{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2025:NGILMKJHW6KLVJPOJEGJFBGPJW","short_pith_number":"pith:NGILMKJH","schema_version":"1.0","canonical_sha256":"6990b62927b794baa5ee490c9284cf4d8b805f581b99188fdf6dcb74783fb725","source":{"kind":"arxiv","id":"2509.10534","version":3},"attestation_state":"computed","paper":{"title":"Decoupling the \"What\" and \"Where\" With Polar Coordinate Positional Embeddings","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.AI","cs.CL"],"primary_cat":"cs.LG","authors_text":"Anand Gopalakrishnan, J\\\"urgen Schmidhuber, Michael C. Mozer, Robert Csord\\'as","submitted_at":"2025-09-05T14:22:27Z","abstract_excerpt":"The attention mechanism in a Transformer architecture matches key to query based on both content -- the what -- and position in a sequence -- the where. We present an analysis indicating that what and where are entangled in the popular RoPE rotary position embedding. This entanglement can impair performance particularly when decisions require independent matches on these two factors. We propose an improvement to RoPE, which we call Polar Coordinate Position Embeddings or PoPE, that eliminates the what-where confound. PoPE is far superior on a diagnostic task requiring indexing solely by positi"},"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":"2509.10534","kind":"arxiv","version":3},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.LG","submitted_at":"2025-09-05T14:22:27Z","cross_cats_sorted":["cs.AI","cs.CL"],"title_canon_sha256":"b78950eac117b241465dac32082aed707d249cc080b5cd424af4f912644ffee6","abstract_canon_sha256":"983386054c2349985d39f2482bdc629e966446322129d02db964f0de79bcec7a"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-06-09T02:08:33.332632Z","signature_b64":"Eif3DIh4qUkRbXSpF0zTPFl4tBC0NKQRa0vW6EI1S8lGY+6+Y1+Yhz940Gvf1Y/o4pNMnQTXLq0lAGW7qMKHDA==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"6990b62927b794baa5ee490c9284cf4d8b805f581b99188fdf6dcb74783fb725","last_reissued_at":"2026-06-09T02:08:33.331291Z","signature_status":"signed_v1","first_computed_at":"2026-06-09T02:08:33.331291Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Decoupling the \"What\" and \"Where\" With Polar Coordinate Positional Embeddings","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.AI","cs.CL"],"primary_cat":"cs.LG","authors_text":"Anand Gopalakrishnan, J\\\"urgen Schmidhuber, Michael C. Mozer, Robert Csord\\'as","submitted_at":"2025-09-05T14:22:27Z","abstract_excerpt":"The attention mechanism in a Transformer architecture matches key to query based on both content -- the what -- and position in a sequence -- the where. We present an analysis indicating that what and where are entangled in the popular RoPE rotary position embedding. This entanglement can impair performance particularly when decisions require independent matches on these two factors. We propose an improvement to RoPE, which we call Polar Coordinate Position Embeddings or PoPE, that eliminates the what-where confound. PoPE is far superior on a diagnostic task requiring indexing solely by positi"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2509.10534","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/2509.10534/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":"2509.10534","created_at":"2026-06-09T02:08:33.331489+00:00"},{"alias_kind":"arxiv_version","alias_value":"2509.10534v3","created_at":"2026-06-09T02:08:33.331489+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2509.10534","created_at":"2026-06-09T02:08:33.331489+00:00"},{"alias_kind":"pith_short_12","alias_value":"NGILMKJHW6KL","created_at":"2026-06-09T02:08:33.331489+00:00"},{"alias_kind":"pith_short_16","alias_value":"NGILMKJHW6KLVJPO","created_at":"2026-06-09T02:08:33.331489+00:00"},{"alias_kind":"pith_short_8","alias_value":"NGILMKJH","created_at":"2026-06-09T02:08:33.331489+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":3,"internal_anchor_count":3,"sample":[{"citing_arxiv_id":"2606.11275","citing_title":"RoVE: Rotary Value Embeddings Attention for Relative Position-dependent Value Pathways","ref_index":9,"is_internal_anchor":true},{"citing_arxiv_id":"2605.12021","citing_title":"What-Where Transformer: A Slot-Centric Visual Backbone for Concurrent Representation and Localization","ref_index":26,"is_internal_anchor":true},{"citing_arxiv_id":"2604.06070","citing_title":"Short Data, Long Context: Distilling Positional Knowledge in Transformers","ref_index":1,"is_internal_anchor":true}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/NGILMKJHW6KLVJPOJEGJFBGPJW","json":"https://pith.science/pith/NGILMKJHW6KLVJPOJEGJFBGPJW.json","graph_json":"https://pith.science/api/pith-number/NGILMKJHW6KLVJPOJEGJFBGPJW/graph.json","events_json":"https://pith.science/api/pith-number/NGILMKJHW6KLVJPOJEGJFBGPJW/events.json","paper":"https://pith.science/paper/NGILMKJH"},"agent_actions":{"view_html":"https://pith.science/pith/NGILMKJHW6KLVJPOJEGJFBGPJW","download_json":"https://pith.science/pith/NGILMKJHW6KLVJPOJEGJFBGPJW.json","view_paper":"https://pith.science/paper/NGILMKJH","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2509.10534&json=true","fetch_graph":"https://pith.science/api/pith-number/NGILMKJHW6KLVJPOJEGJFBGPJW/graph.json","fetch_events":"https://pith.science/api/pith-number/NGILMKJHW6KLVJPOJEGJFBGPJW/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/NGILMKJHW6KLVJPOJEGJFBGPJW/action/timestamp_anchor","attest_storage":"https://pith.science/pith/NGILMKJHW6KLVJPOJEGJFBGPJW/action/storage_attestation","attest_author":"https://pith.science/pith/NGILMKJHW6KLVJPOJEGJFBGPJW/action/author_attestation","sign_citation":"https://pith.science/pith/NGILMKJHW6KLVJPOJEGJFBGPJW/action/citation_signature","submit_replication":"https://pith.science/pith/NGILMKJHW6KLVJPOJEGJFBGPJW/action/replication_record"}},"created_at":"2026-06-09T02:08:33.331489+00:00","updated_at":"2026-06-09T02:08:33.331489+00:00"}