{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2024:LFLEKQV3WXZM4SFGNMRQJFF23Z","short_pith_number":"pith:LFLEKQV3","schema_version":"1.0","canonical_sha256":"59564542bbb5f2ce48a66b230494bade63718eb2537818667eb1c7bd174e5687","source":{"kind":"arxiv","id":"2405.18649","version":2},"attestation_state":"computed","paper":{"title":"LeDex: Training LLMs to Better Self-Debug and Explain Code","license":"http://creativecommons.org/licenses/by-nc-nd/4.0/","headline":"","cross_cats":["cs.AI","cs.SE"],"primary_cat":"cs.CL","authors_text":"Anoop Deoras, Baishakhi Ray, Nan Jiang, Qiang Zhou, Shiqi Wang, Soneya Binta Hossain, Varun Kumar, Xiaofei Ma, Xiaopeng Li","submitted_at":"2024-05-28T23:20:24Z","abstract_excerpt":"In the domain of code generation, self-debugging is crucial. It allows LLMs to refine their generated code based on execution feedback. This is particularly important because generating correct solutions in one attempt proves challenging for complex tasks. Prior works on self-debugging mostly focus on prompting methods by providing LLMs with few-shot examples, which work poorly on small open-sourced LLMs. In this work, we propose LeDex, a training framework that significantly improves the self-debugging capability of LLMs. Intuitively, we observe that a chain of explanations on the wrong code "},"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":"2405.18649","kind":"arxiv","version":2},"metadata":{"license":"http://creativecommons.org/licenses/by-nc-nd/4.0/","primary_cat":"cs.CL","submitted_at":"2024-05-28T23:20:24Z","cross_cats_sorted":["cs.AI","cs.SE"],"title_canon_sha256":"7789c46e8498720473b21fe71f872e4d7729fcdf29d0578d16bc36c597964e79","abstract_canon_sha256":"0138d2f2dcc4dfff8c646c3329cd21939d19457ceb11b6c4e05981cd14671df4"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T10:14:03.457407Z","signature_b64":"vnE5HWfq+mtYbAD9jOe/Jxz1d9ZxfcppF8L2lFv3uBhNkv/SNZZp+h7kFnhhl2R7X5ugtDOFYJ9GVmByd7NlBw==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"59564542bbb5f2ce48a66b230494bade63718eb2537818667eb1c7bd174e5687","last_reissued_at":"2026-07-05T10:14:03.456854Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T10:14:03.456854Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"LeDex: Training LLMs to Better Self-Debug and Explain Code","license":"http://creativecommons.org/licenses/by-nc-nd/4.0/","headline":"","cross_cats":["cs.AI","cs.SE"],"primary_cat":"cs.CL","authors_text":"Anoop Deoras, Baishakhi Ray, Nan Jiang, Qiang Zhou, Shiqi Wang, Soneya Binta Hossain, Varun Kumar, Xiaofei Ma, Xiaopeng Li","submitted_at":"2024-05-28T23:20:24Z","abstract_excerpt":"In the domain of code generation, self-debugging is crucial. It allows LLMs to refine their generated code based on execution feedback. This is particularly important because generating correct solutions in one attempt proves challenging for complex tasks. Prior works on self-debugging mostly focus on prompting methods by providing LLMs with few-shot examples, which work poorly on small open-sourced LLMs. In this work, we propose LeDex, a training framework that significantly improves the self-debugging capability of LLMs. Intuitively, we observe that a chain of explanations on the wrong code "},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2405.18649","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/2405.18649/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":"2405.18649","created_at":"2026-07-05T10:14:03.456919+00:00"},{"alias_kind":"arxiv_version","alias_value":"2405.18649v2","created_at":"2026-07-05T10:14:03.456919+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2405.18649","created_at":"2026-07-05T10:14:03.456919+00:00"},{"alias_kind":"pith_short_12","alias_value":"LFLEKQV3WXZM","created_at":"2026-07-05T10:14:03.456919+00:00"},{"alias_kind":"pith_short_16","alias_value":"LFLEKQV3WXZM4SFG","created_at":"2026-07-05T10:14:03.456919+00:00"},{"alias_kind":"pith_short_8","alias_value":"LFLEKQV3","created_at":"2026-07-05T10:14:03.456919+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":3,"internal_anchor_count":0,"sample":[{"citing_arxiv_id":"2606.31511","citing_title":"Falsification, Not Exposure: An Internally Preregistered Placebo-Controlled Decomposition of Self-Repair Feedback in Frozen Small Code Models","ref_index":21,"is_internal_anchor":false},{"citing_arxiv_id":"2604.03245","citing_title":"FVRuleLearner: Operator-Level Reasoning Tree (Op-Tree)-Based Rules Learning for Formal Verification","ref_index":46,"is_internal_anchor":false},{"citing_arxiv_id":"2604.25737","citing_title":"SAFEdit: Does Multi-Agent Decomposition Resolve the Reliability Challenges of Instructed Code Editing?","ref_index":10,"is_internal_anchor":false}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/LFLEKQV3WXZM4SFGNMRQJFF23Z","json":"https://pith.science/pith/LFLEKQV3WXZM4SFGNMRQJFF23Z.json","graph_json":"https://pith.science/api/pith-number/LFLEKQV3WXZM4SFGNMRQJFF23Z/graph.json","events_json":"https://pith.science/api/pith-number/LFLEKQV3WXZM4SFGNMRQJFF23Z/events.json","paper":"https://pith.science/paper/LFLEKQV3"},"agent_actions":{"view_html":"https://pith.science/pith/LFLEKQV3WXZM4SFGNMRQJFF23Z","download_json":"https://pith.science/pith/LFLEKQV3WXZM4SFGNMRQJFF23Z.json","view_paper":"https://pith.science/paper/LFLEKQV3","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2405.18649&json=true","fetch_graph":"https://pith.science/api/pith-number/LFLEKQV3WXZM4SFGNMRQJFF23Z/graph.json","fetch_events":"https://pith.science/api/pith-number/LFLEKQV3WXZM4SFGNMRQJFF23Z/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/LFLEKQV3WXZM4SFGNMRQJFF23Z/action/timestamp_anchor","attest_storage":"https://pith.science/pith/LFLEKQV3WXZM4SFGNMRQJFF23Z/action/storage_attestation","attest_author":"https://pith.science/pith/LFLEKQV3WXZM4SFGNMRQJFF23Z/action/author_attestation","sign_citation":"https://pith.science/pith/LFLEKQV3WXZM4SFGNMRQJFF23Z/action/citation_signature","submit_replication":"https://pith.science/pith/LFLEKQV3WXZM4SFGNMRQJFF23Z/action/replication_record"}},"created_at":"2026-07-05T10:14:03.456919+00:00","updated_at":"2026-07-05T10:14:03.456919+00:00"}