{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2023:BMBQOAO7BVAOAGCK6KHQDLT6RA","short_pith_number":"pith:BMBQOAO7","schema_version":"1.0","canonical_sha256":"0b030701df0d40e0184af28f01ae7e8835f5ea5031b7031c85e18f4337bdea39","source":{"kind":"arxiv","id":"2306.03268","version":3},"attestation_state":"computed","paper":{"title":"Skill over Scale: The Case for Medium, Domain-Specific Models for SE","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.SE"],"primary_cat":"cs.CL","authors_text":"Manisha Mukherjee, Vincent J. Hellendoorn","submitted_at":"2023-06-05T21:38:30Z","abstract_excerpt":"Recent advancements in AI have sparked a trend in constructing large, generalist language models that handle a multitude of tasks, including many code-related ones. While these models are expensive to train and are often closed-source, they have enjoyed broad adoption because they tend to outperform smaller, domain-specific models of code. In this work, we argue that this is not a foregone conclusion. We show that modestly sized domain-specific models can outperform much larger ones on code labeling tasks, provided they are trained to the same standards. Concretely, we focus on StackOverflow ("},"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":"2306.03268","kind":"arxiv","version":3},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.CL","submitted_at":"2023-06-05T21:38:30Z","cross_cats_sorted":["cs.SE"],"title_canon_sha256":"785f4560aec91b3a14cb8294ffa2ef202a77b5b293270eea7b75691108322666","abstract_canon_sha256":"c32f9c5a80dc44fc5009debf9ffec0a9c5ac8eeb2c47e107b2059ae4973749aa"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T10:17:45.021617Z","signature_b64":"CjuFh4qOgeEYY6dvl4FDCbUxxoUCIjiH334iB82qRDin+FFnJVvddB0bwuZffLXxphg9oX+f1gRZX7KwEFiVDQ==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"0b030701df0d40e0184af28f01ae7e8835f5ea5031b7031c85e18f4337bdea39","last_reissued_at":"2026-07-05T10:17:45.021128Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T10:17:45.021128Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Skill over Scale: The Case for Medium, Domain-Specific Models for SE","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.SE"],"primary_cat":"cs.CL","authors_text":"Manisha Mukherjee, Vincent J. Hellendoorn","submitted_at":"2023-06-05T21:38:30Z","abstract_excerpt":"Recent advancements in AI have sparked a trend in constructing large, generalist language models that handle a multitude of tasks, including many code-related ones. While these models are expensive to train and are often closed-source, they have enjoyed broad adoption because they tend to outperform smaller, domain-specific models of code. In this work, we argue that this is not a foregone conclusion. We show that modestly sized domain-specific models can outperform much larger ones on code labeling tasks, provided they are trained to the same standards. Concretely, we focus on StackOverflow ("},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2306.03268","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/2306.03268/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":"2306.03268","created_at":"2026-07-05T10:17:45.021207+00:00"},{"alias_kind":"arxiv_version","alias_value":"2306.03268v3","created_at":"2026-07-05T10:17:45.021207+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2306.03268","created_at":"2026-07-05T10:17:45.021207+00:00"},{"alias_kind":"pith_short_12","alias_value":"BMBQOAO7BVAO","created_at":"2026-07-05T10:17:45.021207+00:00"},{"alias_kind":"pith_short_16","alias_value":"BMBQOAO7BVAOAGCK","created_at":"2026-07-05T10:17:45.021207+00:00"},{"alias_kind":"pith_short_8","alias_value":"BMBQOAO7","created_at":"2026-07-05T10:17:45.021207+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":1,"internal_anchor_count":1,"sample":[{"citing_arxiv_id":"2505.21069","citing_title":"CXXCrafter: An LLM-Based Agent for Automated C/C++ Open Source Software Building","ref_index":34,"is_internal_anchor":true}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/BMBQOAO7BVAOAGCK6KHQDLT6RA","json":"https://pith.science/pith/BMBQOAO7BVAOAGCK6KHQDLT6RA.json","graph_json":"https://pith.science/api/pith-number/BMBQOAO7BVAOAGCK6KHQDLT6RA/graph.json","events_json":"https://pith.science/api/pith-number/BMBQOAO7BVAOAGCK6KHQDLT6RA/events.json","paper":"https://pith.science/paper/BMBQOAO7"},"agent_actions":{"view_html":"https://pith.science/pith/BMBQOAO7BVAOAGCK6KHQDLT6RA","download_json":"https://pith.science/pith/BMBQOAO7BVAOAGCK6KHQDLT6RA.json","view_paper":"https://pith.science/paper/BMBQOAO7","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2306.03268&json=true","fetch_graph":"https://pith.science/api/pith-number/BMBQOAO7BVAOAGCK6KHQDLT6RA/graph.json","fetch_events":"https://pith.science/api/pith-number/BMBQOAO7BVAOAGCK6KHQDLT6RA/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/BMBQOAO7BVAOAGCK6KHQDLT6RA/action/timestamp_anchor","attest_storage":"https://pith.science/pith/BMBQOAO7BVAOAGCK6KHQDLT6RA/action/storage_attestation","attest_author":"https://pith.science/pith/BMBQOAO7BVAOAGCK6KHQDLT6RA/action/author_attestation","sign_citation":"https://pith.science/pith/BMBQOAO7BVAOAGCK6KHQDLT6RA/action/citation_signature","submit_replication":"https://pith.science/pith/BMBQOAO7BVAOAGCK6KHQDLT6RA/action/replication_record"}},"created_at":"2026-07-05T10:17:45.021207+00:00","updated_at":"2026-07-05T10:17:45.021207+00:00"}