{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2025:Y73VMJFI2HFW5YFBHLWXNHCQZN","short_pith_number":"pith:Y73VMJFI","schema_version":"1.0","canonical_sha256":"c7f75624a8d1cb6ee0a13aed769c50cb7ba648c067905d9ba131d43e42fa7281","source":{"kind":"arxiv","id":"2505.23387","version":3},"attestation_state":"computed","paper":{"title":"Afterburner: Reinforcement Learning Facilitates Self-Improving Code Efficiency Optimization","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.AI"],"primary_cat":"cs.SE","authors_text":"Dong Huang, Luu Anh Tuan, Mingzhe Du, Qian Liu, See-kiong Ng, Xinyi He, Yue Liu, Yuhao Qing, Zejun Ma","submitted_at":"2025-05-29T12:14:29Z","abstract_excerpt":"Large Language Models (LLMs) generate functionally correct solutions but often fall short in code efficiency, a critical bottleneck for real-world deployment. In this paper, we introduce a novel test-time iterative optimization framework to address this, employing a closed-loop system where LLMs iteratively refine code based on empirical performance feedback from an execution sandbox. We explore three training strategies: Supervised Fine-Tuning (SFT), Direct Preference Optimization (DPO), and Group Relative Policy Optimization (GRPO). Experiments on our Venus dataset and the APPS benchmark sho"},"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":"2505.23387","kind":"arxiv","version":3},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.SE","submitted_at":"2025-05-29T12:14:29Z","cross_cats_sorted":["cs.AI"],"title_canon_sha256":"9f6cf7b81ddac9dc10d269f04d1c3845cbda9655d9d5cccfd67b2ba74dbf87c5","abstract_canon_sha256":"891dfba1256e1b3cb6a10b88204c8861858e83d8a84ad1c660a491ed51a8cca9"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T11:15:00.698571Z","signature_b64":"7AyeWjBkoJRuZvIkEtcyKqpvZacUW6DdKcnl5sWw7ivF7KblAFECmNgEWuvT99sYwZPhewiwpm7G0s26jrfXCQ==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"c7f75624a8d1cb6ee0a13aed769c50cb7ba648c067905d9ba131d43e42fa7281","last_reissued_at":"2026-07-05T11:15:00.698058Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T11:15:00.698058Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Afterburner: Reinforcement Learning Facilitates Self-Improving Code Efficiency Optimization","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.AI"],"primary_cat":"cs.SE","authors_text":"Dong Huang, Luu Anh Tuan, Mingzhe Du, Qian Liu, See-kiong Ng, Xinyi He, Yue Liu, Yuhao Qing, Zejun Ma","submitted_at":"2025-05-29T12:14:29Z","abstract_excerpt":"Large Language Models (LLMs) generate functionally correct solutions but often fall short in code efficiency, a critical bottleneck for real-world deployment. In this paper, we introduce a novel test-time iterative optimization framework to address this, employing a closed-loop system where LLMs iteratively refine code based on empirical performance feedback from an execution sandbox. We explore three training strategies: Supervised Fine-Tuning (SFT), Direct Preference Optimization (DPO), and Group Relative Policy Optimization (GRPO). Experiments on our Venus dataset and the APPS benchmark sho"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2505.23387","kind":"arxiv","version":3},"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/2505.23387/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":"2505.23387","created_at":"2026-07-05T11:15:00.698121+00:00"},{"alias_kind":"arxiv_version","alias_value":"2505.23387v3","created_at":"2026-07-05T11:15:00.698121+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2505.23387","created_at":"2026-07-05T11:15:00.698121+00:00"},{"alias_kind":"pith_short_12","alias_value":"Y73VMJFI2HFW","created_at":"2026-07-05T11:15:00.698121+00:00"},{"alias_kind":"pith_short_16","alias_value":"Y73VMJFI2HFW5YFB","created_at":"2026-07-05T11:15:00.698121+00:00"},{"alias_kind":"pith_short_8","alias_value":"Y73VMJFI","created_at":"2026-07-05T11:15:00.698121+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":9,"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":10,"is_internal_anchor":false},{"citing_arxiv_id":"2606.06821","citing_title":"Chiseling Out Efficiency: Structured Skeleton Supervision for Efficient Code Generation","ref_index":10,"is_internal_anchor":false},{"citing_arxiv_id":"2605.28751","citing_title":"Extrapolative Weight Averaging Reveals Correctness-Efficiency Frontiers in Code RL","ref_index":8,"is_internal_anchor":false},{"citing_arxiv_id":"2606.11052","citing_title":"Attention Amnesia in Hybrid LLMs: When CoT Fine-Tuning Breaks Long-Range Recall, and How to Fix It","ref_index":22,"is_internal_anchor":false},{"citing_arxiv_id":"2606.17682","citing_title":"From Trainee to Trainer: LLM-Designed Training Environment for RL with Multi-Agent Reasoning","ref_index":22,"is_internal_anchor":false},{"citing_arxiv_id":"2605.17978","citing_title":"AutoVecCoder: Teaching LLMs to Generate Explicitly Vectorized Code","ref_index":65,"is_internal_anchor":false},{"citing_arxiv_id":"2509.08827","citing_title":"A Survey of Reinforcement Learning for Large Reasoning Models","ref_index":119,"is_internal_anchor":false},{"citing_arxiv_id":"2604.02352","citing_title":"An Initial Exploration of Contrastive Prompt Tuning to Generate Energy-Efficient Code","ref_index":7,"is_internal_anchor":false},{"citing_arxiv_id":"2604.04562","citing_title":"Paper Espresso: From Paper Overload to Research Insight","ref_index":14,"is_internal_anchor":false}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/Y73VMJFI2HFW5YFBHLWXNHCQZN","json":"https://pith.science/pith/Y73VMJFI2HFW5YFBHLWXNHCQZN.json","graph_json":"https://pith.science/api/pith-number/Y73VMJFI2HFW5YFBHLWXNHCQZN/graph.json","events_json":"https://pith.science/api/pith-number/Y73VMJFI2HFW5YFBHLWXNHCQZN/events.json","paper":"https://pith.science/paper/Y73VMJFI"},"agent_actions":{"view_html":"https://pith.science/pith/Y73VMJFI2HFW5YFBHLWXNHCQZN","download_json":"https://pith.science/pith/Y73VMJFI2HFW5YFBHLWXNHCQZN.json","view_paper":"https://pith.science/paper/Y73VMJFI","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2505.23387&json=true","fetch_graph":"https://pith.science/api/pith-number/Y73VMJFI2HFW5YFBHLWXNHCQZN/graph.json","fetch_events":"https://pith.science/api/pith-number/Y73VMJFI2HFW5YFBHLWXNHCQZN/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/Y73VMJFI2HFW5YFBHLWXNHCQZN/action/timestamp_anchor","attest_storage":"https://pith.science/pith/Y73VMJFI2HFW5YFBHLWXNHCQZN/action/storage_attestation","attest_author":"https://pith.science/pith/Y73VMJFI2HFW5YFBHLWXNHCQZN/action/author_attestation","sign_citation":"https://pith.science/pith/Y73VMJFI2HFW5YFBHLWXNHCQZN/action/citation_signature","submit_replication":"https://pith.science/pith/Y73VMJFI2HFW5YFBHLWXNHCQZN/action/replication_record"}},"created_at":"2026-07-05T11:15:00.698121+00:00","updated_at":"2026-07-05T11:15:00.698121+00:00"}