{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2024:G7RWSXJRQI7P64LYE5OPMUVALA","short_pith_number":"pith:G7RWSXJR","schema_version":"1.0","canonical_sha256":"37e3695d31823eff7178275cf652a05819a8d4ad7976a1d581f91d05bd38b9f3","source":{"kind":"arxiv","id":"2406.02536","version":3},"attestation_state":"computed","paper":{"title":"Mitigate Position Bias in Large Language Models via Scaling a Single Dimension","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":["cs.LG"],"primary_cat":"cs.CL","authors_text":"Chin-Yew Lin, Dongsheng Li, Huiqiang Jiang, Lili Qiu, Qianhui Wu, Xufang Luo, Yijiong Yu, Yongfeng Huang, Yuqing Yang","submitted_at":"2024-06-04T17:55:38Z","abstract_excerpt":"Large Language Models (LLMs) are increasingly applied in various real-world scenarios due to their excellent generalization capabilities and robust generative abilities. However, they exhibit position bias, also known as \"lost in the middle\", a phenomenon that is especially pronounced in long-context scenarios, which indicates the placement of the key information in different positions of a prompt can significantly affect accuracy. This paper first explores the micro-level manifestations of position bias, concluding that attention weights are a micro-level expression of position bias. It furth"},"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":"2406.02536","kind":"arxiv","version":3},"metadata":{"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.CL","submitted_at":"2024-06-04T17:55:38Z","cross_cats_sorted":["cs.LG"],"title_canon_sha256":"9e35969d4ecedad250c957aa824fc947ad80586cc7caa09877cdc31dc5778343","abstract_canon_sha256":"0eb6541c20aa802495996a2e4eec59c7942dfdd5666e905e43b9dba31b06ae27"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T11:07:57.836034Z","signature_b64":"G4IvbEDNCZ/NnINJjv0Q9GgflErhxmzzsIHhruu63sBDqfSigBSsANS4pVOLq4b+LcHeDfZuC7lKexM1Q1AeDg==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"37e3695d31823eff7178275cf652a05819a8d4ad7976a1d581f91d05bd38b9f3","last_reissued_at":"2026-07-05T11:07:57.835453Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T11:07:57.835453Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Mitigate Position Bias in Large Language Models via Scaling a Single Dimension","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":["cs.LG"],"primary_cat":"cs.CL","authors_text":"Chin-Yew Lin, Dongsheng Li, Huiqiang Jiang, Lili Qiu, Qianhui Wu, Xufang Luo, Yijiong Yu, Yongfeng Huang, Yuqing Yang","submitted_at":"2024-06-04T17:55:38Z","abstract_excerpt":"Large Language Models (LLMs) are increasingly applied in various real-world scenarios due to their excellent generalization capabilities and robust generative abilities. However, they exhibit position bias, also known as \"lost in the middle\", a phenomenon that is especially pronounced in long-context scenarios, which indicates the placement of the key information in different positions of a prompt can significantly affect accuracy. This paper first explores the micro-level manifestations of position bias, concluding that attention weights are a micro-level expression of position bias. It furth"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2406.02536","kind":"arxiv","version":3},"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/2406.02536/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":"2406.02536","created_at":"2026-07-05T11:07:57.835522+00:00"},{"alias_kind":"arxiv_version","alias_value":"2406.02536v3","created_at":"2026-07-05T11:07:57.835522+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2406.02536","created_at":"2026-07-05T11:07:57.835522+00:00"},{"alias_kind":"pith_short_12","alias_value":"G7RWSXJRQI7P","created_at":"2026-07-05T11:07:57.835522+00:00"},{"alias_kind":"pith_short_16","alias_value":"G7RWSXJRQI7P64LY","created_at":"2026-07-05T11:07:57.835522+00:00"},{"alias_kind":"pith_short_8","alias_value":"G7RWSXJR","created_at":"2026-07-05T11:07:57.835522+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":3,"internal_anchor_count":0,"sample":[{"citing_arxiv_id":"2605.11974","citing_title":"Towards Order Fairness: Mitigating LLMs Order Sensitivity through Dual Group Advantage Optimization","ref_index":25,"is_internal_anchor":false},{"citing_arxiv_id":"2605.11591","citing_title":"Logit-Attention Divergence: Mitigating Position Bias in Multi-Image Retrieval via Attention-Guided Calibration","ref_index":13,"is_internal_anchor":false},{"citing_arxiv_id":"2604.04500","citing_title":"Saliency-R1: Enforcing Interpretable and Faithful Vision-language Reasoning via Saliency-map Alignment Reward","ref_index":85,"is_internal_anchor":false}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/G7RWSXJRQI7P64LYE5OPMUVALA","json":"https://pith.science/pith/G7RWSXJRQI7P64LYE5OPMUVALA.json","graph_json":"https://pith.science/api/pith-number/G7RWSXJRQI7P64LYE5OPMUVALA/graph.json","events_json":"https://pith.science/api/pith-number/G7RWSXJRQI7P64LYE5OPMUVALA/events.json","paper":"https://pith.science/paper/G7RWSXJR"},"agent_actions":{"view_html":"https://pith.science/pith/G7RWSXJRQI7P64LYE5OPMUVALA","download_json":"https://pith.science/pith/G7RWSXJRQI7P64LYE5OPMUVALA.json","view_paper":"https://pith.science/paper/G7RWSXJR","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2406.02536&json=true","fetch_graph":"https://pith.science/api/pith-number/G7RWSXJRQI7P64LYE5OPMUVALA/graph.json","fetch_events":"https://pith.science/api/pith-number/G7RWSXJRQI7P64LYE5OPMUVALA/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/G7RWSXJRQI7P64LYE5OPMUVALA/action/timestamp_anchor","attest_storage":"https://pith.science/pith/G7RWSXJRQI7P64LYE5OPMUVALA/action/storage_attestation","attest_author":"https://pith.science/pith/G7RWSXJRQI7P64LYE5OPMUVALA/action/author_attestation","sign_citation":"https://pith.science/pith/G7RWSXJRQI7P64LYE5OPMUVALA/action/citation_signature","submit_replication":"https://pith.science/pith/G7RWSXJRQI7P64LYE5OPMUVALA/action/replication_record"}},"created_at":"2026-07-05T11:07:57.835522+00:00","updated_at":"2026-07-05T11:07:57.835522+00:00"}