{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2021:L4HBYMSGGLADREI5BTLZWAK5SC","short_pith_number":"pith:L4HBYMSG","schema_version":"1.0","canonical_sha256":"5f0e1c324632c038911d0cd79b015d90a564693dc4cdbaf6b813db2749c95fd3","source":{"kind":"arxiv","id":"2102.10882","version":3},"attestation_state":"computed","paper":{"title":"Conditional Positional Encodings for Vision Transformers","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":["cs.AI","cs.LG"],"primary_cat":"cs.CV","authors_text":"Bo Zhang, Chunhua Shen, Xiangxiang Chu, Xinlong Wang, Zhi Tian","submitted_at":"2021-02-22T10:29:55Z","abstract_excerpt":"We propose a conditional positional encoding (CPE) scheme for vision Transformers. Unlike previous fixed or learnable positional encodings, which are pre-defined and independent of input tokens, CPE is dynamically generated and conditioned on the local neighborhood of the input tokens. As a result, CPE can easily generalize to the input sequences that are longer than what the model has ever seen during training. Besides, CPE can keep the desired translation-invariance in the image classification task, resulting in improved performance. We implement CPE with a simple Position Encoding Generator"},"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":"2102.10882","kind":"arxiv","version":3},"metadata":{"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.CV","submitted_at":"2021-02-22T10:29:55Z","cross_cats_sorted":["cs.AI","cs.LG"],"title_canon_sha256":"cb2c952c0933b93cc9f81bee387a916fb74c075d683930814f4a73c333a2e6a7","abstract_canon_sha256":"70acb8249bd8326cfe92d93dace9f9e4e16092c63d61e7bc1e96b01307f97071"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T05:40:42.336469Z","signature_b64":"3dJQoaI3hrmOsQ3ye26i1aSERAoRdO08N6EFRP4cvLwQKYeBnhQsVhdAgjYtuBAjKlBNAdb47pENORtSKKdAAA==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"5f0e1c324632c038911d0cd79b015d90a564693dc4cdbaf6b813db2749c95fd3","last_reissued_at":"2026-07-05T05:40:42.335962Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T05:40:42.335962Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Conditional Positional Encodings for Vision Transformers","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":["cs.AI","cs.LG"],"primary_cat":"cs.CV","authors_text":"Bo Zhang, Chunhua Shen, Xiangxiang Chu, Xinlong Wang, Zhi Tian","submitted_at":"2021-02-22T10:29:55Z","abstract_excerpt":"We propose a conditional positional encoding (CPE) scheme for vision Transformers. Unlike previous fixed or learnable positional encodings, which are pre-defined and independent of input tokens, CPE is dynamically generated and conditioned on the local neighborhood of the input tokens. As a result, CPE can easily generalize to the input sequences that are longer than what the model has ever seen during training. Besides, CPE can keep the desired translation-invariance in the image classification task, resulting in improved performance. We implement CPE with a simple Position Encoding Generator"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2102.10882","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/2102.10882/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":"2102.10882","created_at":"2026-07-05T05:40:42.336023+00:00"},{"alias_kind":"arxiv_version","alias_value":"2102.10882v3","created_at":"2026-07-05T05:40:42.336023+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2102.10882","created_at":"2026-07-05T05:40:42.336023+00:00"},{"alias_kind":"pith_short_12","alias_value":"L4HBYMSGGLAD","created_at":"2026-07-05T05:40:42.336023+00:00"},{"alias_kind":"pith_short_16","alias_value":"L4HBYMSGGLADREI5","created_at":"2026-07-05T05:40:42.336023+00:00"},{"alias_kind":"pith_short_8","alias_value":"L4HBYMSG","created_at":"2026-07-05T05:40:42.336023+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":13,"internal_anchor_count":0,"sample":[{"citing_arxiv_id":"2606.24498","citing_title":"VistaRef: