{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2024:QJIWELKM5ZWEVAECOVRKETVLOC","short_pith_number":"pith:QJIWELKM","schema_version":"1.0","canonical_sha256":"8251622d4cee6c4a80827562a24eab70aa139a0ec46137e7954678817851c034","source":{"kind":"arxiv","id":"2407.10457","version":1},"attestation_state":"computed","paper":{"title":"The Good, The Bad, and The Greedy: Evaluation of LLMs Should Not Ignore Non-Determinism","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.AI"],"primary_cat":"cs.CL","authors_text":"Bill Yuchen Lin, Guoyin Wang, Sujian Li, Yifan Song","submitted_at":"2024-07-15T06:12:17Z","abstract_excerpt":"Current evaluations of large language models (LLMs) often overlook non-determinism, typically focusing on a single output per example. This limits our understanding of LLM performance variability in real-world applications. Our study addresses this issue by exploring key questions about the performance differences between greedy decoding and sampling, identifying benchmarks' consistency regarding non-determinism, and examining unique model behaviors. Through extensive experiments, we observe that greedy decoding generally outperforms sampling methods for most evaluated tasks. We also observe c"},"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":"2407.10457","kind":"arxiv","version":1},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.CL","submitted_at":"2024-07-15T06:12:17Z","cross_cats_sorted":["cs.AI"],"title_canon_sha256":"e8241db3ddd52ad5ee88d9a3afc4ef892a788372e14b1ffeb8188882ea9e6a49","abstract_canon_sha256":"0ae8a01668ac6da75d58be818805384855d0af4b067e0e756cbc76d3fe39e17b"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T08:43:59.812238Z","signature_b64":"9Db7VEfmFn7zZGxdPOUL8tu2q+cYxiXVYUjDZc7p/xAtSAEGFLK2NHaAbbD3RbL8LFon1FHQkVNz0rjmsCBsCQ==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"8251622d4cee6c4a80827562a24eab70aa139a0ec46137e7954678817851c034","last_reissued_at":"2026-07-05T08:43:59.811741Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T08:43:59.811741Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"The Good, The Bad, and The Greedy: Evaluation of LLMs Should Not Ignore Non-Determinism","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.AI"],"primary_cat":"cs.CL","authors_text":"Bill Yuchen Lin, Guoyin Wang, Sujian Li, Yifan Song","submitted_at":"2024-07-15T06:12:17Z","abstract_excerpt":"Current evaluations of large language models (LLMs) often overlook non-determinism, typically focusing on a single output per example. This limits our understanding of LLM performance variability in real-world applications. Our study addresses this issue by exploring key questions about the performance differences between greedy decoding and sampling, identifying benchmarks' consistency regarding non-determinism, and examining unique model behaviors. Through extensive experiments, we observe that greedy decoding generally outperforms sampling methods for most evaluated tasks. We also observe c"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2407.10457","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/2407.10457/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":"2407.10457","created_at":"2026-07-05T08:43:59.811808+00:00"},{"alias_kind":"arxiv_version","alias_value":"2407.10457v1","created_at":"2026-07-05T08:43:59.811808+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2407.10457","created_at":"2026-07-05T08:43:59.811808+00:00"},{"alias_kind":"pith_short_12","alias_value":"QJIWELKM5ZWE","created_at":"2026-07-05T08:43:59.811808+00:00"},{"alias_kind":"pith_short_16","alias_value":"QJIWELKM5ZWEVAEC","created_at":"2026-07-05T08:43:59.811808+00:00"},{"alias_kind":"pith_short_8","alias_value":"QJIWELKM","created_at":"2026-07-05T08:43:59.811808+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":5,"internal_anchor_count":0,"sample":[{"citing_arxiv_id":"2606.17182","citing_title":"Verified Detection and Prevention of Concurrency Anomalies in Multi-Agent Large Language Model Systems","ref_index":98,"is_internal_anchor":false},{"citing_arxiv_id":"2407.21787","citing_title":"Large Language Monkeys: Scaling Inference Compute with Repeated Sampling","ref_index":54,"is_internal_anchor":false},{"citing_arxiv_id":"2604.22411","citing_title":"Introducing Background Temperature to Characterise Hidden Randomness in Large Language Models","ref_index":11,"is_internal_anchor":false},{"citing_arxiv_id":"2604.19598","citing_title":"Cross-Model Consistency of AI-Generated Exercise Prescriptions: A Repeated Generation Study Across Three Large Language Models","ref_index":42,"is_internal_anchor":false},{"citing_arxiv_id":"2604.10350","citing_title":"LLM-based Generation of Semantically Diverse and Realistic Domain Model Instances","ref_index":27,"is_internal_anchor":false}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/QJIWELKM5ZWEVAECOVRKETVLOC","json":"https://pith.science/pith/QJIWELKM5ZWEVAECOVRKETVLOC.json","graph_json":"https://pith.science/api/pith-number/QJIWELKM5ZWEVAECOVRKETVLOC/graph.json","events_json":"https://pith.science/api/pith-number/QJIWELKM5ZWEVAECOVRKETVLOC/events.json","paper":"https://pith.science/paper/QJIWELKM"},"agent_actions":{"view_html":"https://pith.science/pith/QJIWELKM5ZWEVAECOVRKETVLOC","download_json":"https://pith.science/pith/QJIWELKM5ZWEVAECOVRKETVLOC.json","view_paper":"https://pith.science/paper/QJIWELKM","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2407.10457&json=true","fetch_graph":"https://pith.science/api/pith-number/QJIWELKM5ZWEVAECOVRKETVLOC/graph.json","fetch_events":"https://pith.science/api/pith-number/QJIWELKM5ZWEVAECOVRKETVLOC/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/QJIWELKM5ZWEVAECOVRKETVLOC/action/timestamp_anchor","attest_storage":"https://pith.science/pith/QJIWELKM5ZWEVAECOVRKETVLOC/action/storage_attestation","attest_author":"https://pith.science/pith/QJIWELKM5ZWEVAECOVRKETVLOC/action/author_attestation","sign_citation":"https://pith.science/pith/QJIWELKM5ZWEVAECOVRKETVLOC/action/citation_signature","submit_replication":"https://pith.science/pith/QJIWELKM5ZWEVAECOVRKETVLOC/action/replication_record"}},"created_at":"2026-07-05T08:43:59.811808+00:00","updated_at":"2026-07-05T08:43:59.811808+00:00"}