{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2025:YIB67CLKZ244IOZY3BEFTBW6CB","short_pith_number":"pith:YIB67CLK","schema_version":"1.0","canonical_sha256":"c203ef896aceb9c43b38d8485986de106bc467c7d2a7d7c8b0a271a8eefd631c","source":{"kind":"arxiv","id":"2502.01951","version":4},"attestation_state":"computed","paper":{"title":"On the Emergence of Position Bias in Transformers","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":[],"primary_cat":"cs.LG","authors_text":"Ali Jadbabaie, Stefanie Jegelka, Xinyi Wu, Yifei Wang","submitted_at":"2025-02-04T02:53:07Z","abstract_excerpt":"Recent studies have revealed various manifestations of position bias in transformer architectures, from the \"lost-in-the-middle\" phenomenon to attention sinks, yet a comprehensive theoretical understanding of how attention masks and positional encodings shape these biases remains elusive. This paper presents a graph-theoretic framework for analyzing position bias in multi-layer attention. Modeling attention masks as directed graphs, we quantify how tokens interact with contextual information based on their sequential positions. We uncover two key insights: First, causal masking inherently 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":"2502.01951","kind":"arxiv","version":4},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.LG","submitted_at":"2025-02-04T02:53:07Z","cross_cats_sorted":[],"title_canon_sha256":"c331e1a087be73cb94791c6ce7da0b589ee724948b13eaab1edb1107d09bbc35","abstract_canon_sha256":"4d2086c0c2026be889dc1e0e30fb3fdda19bc0b9ddd260e6f3fb9d0d9d7a6b4e"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T11:51:21.535890Z","signature_b64":"5ThUqe2OQrnAHkmXyDdsuBIlGJB6p0PJAglOImrP4k3gZco7s5wSakhkCgT+MzXEM1WY9YxuG4eydEN5BrMsAg==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"c203ef896aceb9c43b38d8485986de106bc467c7d2a7d7c8b0a271a8eefd631c","last_reissued_at":"2026-07-05T11:51:21.535362Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T11:51:21.535362Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"On the Emergence of Position Bias in Transformers","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":[],"primary_cat":"cs.LG","authors_text":"Ali Jadbabaie, Stefanie Jegelka, Xinyi Wu, Yifei Wang","submitted_at":"2025-02-04T02:53:07Z","abstract_excerpt":"Recent studies have revealed various manifestations of position bias in transformer architectures, from the \"lost-in-the-middle\" phenomenon to attention sinks, yet a comprehensive theoretical understanding of how attention masks and positional encodings shape these biases remains elusive. This paper presents a graph-theoretic framework for analyzing position bias in multi-layer attention. Modeling attention masks as directed graphs, we quantify how tokens interact with contextual information based on their sequential positions. We uncover two key insights: First, causal masking inherently bias"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2502.01951","kind":"arxiv","version":4},"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/2502.01951/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":"2502.01951","created_at":"2026-07-05T11:51:21.535423+00:00"},{"alias_kind":"arxiv_version","alias_value":"2502.01951v4","created_at":"2026-07-05T11:51:21.535423+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2502.01951","created_at":"2026-07-05T11:51:21.535423+00:00"},{"alias_kind":"pith_short_12","alias_value":"YIB67CLKZ244","created_at":"2026-07-05T11:51:21.535423+00:00"},{"alias_kind":"pith_short_16","alias_value":"YIB67CLKZ244IOZY","created_at":"2026-07-05T11:51:21.535423+00:00"},{"alias_kind":"pith_short_8","alias_value":"YIB67CLK","created_at":"2026-07-05T11:51:21.535423+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":9,"internal_anchor_count":0,"sample":[{"citing_arxiv_id":"2605.24279","citing_title":"ContextEcho: A Benchmark for Persona Drift in Long Agentic-Coding Sessions","ref_index":87,"is_internal_anchor":false},{"citing_arxiv_id":"2606.29139","citing_title":"How Token Influence Decays with Distance: A Green-Function View of Trained Language Models","ref_index":23,"is_internal_anchor":false},{"citing_arxiv_id":"2506.14092","citing_title":"Fragile Preferences: A Deep Dive Into Order Effects in Large Language Models","ref_index":40,"is_internal_anchor":false},{"citing_arxiv_id":"2508.15815","citing_title":"User-Assistant Bias in LLMs","ref_index":20,"is_internal_anchor":false},{"citing_arxiv_id":"2604.19777","citing_title":"Self-Describing Structured Data with Dual-Layer Guidance: A Lightweight Alternative to RAG for Precision Retrieval in Large-Scale LLM Knowledge Navigation","ref_index":6,"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":10,"is_internal_anchor":false},{"citing_arxiv_id":"2605.08539","citing_title":"Continuity Laws for Sequential Models","ref_index":36,"is_internal_anchor":false},{"citing_arxiv_id":"2605.10039","citing_title":"Instruction Adherence in Coding Agent Configuration Files: A Factorial Study of Four File-Structure Variables","ref_index":15,"is_internal_anchor":false},{"citing_arxiv_id":"2605.05696","citing_title":"Irminsul: MLA-Native Position-Independent Caching for Agentic LLM Serving","ref_index":30,"is_internal_anchor":false}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/YIB67CLKZ244IOZY3BEFTBW6CB","json":"https://pith.science/pith/YIB67CLKZ244IOZY3BEFTBW6CB.json","graph_json":"https://pith.science/api/pith-number/YIB67CLKZ244IOZY3BEFTBW6CB/graph.json","events_json":"https://pith.science/api/pith-number/YIB67CLKZ244IOZY3BEFTBW6CB/events.json","paper":"https://pith.science/paper/YIB67CLK"},"agent_actions":{"view_html":"https://pith.science/pith/YIB67CLKZ244IOZY3BEFTBW6CB","download_json":"https://pith.science/pith/YIB67CLKZ244IOZY3BEFTBW6CB.json","view_paper":"https://pith.science/paper/YIB67CLK","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2502.01951&json=true","fetch_graph":"https://pith.science/api/pith-number/YIB67CLKZ244IOZY3BEFTBW6CB/graph.json","fetch_events":"https://pith.science/api/pith-number/YIB67CLKZ244IOZY3BEFTBW6CB/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/YIB67CLKZ244IOZY3BEFTBW6CB/action/timestamp_anchor","attest_storage":"https://pith.science/pith/YIB67CLKZ244IOZY3BEFTBW6CB/action/storage_attestation","attest_author":"https://pith.science/pith/YIB67CLKZ244IOZY3BEFTBW6CB/action/author_attestation","sign_citation":"https://pith.science/pith/YIB67CLKZ244IOZY3BEFTBW6CB/action/citation_signature","submit_replication":"https://pith.science/pith/YIB67CLKZ244IOZY3BEFTBW6CB/action/replication_record"}},"created_at":"2026-07-05T11:51:21.535423+00:00","updated_at":"2026-07-05T11:51:21.535423+00:00"}