{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2025:NETQRCQWEJD6OAXLKV3JKFZHUP","short_pith_number":"pith:NETQRCQW","schema_version":"1.0","canonical_sha256":"6927088a162247e702eb5576951727a3f5041cc06446fdc0c45a0a836a9a2674","source":{"kind":"arxiv","id":"2501.01054","version":1},"attestation_state":"computed","paper":{"title":"Dynamic Scaling of Unit Tests for Code Reward Modeling","license":"http://creativecommons.org/licenses/by-sa/4.0/","headline":"","cross_cats":["cs.SE"],"primary_cat":"cs.CL","authors_text":"Jie Tang, Jifan Yu, Jing Zhang, Sijia Luo, Xiaokang Zhang, Zeyao Ma","submitted_at":"2025-01-02T04:33:31Z","abstract_excerpt":"Current large language models (LLMs) often struggle to produce accurate responses on the first attempt for complex reasoning tasks like code generation. Prior research tackles this challenge by generating multiple candidate solutions and validating them with LLM-generated unit tests. The execution results of unit tests serve as reward signals to identify correct solutions. As LLMs always confidently make mistakes, these unit tests are not reliable, thereby diminishing the quality of reward signals. Motivated by the observation that scaling the number of solutions improves LLM performance, we e"},"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":"2501.01054","kind":"arxiv","version":1},"metadata":{"license":"http://creativecommons.org/licenses/by-sa/4.0/","primary_cat":"cs.CL","submitted_at":"2025-01-02T04:33:31Z","cross_cats_sorted":["cs.SE"],"title_canon_sha256":"7bb3263b72703751df1c58f4f57ea22d3553aeae7f5683ba989c779f545f101f","abstract_canon_sha256":"232b6fe5be8e89aeb7302a482ba6225e3d5281492be1d7e1bcf367232e20f02c"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T09:56:13.630131Z","signature_b64":"CNKO1ZMsV0VEshVRkbnKbNN+dTXa0L67K9sHYqrDVcNipSnsoS8OpTlI8nqM+aGNQqMJjLI8YesrERY0GoH6CA==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"6927088a162247e702eb5576951727a3f5041cc06446fdc0c45a0a836a9a2674","last_reissued_at":"2026-07-05T09:56:13.629717Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T09:56:13.629717Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Dynamic Scaling of Unit Tests for Code Reward Modeling","license":"http://creativecommons.org/licenses/by-sa/4.0/","headline":"","cross_cats":["cs.SE"],"primary_cat":"cs.CL","authors_text":"Jie Tang, Jifan Yu, Jing Zhang, Sijia Luo, Xiaokang Zhang, Zeyao Ma","submitted_at":"2025-01-02T04:33:31Z","abstract_excerpt":"Current large language models (LLMs) often struggle to produce accurate responses on the first attempt for complex reasoning tasks like code generation. Prior research tackles this challenge by generating multiple candidate solutions and validating them with LLM-generated unit tests. The execution results of unit tests serve as reward signals to identify correct solutions. As LLMs always confidently make mistakes, these unit tests are not reliable, thereby diminishing the quality of reward signals. Motivated by the observation that scaling the number of solutions improves LLM performance, we e"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2501.01054","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/2501.01054/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":"2501.01054","created_at":"2026-07-05T09:56:13.629773+00:00"},{"alias_kind":"arxiv_version","alias_value":"2501.01054v1","created_at":"2026-07-05T09:56:13.629773+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2501.01054","created_at":"2026-07-05T09:56:13.629773+00:00"},{"alias_kind":"pith_short_12","alias_value":"NETQRCQWEJD6","created_at":"2026-07-05T09:56:13.629773+00:00"},{"alias_kind":"pith_short_16","alias_value":"NETQRCQWEJD6OAXL","created_at":"2026-07-05T09:56:13.629773+00:00"},{"alias_kind":"pith_short_8","alias_value":"NETQRCQW","created_at":"2026-07-05T09:56:13.629773+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":5,"internal_anchor_count":0,"sample":[{"citing_arxiv_id":"2606.25450","citing_title":"The Generalization Spectrum: A Chromatographic Approach to Evaluating Learning Algorithms","ref_index":31,"is_internal_anchor":false},{"citing_arxiv_id":"2606.25450","citing_title":"The Generalization Spectrum: A Chromatographic Approach to Evaluating Learning Algorithms","ref_index":31,"is_internal_anchor":false},{"citing_arxiv_id":"2605.17497","citing_title":"Self-Supervised On-Policy Distillation for Reasoning Language Models","ref_index":108,"is_internal_anchor":false},{"citing_arxiv_id":"2605.12947","citing_title":"When Should an AI Workflow Release? Always-Valid Inference for Black-Box Generate-Verify Systems","ref_index":14,"is_internal_anchor":false},{"citing_arxiv_id":"2604.05560","citing_title":"An Iterative Test-and-Repair Framework for Competitive Code Generation","ref_index":35,"is_internal_anchor":false}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/NETQRCQWEJD6OAXLKV3JKFZHUP","json":"https://pith.science/pith/NETQRCQWEJD6OAXLKV3JKFZHUP.json","graph_json":"https://pith.science/api/pith-number/NETQRCQWEJD6OAXLKV3JKFZHUP/graph.json","events_json":"https://pith.science/api/pith-number/NETQRCQWEJD6OAXLKV3JKFZHUP/events.json","paper":"https://pith.science/paper/NETQRCQW"},"agent_actions":{"view_html":"https://pith.science/pith/NETQRCQWEJD6OAXLKV3JKFZHUP","download_json":"https://pith.science/pith/NETQRCQWEJD6OAXLKV3JKFZHUP.json","view_paper":"https://pith.science/paper/NETQRCQW","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2501.01054&json=true","fetch_graph":"https://pith.science/api/pith-number/NETQRCQWEJD6OAXLKV3JKFZHUP/graph.json","fetch_events":"https://pith.science/api/pith-number/NETQRCQWEJD6OAXLKV3JKFZHUP/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/NETQRCQWEJD6OAXLKV3JKFZHUP/action/timestamp_anchor","attest_storage":"https://pith.science/pith/NETQRCQWEJD6OAXLKV3JKFZHUP/action/storage_attestation","attest_author":"https://pith.science/pith/NETQRCQWEJD6OAXLKV3JKFZHUP/action/author_attestation","sign_citation":"https://pith.science/pith/NETQRCQWEJD6OAXLKV3JKFZHUP/action/citation_signature","submit_replication":"https://pith.science/pith/NETQRCQWEJD6OAXLKV3JKFZHUP/action/replication_record"}},"created_at":"2026-07-05T09:56:13.629773+00:00","updated_at":"2026-07-05T09:56:13.629773+00:00"}