{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2024:DD52LTREYE4RDLAKFUDYM6453Z","short_pith_number":"pith:DD52LTRE","schema_version":"1.0","canonical_sha256":"18fba5ce24c13911ac0a2d07867b9dde7335a00ef1e94004e31797d8b52c640f","source":{"kind":"arxiv","id":"2407.12847","version":1},"attestation_state":"computed","paper":{"title":"Aligning Model Evaluations with Human Preferences: Mitigating Token Count Bias in Language Model Assessments","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.AI","cs.HC"],"primary_cat":"cs.CL","authors_text":"Jason Mars, Roland Daynauth","submitted_at":"2024-07-05T09:26:40Z","abstract_excerpt":"The SLAM paper demonstrated that on-device Small Language Models (SLMs) are a viable and cost-effective alternative to API-based Large Language Models (LLMs), such as OpenAI's GPT-4, offering comparable performance and stability. However, SLAM also identified discrepancies between human preferences and traditional auto-evaluators. This follow-up paper explores methods to align LLM evaluator preferences with human evaluations by addressing biases, particularly toward higher token counts. We employed Bayesian statistics and a t-test to quantify this bias and developed a recalibration procedure t"},"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":"2407.12847","kind":"arxiv","version":1},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.CL","submitted_at":"2024-07-05T09:26:40Z","cross_cats_sorted":["cs.AI","cs.HC"],"title_canon_sha256":"375c04ed5b44422a4a0a5b2ef183526687e50f9a9e76bba5146f1e039da0a927","abstract_canon_sha256":"bbff29aa5acf06fd22019f528fb902b9ae99bf7a2ef431c7859da68604110c46"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T08:45:12.478417Z","signature_b64":"5yUfWE4USczhA8ZCJUp00G02d32L1hBBSLvkxTI2Xu9Rc9CD5J/PjrQJEt56XfLz9GcSk7QHGnD3q5rpOHNuDg==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"18fba5ce24c13911ac0a2d07867b9dde7335a00ef1e94004e31797d8b52c640f","last_reissued_at":"2026-07-05T08:45:12.478013Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T08:45:12.478013Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Aligning Model Evaluations with Human Preferences: Mitigating Token Count Bias in Language Model Assessments","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.AI","cs.HC"],"primary_cat":"cs.CL","authors_text":"Jason Mars, Roland Daynauth","submitted_at":"2024-07-05T09:26:40Z","abstract_excerpt":"The SLAM paper demonstrated that on-device Small Language Models (SLMs) are a viable and cost-effective alternative to API-based Large Language Models (LLMs), such as OpenAI's GPT-4, offering comparable performance and stability. However, SLAM also identified discrepancies between human preferences and traditional auto-evaluators. This follow-up paper explores methods to align LLM evaluator preferences with human evaluations by addressing biases, particularly toward higher token counts. We employed Bayesian statistics and a t-test to quantify this bias and developed a recalibration procedure t"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2407.12847","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/2407.12847/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":"2407.12847","created_at":"2026-07-05T08:45:12.478067+00:00"},{"alias_kind":"arxiv_version","alias_value":"2407.12847v1","created_at":"2026-07-05T08:45:12.478067+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2407.12847","created_at":"2026-07-05T08:45:12.478067+00:00"},{"alias_kind":"pith_short_12","alias_value":"DD52LTREYE4R","created_at":"2026-07-05T08:45:12.478067+00:00"},{"alias_kind":"pith_short_16","alias_value":"DD52LTREYE4RDLAK","created_at":"2026-07-05T08:45:12.478067+00:00"},{"alias_kind":"pith_short_8","alias_value":"DD52LTRE","created_at":"2026-07-05T08:45:12.478067+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":2,"internal_anchor_count":0,"sample":[{"citing_arxiv_id":"2512.23213","citing_title":"Scoring, Reasoning, and Selecting the Best! Ensembling Large Language Models via a Peer-Review Process","ref_index":15,"is_internal_anchor":false},{"citing_arxiv_id":"2412.05579","citing_title":"LLMs-as-Judges: A Comprehensive Survey on LLM-based Evaluation Methods","ref_index":45,"is_internal_anchor":false}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/DD52LTREYE4RDLAKFUDYM6453Z","json":"https://pith.science/pith/DD52LTREYE4RDLAKFUDYM6453Z.json","graph_json":"https://pith.science/api/pith-number/DD52LTREYE4RDLAKFUDYM6453Z/graph.json","events_json":"https://pith.science/api/pith-number/DD52LTREYE4RDLAKFUDYM6453Z/events.json","paper":"https://pith.science/paper/DD52LTRE"},"agent_actions":{"view_html":"https://pith.science/pith/DD52LTREYE4RDLAKFUDYM6453Z","download_json":"https://pith.science/pith/DD52LTREYE4RDLAKFUDYM6453Z.json","view_paper":"https://pith.science/paper/DD52LTRE","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2407.12847&json=true","fetch_graph":"https://pith.science/api/pith-number/DD52LTREYE4RDLAKFUDYM6453Z/graph.json","fetch_events":"https://pith.science/api/pith-number/DD52LTREYE4RDLAKFUDYM6453Z/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/DD52LTREYE4RDLAKFUDYM6453Z/action/timestamp_anchor","attest_storage":"https://pith.science/pith/DD52LTREYE4RDLAKFUDYM6453Z/action/storage_attestation","attest_author":"https://pith.science/pith/DD52LTREYE4RDLAKFUDYM6453Z/action/author_attestation","sign_citation":"https://pith.science/pith/DD52LTREYE4RDLAKFUDYM6453Z/action/citation_signature","submit_replication":"https://pith.science/pith/DD52LTREYE4RDLAKFUDYM6453Z/action/replication_record"}},"created_at":"2026-07-05T08:45:12.478067+00:00","updated_at":"2026-07-05T08:45:12.478067+00:00"}