{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2023:I6CK3IOAHCOGDUV6IRGKGHBEV2","short_pith_number":"pith:I6CK3IOA","schema_version":"1.0","canonical_sha256":"4784ada1c0389c61d2be444ca31c24ae90659b5cfd147cccb44bb6eacd475490","source":{"kind":"arxiv","id":"2310.01427","version":1},"attestation_state":"computed","paper":{"title":"Attention Sorting Combats Recency Bias In Long Context Language Models","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.AI"],"primary_cat":"cs.CL","authors_text":"Adam Lerer, Alexander Peysakhovich","submitted_at":"2023-09-28T05:19:06Z","abstract_excerpt":"Current language models often fail to incorporate long contexts efficiently during generation. We show that a major contributor to this issue are attention priors that are likely learned during pre-training: relevant information located earlier in context is attended to less on average. Yet even when models fail to use the information from a relevant document in their response, they still pay preferential attention to that document compared to an irrelevant document at the same position. We leverage this fact to introduce ``attention sorting'': perform one step of decoding, sort documents by 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":"2310.01427","kind":"arxiv","version":1},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.CL","submitted_at":"2023-09-28T05:19:06Z","cross_cats_sorted":["cs.AI"],"title_canon_sha256":"c23e7d0c49373272714d9715dc514f23de1c9c3293077b829af2f9358fd9f6e8","abstract_canon_sha256":"5d8cc434fd565fee02dc40086d4c575be1cdaa83a83bdce2417425c31eabcad1"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T06:56:40.103575Z","signature_b64":"Gv0pGrQNEz5nJfGGrqx4EmWtC1Z8urTA3155TAJ0nMRh1q1ILZKZHCOH98ITHcSLx2vZV3efJA93jmej7pvKBw==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"4784ada1c0389c61d2be444ca31c24ae90659b5cfd147cccb44bb6eacd475490","last_reissued_at":"2026-07-05T06:56:40.103030Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T06:56:40.103030Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Attention Sorting Combats Recency Bias In Long Context Language Models","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.AI"],"primary_cat":"cs.CL","authors_text":"Adam Lerer, Alexander Peysakhovich","submitted_at":"2023-09-28T05:19:06Z","abstract_excerpt":"Current language models often fail to incorporate long contexts efficiently during generation. We show that a major contributor to this issue are attention priors that are likely learned during pre-training: relevant information located earlier in context is attended to less on average. Yet even when models fail to use the information from a relevant document in their response, they still pay preferential attention to that document compared to an irrelevant document at the same position. We leverage this fact to introduce ``attention sorting'': perform one step of decoding, sort documents by t"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2310.01427","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/2310.01427/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":"2310.01427","created_at":"2026-07-05T06:56:40.103113+00:00"},{"alias_kind":"arxiv_version","alias_value":"2310.01427v1","created_at":"2026-07-05T06:56:40.103113+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2310.01427","created_at":"2026-07-05T06:56:40.103113+00:00"},{"alias_kind":"pith_short_12","alias_value":"I6CK3IOAHCOG","created_at":"2026-07-05T06:56:40.103113+00:00"},{"alias_kind":"pith_short_16","alias_value":"I6CK3IOAHCOGDUV6","created_at":"2026-07-05T06:56:40.103113+00:00"},{"alias_kind":"pith_short_8","alias_value":"I6CK3IOA","created_at":"2026-07-05T06:56:40.103113+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":13,"internal_anchor_count":0,"sample":[{"citing_arxiv_id":"2606.31522","citing_title":"FinPersona-Bench: A Benchmark for Longitudinal Psychometric Stability of Autonomous Financial Agents","ref_index":53,"is_internal_anchor":false},{"citing_arxiv_id":"2606.31522","citing_title":"FinPersona-Bench: A Benchmark for Longitudinal Psychometric Stability of Autonomous Financial Agents","ref_index":53,"is_internal_anchor":false},{"citing_arxiv_id":"2606.28327","citing_title":"The Interference Gap: Comparing Retrieval Bounds in Human Memory and RAG Systems","ref_index":21,"is_internal_anchor":false},{"citing_arxiv_id":"2606.27705","citing_title":"Mitigating Position Bias in Transformers via Layer-Specific Positional Embedding Scaling","ref_index":23,"is_internal_anchor":false},{"citing_arxiv_id":"2506.04390","citing_title":"Through the Stealth Lens: Attention-Aware Defenses Against Poisoning in RAG","ref_index":38,"is_internal_anchor":false},{"citing_arxiv_id":"2604.10027","citing_title":"SinkTrack: Attention Sink based Context Anchoring for Large Language Models","ref_index":10,"is_internal_anchor":false},{"citing_arxiv_id":"2605.00505","citing_title":"LLM-Oriented Information Retrieval: A Denoising-First Perspective","ref_index":151,"is_internal_anchor":false},{"citing_arxiv_id":"2605.11161","citing_title":"Interpretability Can Be Actionable","ref_index":161,"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":4,"is_internal_anchor":false},{"citing_arxiv_id":"2605.08295","citing_title":"In-Context Fixation: When Demonstrated Labels Override Semantics in Few-Shot Classification","ref_index":34,"is_internal_anchor":false},{"citing_arxiv_id":"2605.00505","citing_title":"LLM-Oriented Information Retrieval: A Denoising-First Perspective","ref_index":148,"is_internal_anchor":false},{"citing_arxiv_id":"2604.10027","citing_title":"SinkTrack: Attention Sink based Context Anchoring for Large Language Models","ref_index":10,"is_internal_anchor":false},{"citing_arxiv_id":"2604.17237","citing_title":"HeadRank: Decoding-Free Passage Reranking via Preference-Aligned Attention Heads","ref_index":47,"is_internal_anchor":false}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/I6CK3IOAHCOGDUV6IRGKGHBEV2","json":"https://pith.science/pith/I6CK3IOAHCOGDUV6IRGKGHBEV2.json","graph_json":"https://pith.science/api/pith-number/I6CK3IOAHCOGDUV6IRGKGHBEV2/graph.json","events_json":"https://pith.science/api/pith-number/I6CK3IOAHCOGDUV6IRGKGHBEV2/events.json","paper":"https://pith.science/paper/I6CK3IOA"},"agent_actions":{"view_html":"https://pith.science/pith/I6CK3IOAHCOGDUV6IRGKGHBEV2","download_json":"https://pith.science/pith/I6CK3IOAHCOGDUV6IRGKGHBEV2.json","view_paper":"https://pith.science/paper/I6CK3IOA","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2310.01427&json=true","fetch_graph":"https://pith.science/api/pith-number/I6CK3IOAHCOGDUV6IRGKGHBEV2/graph.json","fetch_events":"https://pith.science/api/pith-number/I6CK3IOAHCOGDUV6IRGKGHBEV2/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/I6CK3IOAHCOGDUV6IRGKGHBEV2/action/timestamp_anchor","attest_storage":"https://pith.science/pith/I6CK3IOAHCOGDUV6IRGKGHBEV2/action/storage_attestation","attest_author":"https://pith.science/pith/I6CK3IOAHCOGDUV6IRGKGHBEV2/action/author_attestation","sign_citation":"https://pith.science/pith/I6CK3IOAHCOGDUV6IRGKGHBEV2/action/citation_signature","submit_replication":"https://pith.science/pith/I6CK3IOAHCOGDUV6IRGKGHBEV2/action/replication_record"}},"created_at":"2026-07-05T06:56:40.103113+00:00","updated_at":"2026-07-05T06:56:40.103113+00:00"}