":false},{"citing_arxiv_id":"2607.02390","citing_title":"DecompRL: Solving Harder Problems by Learning Modular Code Generation","ref_index":21,"is_internal_anchor":false},{"citing_arxiv_id":"2606.03489","citing_title":"Learn from Your Mistakes: Tree-like Self-Play for Secure Code LLMs","ref_index":37,"is_internal_anchor":false},{"citing_arxiv_id":"2606.28438","citing_title":"When AI Reviews Its Own Code: Recursive Self-Training Collapse in Code LLMs","ref_index":167,"is_internal_anchor":false},{"citing_arxiv_id":"2605.28328","citing_title":"Learning the Error Patterns of Language Models","ref_index":13,"is_internal_anchor":false},{"citing_arxiv_id":"2503.19786","citing_title":"Gemma 3 Technical Report","ref_index":19,"is_internal_anchor":false},{"citing_arxiv_id":"2504.12501","citing_title":"Reinforcement Learning from Human Feedback","ref_index":166,"is_internal_anchor":false},{"citing_arxiv_id":"2605.22675","citing_title":"Self-Policy Distillation via Capability-Selective Subspace Projection","ref_index":19,"is_internal_anchor":false},{"citing_arxiv_id":"2605.21235","citing_title":"LamPO: A Lambda Style Policy Optimization for Reasoning Language Models","ref_index":4,"is_internal_anchor":false},{"citing_arxiv_id":"2605.17497","citing_title":"Self-Supervised On-Policy Distillation for Reasoning Language Models","ref_index":107,"is_internal_anchor":false},{"citing_arxiv_id":"2605.17792","citing_title":"HydroAgent: Closing the Gap Between Frontier LLMs and Human Experts in Hydrologic Model Calibration via Simulator-Grounded RL","ref_index":13,"is_internal_anchor":false},{"citing_arxiv_id":"2605.18747","citing_title":"Code as Agent Harness","ref_index":104,"is_internal_anchor":false},{"citing_arxiv_id":"2509.24552","citing_title":"Short window attention enables long-term memorization","ref_index":14,"is_internal_anchor":false},{"citing_arxiv_id":"2501.09686","citing_title":"Towards Large Reasoning Models: A Survey of Reinforced Reasoning with Large Language Models","ref_index":41,"is_internal_anchor":false},{"citing_arxiv_id":"2604.16335","citing_title":"Beyond Verifiable Rewards: Rubric-Based GRM for Reinforced Fine-Tuning SWE Agents","ref_index":8,"is_internal_anchor":false},{"citing_arxiv_id":"2502.18449","citing_title":"SWE-RL: Advancing LLM Reasoning via Reinforcement Learning on Open Software Evolution","ref_index":15,"is_internal_anchor":false},{"citing_arxiv_id":"2605.14539","citing_title":"Learning from Failures: Correction-Oriented Policy Optimization with Verifiable Rewards","ref_index":46,"is_internal_anchor":false},{"citing_arxiv_id":"2605.09134","citing_title":"BoostAPR: Boosting Automated Program Repair via Execution-Grounded Reinforcement Learning with Dual Reward Models","ref_index":89,"is_internal_anchor":false},{"citing_arxiv_id":"2503.09567","citing_title":"Towards Reasoning Era: A Survey of Long Chain-of-Thought for Reasoning Large Language Models","ref_index":206,"is_internal_anchor":false},{"citing_arxiv_id":"2605.08468","citing_title":"PYTHALAB-MERA: Validation-Grounded Memory, Retrieval, and Acceptance Control for Frozen-LLM Coding Agents","ref_index":15,"is_internal_anchor":false},{"citing_arxiv_id":"2605.09134","citing_title":"BoostAPR: Boosting Automated Program Repair via Execution-Grounded Reinforcement Learning with Dual Reward Models","ref_index":82,"is_internal_anchor":false},{"citing_arxiv_id":"2604.09813","citing_title":"Controllable and Verifiable Tool-Use Data Synthesis for Agentic Reinforcement Learning","ref_index":8,"is_internal_anchor":false},{"citing_arxiv_id":"2604.05560","citing_title":"An Iterative Test-and-Repair Framework for Competitive Code Generation","ref_index":13,"is_internal_anchor":false},{"citing_arxiv_id":"2604.06079","citing_title":"Scientific Graphics Program Synthesis via Dual Self-Consistency Reinforcement Learning","ref_index":17,"is_internal_anchor":false},{"citing_arxiv_id":"2604.04580","citing_title":"Beyond Fixed Tests: Repository-Level Issue Resolution as Coevolution of Code and Behavioral Constraints","ref_index":17,"is_internal_anchor":false}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/ZMMNP36SSQ4XYFGSYXLHFYE2LG","json":"https://pith.science/pith/ZMMNP36SSQ4XYFGSYXLHFYE2LG.json","graph_json":"https://pith.science/api/pith-number/ZMMNP36SSQ4XYFGSYXLHFYE2LG/graph.json","events_json":"https://pith.science/api/pith-number/ZMMNP36SSQ4XYFGSYXLHFYE2LG/events.json","paper":"https://pith.science/paper/ZMMNP36S"},"agent_actions":{"view_html":"https://pith.science/pith/ZMMNP36SSQ4XYFGSYXLHFYE2LG","download_json":"https://pith.science/pith/ZMMNP36SSQ4XYFGSYXLHFYE2LG.json","view_paper":"https://pith.science/paper/ZMMNP36S","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2410.02089&json=true","fetch_graph":"https://pith.science/api/pith-number/ZMMNP36SSQ4XYFGSYXLHFYE2LG/graph.json","fetch_events":"https://pith.science/api/pith-number/ZMMNP36SSQ4XYFGSYXLHFYE2LG/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/ZMMNP36SSQ4XYFGSYXLHFYE2LG/action/timestamp_anchor","attest_storage":"https://pith.science/pith/ZMMNP36SSQ4XYFGSYXLHFYE2LG/action/storage_attestation","attest_author":"https://pith.science/pith/ZMMNP36SSQ4XYFGSYXLHFYE2LG/action/author_attestation","sign_citation":"https://pith.science/pith/ZMMNP36SSQ4XYFGSYXLHFYE2LG/action/citation_signature","submit_replication":"https://pith.science/pith/ZMMNP36SSQ4XYFGSYXLHFYE2LG/action/replication_record"}},"created_at":"2026-07-05T10:16:13.316146+00:00","updated_at":"2026-07-05T10:16:13.316146+00:00"}