{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2024:PCP4RUSZHGWN757TEDGFY467J4","short_pith_number":"pith:PCP4RUSZ","schema_version":"1.0","canonical_sha256":"789fc8d25939acdff7f320cc5c73df4f04686314ea4f230feded2921d2664b40","source":{"kind":"arxiv","id":"2402.08699","version":2},"attestation_state":"computed","paper":{"title":"Unsupervised Evaluation of Code LLMs with Round-Trip Correctness","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":["cs.LG"],"primary_cat":"cs.SE","authors_text":"Miltiadis Allamanis, Pengcheng Yin, Sheena Panthaplackel","submitted_at":"2024-02-13T11:08:08Z","abstract_excerpt":"To evaluate code large language models (LLMs), research has relied on a few small manually curated benchmarks, such as HumanEval and MBPP, which represent a narrow part of the real-world software domains. In this work, we introduce round-trip correctness (RTC) as an alternative evaluation method. RTC allows Code LLM evaluation on a broader spectrum of real-world software domains without the need for costly human curation. RTC rests on the idea that we can ask a model to make a prediction (e.g., describe some code using natural language), feed that prediction back (e.g., synthesize code from th"},"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":"2402.08699","kind":"arxiv","version":2},"metadata":{"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.SE","submitted_at":"2024-02-13T11:08:08Z","cross_cats_sorted":["cs.LG"],"title_canon_sha256":"124432447114444b37924d396132ef9bb7b3ed63de4e0eeddc6f56f670dad72d","abstract_canon_sha256":"5460b54378ade90e97958da63f5b3291f0baef9e9d3150174c179678926764f9"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T08:23:17.959255Z","signature_b64":"vkHOXoJIHTuB/eUxQUNyCTOxBcv7pIyni/qo+If+vegQ9S1rIutuBhCxg59rDbsxRm7Ze40EKDcUW4wTv2mHBA==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"789fc8d25939acdff7f320cc5c73df4f04686314ea4f230feded2921d2664b40","last_reissued_at":"2026-07-05T08:23:17.958783Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T08:23:17.958783Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Unsupervised Evaluation of Code LLMs with Round-Trip Correctness","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":["cs.LG"],"primary_cat":"cs.SE","authors_text":"Miltiadis Allamanis, Pengcheng Yin, Sheena Panthaplackel","submitted_at":"2024-02-13T11:08:08Z","abstract_excerpt":"To evaluate code large language models (LLMs), research has relied on a few small manually curated benchmarks, such as HumanEval and MBPP, which represent a narrow part of the real-world software domains. In this work, we introduce round-trip correctness (RTC) as an alternative evaluation method. RTC allows Code LLM evaluation on a broader spectrum of real-world software domains without the need for costly human curation. RTC rests on the idea that we can ask a model to make a prediction (e.g., describe some code using natural language), feed that prediction back (e.g., synthesize code from th"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2402.08699","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/2402.08699/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":"2402.08699","created_at":"2026-07-05T08:23:17.958841+00:00"},{"alias_kind":"arxiv_version","alias_value":"2402.08699v2","created_at":"2026-07-05T08:23:17.958841+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2402.08699","created_at":"2026-07-05T08:23:17.958841+00:00"},{"alias_kind":"pith_short_12","alias_value":"PCP4RUSZHGWN","created_at":"2026-07-05T08:23:17.958841+00:00"},{"alias_kind":"pith_short_16","alias_value":"PCP4RUSZHGWN757T","created_at":"2026-07-05T08:23:17.958841+00:00"},{"alias_kind":"pith_short_8","alias_value":"PCP4RUSZ","created_at":"2026-07-05T08:23:17.958841+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":2,"internal_anchor_count":0,"sample":[{"citing_arxiv_id":"2606.09145","citing_title":"PrivCode++: Latent-Conditioned Differentially Private Code Generation for Comprehensive Guarantees","ref_index":68,"is_internal_anchor":false},{"citing_arxiv_id":"2605.13817","citing_title":"Neurosymbolic Auditing of Natural-Language Software Requirements","ref_index":2,"is_internal_anchor":false}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/PCP4RUSZHGWN757TEDGFY467J4","json":"https://pith.science/pith/PCP4RUSZHGWN757TEDGFY467J4.json","graph_json":"https://pith.science/api/pith-number/PCP4RUSZHGWN757TEDGFY467J4/graph.json","events_json":"https://pith.science/api/pith-number/PCP4RUSZHGWN757TEDGFY467J4/events.json","paper":"https://pith.science/paper/PCP4RUSZ"},"agent_actions":{"view_html":"https://pith.science/pith/PCP4RUSZHGWN757TEDGFY467J4","download_json":"https://pith.science/pith/PCP4RUSZHGWN757TEDGFY467J4.json","view_paper":"https://pith.science/paper/PCP4RUSZ","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2402.08699&json=true","fetch_graph":"https://pith.science/api/pith-number/PCP4RUSZHGWN757TEDGFY467J4/graph.json","fetch_events":"https://pith.science/api/pith-number/PCP4RUSZHGWN757TEDGFY467J4/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/PCP4RUSZHGWN757TEDGFY467J4/action/timestamp_anchor","attest_storage":"https://pith.science/pith/PCP4RUSZHGWN757TEDGFY467J4/action/storage_attestation","attest_author":"https://pith.science/pith/PCP4RUSZHGWN757TEDGFY467J4/action/author_attestation","sign_citation":"https://pith.science/pith/PCP4RUSZHGWN757TEDGFY467J4/action/citation_signature","submit_replication":"https://pith.science/pith/PCP4RUSZHGWN757TEDGFY467J4/action/replication_record"}},"created_at":"2026-07-05T08:23:17.958841+00:00","updated_at":"2026-07-05T08:23:17.958841+00:00"}