{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2025:AGL3PBJGVDCMQBV2OMW2SL3RPN","short_pith_number":"pith:AGL3PBJG","schema_version":"1.0","canonical_sha256":"0197b78526a8c4c806ba732da92f717b6588ca312f97cd810ff4d503fc66696a","source":{"kind":"arxiv","id":"2508.02020","version":1},"attestation_state":"computed","paper":{"title":"Evaluating Position Bias in Large Language Model Recommendations","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":[],"primary_cat":"cs.IR","authors_text":"Estrid He, Ethan Bito, Yongli Ren","submitted_at":"2025-08-04T03:30:26Z","abstract_excerpt":"Large Language Models (LLMs) are being increasingly explored as general-purpose tools for recommendation tasks, enabling zero-shot and instruction-following capabilities without the need for task-specific training. While the research community is enthusiastically embracing LLMs, there are important caveats to directly adapting them for recommendation tasks. In this paper, we show that LLM-based recommendation models suffer from position bias, where the order of candidate items in a prompt can disproportionately influence the recommendations produced by LLMs. First, we analyse the position bias"},"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":"2508.02020","kind":"arxiv","version":1},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.IR","submitted_at":"2025-08-04T03:30:26Z","cross_cats_sorted":[],"title_canon_sha256":"29d2d88dae8e0d8582fa8cbd944251f9b9bcf15974ca0789de6833240bfe4526","abstract_canon_sha256":"942751ba61e9bf6e9865a338e50dd972c419f51310dea9556edc524a81bfd398"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T11:47:49.687002Z","signature_b64":"VF3hBTXPUOn7t8B47bwHBKSTfgWdFLUQeiHWg5dBLW6QUIdJ4qpsX+d5//BIFs5TvXnrSqSgVnw8r11DfFIGDA==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"0197b78526a8c4c806ba732da92f717b6588ca312f97cd810ff4d503fc66696a","last_reissued_at":"2026-07-05T11:47:49.686482Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T11:47:49.686482Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Evaluating Position Bias in Large Language Model Recommendations","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":[],"primary_cat":"cs.IR","authors_text":"Estrid He, Ethan Bito, Yongli Ren","submitted_at":"2025-08-04T03:30:26Z","abstract_excerpt":"Large Language Models (LLMs) are being increasingly explored as general-purpose tools for recommendation tasks, enabling zero-shot and instruction-following capabilities without the need for task-specific training. While the research community is enthusiastically embracing LLMs, there are important caveats to directly adapting them for recommendation tasks. In this paper, we show that LLM-based recommendation models suffer from position bias, where the order of candidate items in a prompt can disproportionately influence the recommendations produced by LLMs. First, we analyse the position bias"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2508.02020","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/2508.02020/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":"2508.02020","created_at":"2026-07-05T11:47:49.686537+00:00"},{"alias_kind":"arxiv_version","alias_value":"2508.02020v1","created_at":"2026-07-05T11:47:49.686537+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2508.02020","created_at":"2026-07-05T11:47:49.686537+00:00"},{"alias_kind":"pith_short_12","alias_value":"AGL3PBJGVDCM","created_at":"2026-07-05T11:47:49.686537+00:00"},{"alias_kind":"pith_short_16","alias_value":"AGL3PBJGVDCMQBV2","created_at":"2026-07-05T11:47:49.686537+00:00"},{"alias_kind":"pith_short_8","alias_value":"AGL3PBJG","created_at":"2026-07-05T11:47:49.686537+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":3,"internal_anchor_count":0,"sample":[{"citing_arxiv_id":"2606.19710","citing_title":"FineREX: Fine-Tuned NER-RE for Human Smuggling Knowledge Graphs","ref_index":7,"is_internal_anchor":false},{"citing_arxiv_id":"2606.00540","citing_title":"Trustworthy Recommendation in the Era of Large Language Models: Opportunities and Challenges","ref_index":15,"is_internal_anchor":false},{"citing_arxiv_id":"2604.27599","citing_title":"One Pass, Any Order: Position-Invariant Listwise Reranking for LLM-Based Recommendation","ref_index":1,"is_internal_anchor":false}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/AGL3PBJGVDCMQBV2OMW2SL3RPN","json":"https://pith.science/pith/AGL3PBJGVDCMQBV2OMW2SL3RPN.json","graph_json":"https://pith.science/api/pith-number/AGL3PBJGVDCMQBV2OMW2SL3RPN/graph.json","events_json":"https://pith.science/api/pith-number/AGL3PBJGVDCMQBV2OMW2SL3RPN/events.json","paper":"https://pith.science/paper/AGL3PBJG"},"agent_actions":{"view_html":"https://pith.science/pith/AGL3PBJGVDCMQBV2OMW2SL3RPN","download_json":"https://pith.science/pith/AGL3PBJGVDCMQBV2OMW2SL3RPN.json","view_paper":"https://pith.science/paper/AGL3PBJG","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2508.02020&json=true","fetch_graph":"https://pith.science/api/pith-number/AGL3PBJGVDCMQBV2OMW2SL3RPN/graph.json","fetch_events":"https://pith.science/api/pith-number/AGL3PBJGVDCMQBV2OMW2SL3RPN/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/AGL3PBJGVDCMQBV2OMW2SL3RPN/action/timestamp_anchor","attest_storage":"https://pith.science/pith/AGL3PBJGVDCMQBV2OMW2SL3RPN/action/storage_attestation","attest_author":"https://pith.science/pith/AGL3PBJGVDCMQBV2OMW2SL3RPN/action/author_attestation","sign_citation":"https://pith.science/pith/AGL3PBJGVDCMQBV2OMW2SL3RPN/action/citation_signature","submit_replication":"https://pith.science/pith/AGL3PBJGVDCMQBV2OMW2SL3RPN/action/replication_record"}},"created_at":"2026-07-05T11:47:49.686537+00:00","updated_at":"2026-07-05T11:47:49.686537+00:00"}