{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2024:FXVIQKQDZUGKQS4DCBQKMM6LXS","short_pith_number":"pith:FXVIQKQD","schema_version":"1.0","canonical_sha256":"2dea882a03cd0ca84b831060a633cbbc93955e25e205f742c04a6fdede06d580","source":{"kind":"arxiv","id":"2404.00725","version":2},"attestation_state":"computed","paper":{"title":"The Larger the Better? Improved LLM Code-Generation via Budget Reallocation","license":"http://creativecommons.org/publicdomain/zero/1.0/","headline":"","cross_cats":["cs.AI","cs.CL","cs.LG"],"primary_cat":"cs.SE","authors_text":"Jonas Gehring, Michael Hassid, Roy Schwartz, Tal Remez, Yossi Adi","submitted_at":"2024-03-31T15:55:49Z","abstract_excerpt":"It is a common belief that large language models (LLMs) are better than smaller-sized ones. However, larger models also require significantly more time and compute during inference. This begs the question: what happens when both models operate under the same budget? (e.g., compute, run-time). To address this question, we analyze code generation LLMs of various sizes and make comparisons such as running a 70B model once vs. generating five outputs from a 13B model. We consider a standard unit-test setup, which can be used to select the correct output from the smaller model. Our findings reveal "},"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":"2404.00725","kind":"arxiv","version":2},"metadata":{"license":"http://creativecommons.org/publicdomain/zero/1.0/","primary_cat":"cs.SE","submitted_at":"2024-03-31T15:55:49Z","cross_cats_sorted":["cs.AI","cs.CL","cs.LG"],"title_canon_sha256":"755d64585390b318ebb550d36a808aeacadb4095792019ec091558435b13dcaf","abstract_canon_sha256":"1c3f1dbad7048ab04a20d545a439ca6469a00cc156d97e2799d3419524e1eec5"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T08:48:13.341832Z","signature_b64":"mwahFdoVoMV+84+G1hQRWKrkVlikrcPtQ6mzbTa45Ni9itlnYJaGAASQ7SJFjp6M2opmDtCx8TbquoNfsfpQAA==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"2dea882a03cd0ca84b831060a633cbbc93955e25e205f742c04a6fdede06d580","last_reissued_at":"2026-07-05T08:48:13.341328Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T08:48:13.341328Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"The Larger the Better? Improved LLM Code-Generation via Budget Reallocation","license":"http://creativecommons.org/publicdomain/zero/1.0/","headline":"","cross_cats":["cs.AI","cs.CL","cs.LG"],"primary_cat":"cs.SE","authors_text":"Jonas Gehring, Michael Hassid, Roy Schwartz, Tal Remez, Yossi Adi","submitted_at":"2024-03-31T15:55:49Z","abstract_excerpt":"It is a common belief that large language models (LLMs) are better than smaller-sized ones. However, larger models also require significantly more time and compute during inference. This begs the question: what happens when both models operate under the same budget? (e.g., compute, run-time). To address this question, we analyze code generation LLMs of various sizes and make comparisons such as running a 70B model once vs. generating five outputs from a 13B model. We consider a standard unit-test setup, which can be used to select the correct output from the smaller model. Our findings reveal "},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2404.00725","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/2404.00725/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":"2404.00725","created_at":"2026-07-05T08:48:13.341389+00:00"},{"alias_kind":"arxiv_version","alias_value":"2404.00725v2","created_at":"2026-07-05T08:48:13.341389+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2404.00725","created_at":"2026-07-05T08:48:13.341389+00:00"},{"alias_kind":"pith_short_12","alias_value":"FXVIQKQDZUGK","created_at":"2026-07-05T08:48:13.341389+00:00"},{"alias_kind":"pith_short_16","alias_value":"FXVIQKQDZUGKQS4D","created_at":"2026-07-05T08:48:13.341389+00:00"},{"alias_kind":"pith_short_8","alias_value":"FXVIQKQD","created_at":"2026-07-05T08:48:13.341389+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":4,"internal_anchor_count":0,"sample":[{"citing_arxiv_id":"2606.11755","citing_title":"Acoda: Adversarial Code Obfuscation for Defending against LLM-based Analysis","ref_index":13,"is_internal_anchor":false},{"citing_arxiv_id":"2407.21787","citing_title":"Large Language Monkeys: Scaling Inference Compute with Repeated Sampling","ref_index":26,"is_internal_anchor":false},{"citing_arxiv_id":"2604.14853","citing_title":"Adaptive Test-Time Compute Allocation for Reasoning LLMs via Constrained Policy Optimization","ref_index":37,"is_internal_anchor":false},{"citing_arxiv_id":"2604.17433","citing_title":"Self-Consistency from Only Two Samples: CoT-PoT Ensembling for Efficient LLM Reasoning","ref_index":49,"is_internal_anchor":false}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/FXVIQKQDZUGKQS4DCBQKMM6LXS","json":"https://pith.science/pith/FXVIQKQDZUGKQS4DCBQKMM6LXS.json","graph_json":"https://pith.science/api/pith-number/FXVIQKQDZUGKQS4DCBQKMM6LXS/graph.json","events_json":"https://pith.science/api/pith-number/FXVIQKQDZUGKQS4DCBQKMM6LXS/events.json","paper":"https://pith.science/paper/FXVIQKQD"},"agent_actions":{"view_html":"https://pith.science/pith/FXVIQKQDZUGKQS4DCBQKMM6LXS","download_json":"https://pith.science/pith/FXVIQKQDZUGKQS4DCBQKMM6LXS.json","view_paper":"https://pith.science/paper/FXVIQKQD","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2404.00725&json=true","fetch_graph":"https://pith.science/api/pith-number/FXVIQKQDZUGKQS4DCBQKMM6LXS/graph.json","fetch_events":"https://pith.science/api/pith-number/FXVIQKQDZUGKQS4DCBQKMM6LXS/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/FXVIQKQDZUGKQS4DCBQKMM6LXS/action/timestamp_anchor","attest_storage":"https://pith.science/pith/FXVIQKQDZUGKQS4DCBQKMM6LXS/action/storage_attestation","attest_author":"https://pith.science/pith/FXVIQKQDZUGKQS4DCBQKMM6LXS/action/author_attestation","sign_citation":"https://pith.science/pith/FXVIQKQDZUGKQS4DCBQKMM6LXS/action/citation_signature","submit_replication":"https://pith.science/pith/FXVIQKQDZUGKQS4DCBQKMM6LXS/action/replication_record"}},"created_at":"2026-07-05T08:48:13.341389+00:00","updated_at":"2026-07-05T08:48:13.341389+00:00"}