{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2026:YVMFQLNWOZKQF3KLAL637FOSJF","short_pith_number":"pith:YVMFQLNW","schema_version":"1.0","canonical_sha256":"c558582db6765502ed4b02fdbf95d2496c9af7e0f44a9228c3a3717763ac1b60","source":{"kind":"arxiv","id":"2607.21069","version":1},"attestation_state":"computed","paper":{"title":"From Evaluation to Optimisation: Hierarchy-Aware Training Signals for CWE Prediction in Python","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":[],"primary_cat":"cs.LG","authors_text":"Carlos C. N. Kuhn, Manile Srun, Muntasir Adnan","submitted_at":"2026-07-23T09:02:46Z","abstract_excerpt":"The original ALPHA benchmark introduced a taxonomy-aware penalty for evaluating CWE-level vulnerability prediction in Python and proposed that the penalty could theoretically also serve as a training signal. This paper provides that validation. We compare three delivery mechanisms: supervised fine-tuning, a dual-head classification loss, and reinforcement learning with a dense reward derived from the normalised penalty. We find that supervised approaches consistently regress below the zero-shot baseline under distribution shift, while GRPO succeeds. Our best policy reduces the cumulative ALPHA"},"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":"2607.21069","kind":"arxiv","version":1},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.LG","submitted_at":"2026-07-23T09:02:46Z","cross_cats_sorted":[],"title_canon_sha256":"19fb16d3bf39a5f5f4166426db55a7f98131a3b36093b5383e1ad317e064f54c","abstract_canon_sha256":"fab90561fa6d1cc3d8ae4dd7f6c022e1e62119c0ad87fb61ec0647a649d4ae6d"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-24T01:23:50.246516Z","signature_b64":"02X6hn37REUsMffkVQegskAAjLJTi7iMc/BVKRa+MxQC87p5iqQPcCC5uxhYtDpXiWGiGTY8oYTO4rh3l+RdDw==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"c558582db6765502ed4b02fdbf95d2496c9af7e0f44a9228c3a3717763ac1b60","last_reissued_at":"2026-07-24T01:23:50.245655Z","signature_status":"signed_v1","first_computed_at":"2026-07-24T01:23:50.245655Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"From Evaluation to Optimisation: Hierarchy-Aware Training Signals for CWE Prediction in Python","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":[],"primary_cat":"cs.LG","authors_text":"Carlos C. N. Kuhn, Manile Srun, Muntasir Adnan","submitted_at":"2026-07-23T09:02:46Z","abstract_excerpt":"The original ALPHA benchmark introduced a taxonomy-aware penalty for evaluating CWE-level vulnerability prediction in Python and proposed that the penalty could theoretically also serve as a training signal. This paper provides that validation. We compare three delivery mechanisms: supervised fine-tuning, a dual-head classification loss, and reinforcement learning with a dense reward derived from the normalised penalty. We find that supervised approaches consistently regress below the zero-shot baseline under distribution shift, while GRPO succeeds. Our best policy reduces the cumulative ALPHA"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2607.21069","kind":"arxiv","version":1},"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/2607.21069/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":"2607.21069","created_at":"2026-07-24T01:23:50.246098+00:00"},{"alias_kind":"arxiv_version","alias_value":"2607.21069v1","created_at":"2026-07-24T01:23:50.246098+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2607.21069","created_at":"2026-07-24T01:23:50.246098+00:00"},{"alias_kind":"pith_short_12","alias_value":"YVMFQLNWOZKQ","created_at":"2026-07-24T01:23:50.246098+00:00"},{"alias_kind":"pith_short_16","alias_value":"YVMFQLNWOZKQF3KL","created_at":"2026-07-24T01:23:50.246098+00:00"},{"alias_kind":"pith_short_8","alias_value":"YVMFQLNW","created_at":"2026-07-24T01:23:50.246098+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":0,"internal_anchor_count":0,"sample":[]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/YVMFQLNWOZKQF3KLAL637FOSJF","json":"https://pith.science/pith/YVMFQLNWOZKQF3KLAL637FOSJF.json","graph_json":"https://pith.science/api/pith-number/YVMFQLNWOZKQF3KLAL637FOSJF/graph.json","events_json":"https://pith.science/api/pith-number/YVMFQLNWOZKQF3KLAL637FOSJF/events.json","paper":"https://pith.science/paper/YVMFQLNW"},"agent_actions":{"view_html":"https://pith.science/pith/YVMFQLNWOZKQF3KLAL637FOSJF","download_json":"https://pith.science/pith/YVMFQLNWOZKQF3KLAL637FOSJF.json","view_paper":"https://pith.science/paper/YVMFQLNW","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2607.21069&json=true","fetch_graph":"https://pith.science/api/pith-number/YVMFQLNWOZKQF3KLAL637FOSJF/graph.json","fetch_events":"https://pith.science/api/pith-number/YVMFQLNWOZKQF3KLAL637FOSJF/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/YVMFQLNWOZKQF3KLAL637FOSJF/action/timestamp_anchor","attest_storage":"https://pith.science/pith/YVMFQLNWOZKQF3KLAL637FOSJF/action/storage_attestation","attest_author":"https://pith.science/pith/YVMFQLNWOZKQF3KLAL637FOSJF/action/author_attestation","sign_citation":"https://pith.science/pith/YVMFQLNWOZKQF3KLAL637FOSJF/action/citation_signature","submit_replication":"https://pith.science/pith/YVMFQLNWOZKQF3KLAL637FOSJF/action/replication_record"}},"created_at":"2026-07-24T01:23:50.246098+00:00","updated_at":"2026-07-24T01:23:50.246098+00:00"}