{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2026:CWCWQBLGQ35PY6ZMOBXHWWKPCU","short_pith_number":"pith:CWCWQBLG","schema_version":"1.0","canonical_sha256":"158568056686fafc7b2c706e7b594f150035af552dd2cf2819ad0438862dbdd7","source":{"kind":"arxiv","id":"2602.16837","version":2},"attestation_state":"computed","paper":{"title":"A Structural Theory of Position Bias in Transformers","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":[],"primary_cat":"cs.LG","authors_text":"Hanna Herasimchyk, Robin Labryga, S\\\"oren Laue, Tomislav Prusina","submitted_at":"2026-02-18T20:01:39Z","abstract_excerpt":"Transformer models systematically favor certain token positions, yet the architectural origins of this position bias remain poorly understood. This bias is closely connected to the Lost-in-the-Middle phenomenon, where models underutilize information placed in the middle of the context. We show that Lost-in-the-Middle-type behavior can arise from the architecture of causal Transformers itself. To do so, we develop a structural theory of position bias based on residual-aware cumulative attention rollout. At finite depth, causal masking and residual connections induce broad, often U-shaped, influ"},"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":"2602.16837","kind":"arxiv","version":2},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.LG","submitted_at":"2026-02-18T20:01:39Z","cross_cats_sorted":[],"title_canon_sha256":"6efb6bc7bf34ccc83402ea3159d42ba405501760491f0e478ff10b929aafc20e","abstract_canon_sha256":"63b944dedfbbe348d7ac50c7d22dbf0300f67bccc31ac36dfabff9e394f5f12a"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-05-28T01:04:37.157147Z","signature_b64":"D5VppMZp2gxhkZV177keZmpLE7JOLRuJa/Plt907IHEHd6to0iqqhnvKJzo3ll+cdtgYxyIc88l2EluyK1t3CQ==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"158568056686fafc7b2c706e7b594f150035af552dd2cf2819ad0438862dbdd7","last_reissued_at":"2026-05-28T01:04:37.156644Z","signature_status":"signed_v1","first_computed_at":"2026-05-28T01:04:37.156644Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"A Structural Theory of Position Bias in Transformers","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":[],"primary_cat":"cs.LG","authors_text":"Hanna Herasimchyk, Robin Labryga, S\\\"oren Laue, Tomislav Prusina","submitted_at":"2026-02-18T20:01:39Z","abstract_excerpt":"Transformer models systematically favor certain token positions, yet the architectural origins of this position bias remain poorly understood. This bias is closely connected to the Lost-in-the-Middle phenomenon, where models underutilize information placed in the middle of the context. We show that Lost-in-the-Middle-type behavior can arise from the architecture of causal Transformers itself. To do so, we develop a structural theory of position bias based on residual-aware cumulative attention rollout. At finite depth, causal masking and residual connections induce broad, often U-shaped, influ"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2602.16837","kind":"arxiv","version":2},"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/2602.16837/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":"2602.16837","created_at":"2026-05-28T01:04:37.156705+00:00"},{"alias_kind":"arxiv_version","alias_value":"2602.16837v2","created_at":"2026-05-28T01:04:37.156705+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2602.16837","created_at":"2026-05-28T01:04:37.156705+00:00"},{"alias_kind":"pith_short_12","alias_value":"CWCWQBLGQ35P","created_at":"2026-05-28T01:04:37.156705+00:00"},{"alias_kind":"pith_short_16","alias_value":"CWCWQBLGQ35PY6ZM","created_at":"2026-05-28T01:04:37.156705+00:00"},{"alias_kind":"pith_short_8","alias_value":"CWCWQBLG","created_at":"2026-05-28T01:04:37.156705+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":1,"internal_anchor_count":1,"sample":[{"citing_arxiv_id":"2605.09213","citing_title":"Kinetic theory for Transformers and the lost-in-the-middle phenomenon","ref_index":22,"is_internal_anchor":true}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/CWCWQBLGQ35PY6ZMOBXHWWKPCU","json":"https://pith.science/pith/CWCWQBLGQ35PY6ZMOBXHWWKPCU.json","graph_json":"https://pith.science/api/pith-number/CWCWQBLGQ35PY6ZMOBXHWWKPCU/graph.json","events_json":"https://pith.science/api/pith-number/CWCWQBLGQ35PY6ZMOBXHWWKPCU/events.json","paper":"https://pith.science/paper/CWCWQBLG"},"agent_actions":{"view_html":"https://pith.science/pith/CWCWQBLGQ35PY6ZMOBXHWWKPCU","download_json":"https://pith.science/pith/CWCWQBLGQ35PY6ZMOBXHWWKPCU.json","view_paper":"https://pith.science/paper/CWCWQBLG","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2602.16837&json=true","fetch_graph":"https://pith.science/api/pith-number/CWCWQBLGQ35PY6ZMOBXHWWKPCU/graph.json","fetch_events":"https://pith.science/api/pith-number/CWCWQBLGQ35PY6ZMOBXHWWKPCU/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/CWCWQBLGQ35PY6ZMOBXHWWKPCU/action/timestamp_anchor","attest_storage":"https://pith.science/pith/CWCWQBLGQ35PY6ZMOBXHWWKPCU/action/storage_attestation","attest_author":"https://pith.science/pith/CWCWQBLGQ35PY6ZMOBXHWWKPCU/action/author_attestation","sign_citation":"https://pith.science/pith/CWCWQBLGQ35PY6ZMOBXHWWKPCU/action/citation_signature","submit_replication":"https://pith.science/pith/CWCWQBLGQ35PY6ZMOBXHWWKPCU/action/replication_record"}},"created_at":"2026-05-28T01:04:37.156705+00:00","updated_at":"2026-05-28T01:04:37.156705+00:00"}