{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2024:JJPT672SVMDG5HN3CS5ZVZ72KF","short_pith_number":"pith:JJPT672S","schema_version":"1.0","canonical_sha256":"4a5f3f7f52ab066e9dbb14bb9ae7fa516d6d860c0eeb515aa7e1db579f6acbf5","source":{"kind":"arxiv","id":"2405.18719","version":2},"attestation_state":"computed","paper":{"title":"Contextual Position Encoding: Learning to Count What's Important","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":["cs.AI"],"primary_cat":"cs.CL","authors_text":"Jason Weston, Olga Golovneva, Sainbayar Sukhbaatar, Tianlu Wang","submitted_at":"2024-05-29T02:57:15Z","abstract_excerpt":"The attention mechanism is a critical component of Large Language Models (LLMs) that allows tokens in a sequence to interact with each other, but is order-invariant. Incorporating position encoding (PE) makes it possible to address by position, such as attending to the i-th token. However, current PE methods use token counts to derive position, and thus cannot generalize to higher levels of abstraction, such as attending to the i-th sentence. In this paper, we propose a new position encoding method, Contextual Position Encoding (CoPE), that allows positions to be conditioned on context by incr"},"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":"2405.18719","kind":"arxiv","version":2},"metadata":{"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.CL","submitted_at":"2024-05-29T02:57:15Z","cross_cats_sorted":["cs.AI"],"title_canon_sha256":"dfefcaf25d8a54263c5c180b6930df162a022cc025fa59e307e94f319893a7db","abstract_canon_sha256":"5cdcd0f5fb463c0b94cc334a43e4defb926f7d522069c7511e57096004cae2ad"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T08:25:06.111199Z","signature_b64":"BCIpcdYwGUVHmPiwsbcN+WGLvfhpi4Dh0khjv6Nv/J/Hfonu17WM1vetGgRE//+6xYyl+WOB2bu0GuwEPkbMBA==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"4a5f3f7f52ab066e9dbb14bb9ae7fa516d6d860c0eeb515aa7e1db579f6acbf5","last_reissued_at":"2026-07-05T08:25:06.110718Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T08:25:06.110718Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Contextual Position Encoding: Learning to Count What's Important","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":["cs.AI"],"primary_cat":"cs.CL","authors_text":"Jason Weston, Olga Golovneva, Sainbayar Sukhbaatar, Tianlu Wang","submitted_at":"2024-05-29T02:57:15Z","abstract_excerpt":"The attention mechanism is a critical component of Large Language Models (LLMs) that allows tokens in a sequence to interact with each other, but is order-invariant. Incorporating position encoding (PE) makes it possible to address by position, such as attending to the i-th token. However, current PE methods use token counts to derive position, and thus cannot generalize to higher levels of abstraction, such as attending to the i-th sentence. In this paper, we propose a new position encoding method, Contextual Position Encoding (CoPE), that allows positions to be conditioned on context by incr"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2405.18719","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/2405.18719/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":"2405.18719","created_at":"2026-07-05T08:25:06.110778+00:00"},{"alias_kind":"arxiv_version","alias_value":"2405.18719v2","created_at":"2026-07-05T08:25:06.110778+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2405.18719","created_at":"2026-07-05T08:25:06.110778+00:00"},{"alias_kind":"pith_short_12","alias_value":"JJPT672SVMDG","created_at":"2026-07-05T08:25:06.110778+00:00"},{"alias_kind":"pith_short_16","alias_value":"JJPT672SVMDG5HN3","created_at":"2026-07-05T08:25:06.110778+00:00"},{"alias_kind":"pith_short_8","alias_value":"JJPT672S","created_at":"2026-07-05T08:25:06.110778+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":5,"internal_anchor_count":0,"sample":[{"citing_arxiv_id":"2605.30022","citing_title":"Give it Space! Explicit Disentangling of Positional and Semantic Representations in Encoders","ref_index":8,"is_internal_anchor":false},{"citing_arxiv_id":"2502.05564","citing_title":"TabICL: A Tabular Foundation Model for In-Context Learning on Large Data","ref_index":187,"is_internal_anchor":false},{"citing_arxiv_id":"2512.07805","citing_title":"Group Representational Position Encoding","ref_index":6,"is_internal_anchor":false},{"citing_arxiv_id":"2604.18603","citing_title":"Dual Triangle Attention: Effective Bidirectional Attention Without Positional Embeddings","ref_index":32,"is_internal_anchor":false},{"citing_arxiv_id":"2605.05851","citing_title":"Hypothesis generation and updating in large language models","ref_index":59,"is_internal_anchor":false}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/JJPT672SVMDG5HN3CS5ZVZ72KF","json":"https://pith.science/pith/JJPT672SVMDG5HN3CS5ZVZ72KF.json","graph_json":"https://pith.science/api/pith-number/JJPT672SVMDG5HN3CS5ZVZ72KF/graph.json","events_json":"https://pith.science/api/pith-number/JJPT672SVMDG5HN3CS5ZVZ72KF/events.json","paper":"https://pith.science/paper/JJPT672S"},"agent_actions":{"view_html":"https://pith.science/pith/JJPT672SVMDG5HN3CS5ZVZ72KF","download_json":"https://pith.science/pith/JJPT672SVMDG5HN3CS5ZVZ72KF.json","view_paper":"https://pith.science/paper/JJPT672S","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2405.18719&json=true","fetch_graph":"https://pith.science/api/pith-number/JJPT672SVMDG5HN3CS5ZVZ72KF/graph.json","fetch_events":"https://pith.science/api/pith-number/JJPT672SVMDG5HN3CS5ZVZ72KF/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/JJPT672SVMDG5HN3CS5ZVZ72KF/action/timestamp_anchor","attest_storage":"https://pith.science/pith/JJPT672SVMDG5HN3CS5ZVZ72KF/action/storage_attestation","attest_author":"https://pith.science/pith/JJPT672SVMDG5HN3CS5ZVZ72KF/action/author_attestation","sign_citation":"https://pith.science/pith/JJPT672SVMDG5HN3CS5ZVZ72KF/action/citation_signature","submit_replication":"https://pith.science/pith/JJPT672SVMDG5HN3CS5ZVZ72KF/action/replication_record"}},"created_at":"2026-07-05T08:25:06.110778+00:00","updated_at":"2026-07-05T08:25:06.110778+00:00"}