Boosting Visual Spatial Orientation Awareness for Pointing-to-Object Detection","ref_index":4,"is_internal_anchor":false},{"citing_arxiv_id":"2606.09659","citing_title":"End-to-End Context Compression at Scale","ref_index":12,"is_internal_anchor":false},{"citing_arxiv_id":"2607.00249","citing_title":"Device Passport: Enabling Spatio-Temporal Pretrained Models to Generalize Across Input Layouts","ref_index":4,"is_internal_anchor":false},{"citing_arxiv_id":"2606.31585","citing_title":"DPPE: Rethinking Camera-Based Positional Encoding for Scaling Multi-View Transformers","ref_index":8,"is_internal_anchor":false},{"citing_arxiv_id":"2605.25949","citing_title":"Small Models, Strong Priors: Architectural Inductive Bias for Parameter-Efficient Neural PDE Solvers","ref_index":7,"is_internal_anchor":false},{"citing_arxiv_id":"2408.12406","citing_title":"Generalized SAM: Efficient Fine-Tuning of SAM for Variable Input Image Sizes","ref_index":10,"is_internal_anchor":false},{"citing_arxiv_id":"2504.14386","citing_title":"LOOPE: Learnable Optimal Patch Order in Positional Embeddings for Vision Transformers","ref_index":6,"is_internal_anchor":false},{"citing_arxiv_id":"2402.17762","citing_title":"Massive Activations in Large Language Models","ref_index":14,"is_internal_anchor":false},{"citing_arxiv_id":"2103.14030","citing_title":"Swin Transformer: Hierarchical Vision Transformer using Shifted Windows","ref_index":15,"is_internal_anchor":false},{"citing_arxiv_id":"2303.16199","citing_title":"LLaMA-Adapter: Efficient Fine-tuning of Language Models with Zero-init Attention","ref_index":211,"is_internal_anchor":false},{"citing_arxiv_id":"2605.11131","citing_title":"USEMA: a Scalable Efficient Mamba Like Attention for Medical Image Segmentation","ref_index":3,"is_internal_anchor":false},{"citing_arxiv_id":"2604.21035","citing_title":"Masked-Token Prediction for Anomaly Detection at the Large Hadron Collider","ref_index":14,"is_internal_anchor":false},{"citing_arxiv_id":"2604.05215","citing_title":"Hierarchical Mesh Transformers with Topology-Guided Pretraining for Morphometric Analysis of Brain Structures","ref_index":5,"is_internal_anchor":false}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/L4HBYMSGGLADREI5BTLZWAK5SC","json":"https://pith.science/pith/L4HBYMSGGLADREI5BTLZWAK5SC.json","graph_json":"https://pith.science/api/pith-number/L4HBYMSGGLADREI5BTLZWAK5SC/graph.json","events_json":"https://pith.science/api/pith-number/L4HBYMSGGLADREI5BTLZWAK5SC/events.json","paper":"https://pith.science/paper/L4HBYMSG"},"agent_actions":{"view_html":"https://pith.science/pith/L4HBYMSGGLADREI5BTLZWAK5SC","download_json":"https://pith.science/pith/L4HBYMSGGLADREI5BTLZWAK5SC.json","view_paper":"https://pith.science/paper/L4HBYMSG","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2102.10882&json=true","fetch_graph":"https://pith.science/api/pith-number/L4HBYMSGGLADREI5BTLZWAK5SC/graph.json","fetch_events":"https://pith.science/api/pith-number/L4HBYMSGGLADREI5BTLZWAK5SC/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/L4HBYMSGGLADREI5BTLZWAK5SC/action/timestamp_anchor","attest_storage":"https://pith.science/pith/L4HBYMSGGLADREI5BTLZWAK5SC/action/storage_attestation","attest_author":"https://pith.science/pith/L4HBYMSGGLADREI5BTLZWAK5SC/action/author_attestation","sign_citation":"https://pith.science/pith/L4HBYMSGGLADREI5BTLZWAK5SC/action/citation_signature","submit_replication":"https://pith.science/pith/L4HBYMSGGLADREI5BTLZWAK5SC/action/replication_record"}},"created_at":"2026-07-05T05:40:42.336023+00:00","updated_at":"2026-07-05T05:40:42.336023+00:00"}