{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2024:CW3IKBEHPUL5Y3UDVZ6ILT7M2C","short_pith_number":"pith:CW3IKBEH","schema_version":"1.0","canonical_sha256":"15b68504877d17dc6e83ae7c85cfecd0b9739e7834a30f1021b2e95f464f2749","source":{"kind":"arxiv","id":"2405.13956","version":2},"attestation_state":"computed","paper":{"title":"Attention as an RNN","license":"http://creativecommons.org/licenses/by-nc-sa/4.0/","headline":"","cross_cats":[],"primary_cat":"cs.LG","authors_text":"Frederick Tung, Greg Mori, Hossein Hajimirsadeghi, Leo Feng, Mohamed Osama Ahmed, Yoshua Bengio","submitted_at":"2024-05-22T19:45:01Z","abstract_excerpt":"The advent of Transformers marked a significant breakthrough in sequence modelling, providing a highly performant architecture capable of leveraging GPU parallelism. However, Transformers are computationally expensive at inference time, limiting their applications, particularly in low-resource settings (e.g., mobile and embedded devices). Addressing this, we (1) begin by showing that attention can be viewed as a special Recurrent Neural Network (RNN) with the ability to compute its \\textit{many-to-one} RNN output efficiently. We then (2) show that popular attention-based models such as Transfo"},"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":"2405.13956","kind":"arxiv","version":2},"metadata":{"license":"http://creativecommons.org/licenses/by-nc-sa/4.0/","primary_cat":"cs.LG","submitted_at":"2024-05-22T19:45:01Z","cross_cats_sorted":[],"title_canon_sha256":"edc833e7aaffd91a68acdf7962ec318bc5bb19c46e7768b552d6b9c327808e6f","abstract_canon_sha256":"9b3ebcb8eceec69dfc9efaa7b64656e6de738b9a83aaf39e2cc3b81007506914"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T08:24:06.149266Z","signature_b64":"cY3bJZaRy4Hi4Njt8wMjZ+pJ0Ta4Rkmp1j0GqRFAbY/ACzY9F5e0cwv91Tn4+IJ7MpJ87gIYlLDXKyUXzUbpBA==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"15b68504877d17dc6e83ae7c85cfecd0b9739e7834a30f1021b2e95f464f2749","last_reissued_at":"2026-07-05T08:24:06.148780Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T08:24:06.148780Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Attention as an RNN","license":"http://creativecommons.org/licenses/by-nc-sa/4.0/","headline":"","cross_cats":[],"primary_cat":"cs.LG","authors_text":"Frederick Tung, Greg Mori, Hossein Hajimirsadeghi, Leo Feng, Mohamed Osama Ahmed, Yoshua Bengio","submitted_at":"2024-05-22T19:45:01Z","abstract_excerpt":"The advent of Transformers marked a significant breakthrough in sequence modelling, providing a highly performant architecture capable of leveraging GPU parallelism. However, Transformers are computationally expensive at inference time, limiting their applications, particularly in low-resource settings (e.g., mobile and embedded devices). Addressing this, we (1) begin by showing that attention can be viewed as a special Recurrent Neural Network (RNN) with the ability to compute its \\textit{many-to-one} RNN output efficiently. We then (2) show that popular attention-based models such as Transfo"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2405.13956","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/2405.13956/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":"2405.13956","created_at":"2026-07-05T08:24:06.148845+00:00"},{"alias_kind":"arxiv_version","alias_value":"2405.13956v2","created_at":"2026-07-05T08:24:06.148845+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2405.13956","created_at":"2026-07-05T08:24:06.148845+00:00"},{"alias_kind":"pith_short_12","alias_value":"CW3IKBEHPUL5","created_at":"2026-07-05T08:24:06.148845+00:00"},{"alias_kind":"pith_short_16","alias_value":"CW3IKBEHPUL5Y3UD","created_at":"2026-07-05T08:24:06.148845+00:00"},{"alias_kind":"pith_short_8","alias_value":"CW3IKBEH","created_at":"2026-07-05T08:24:06.148845+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":2,"internal_anchor_count":0,"sample":[{"citing_arxiv_id":"2507.02259","citing_title":"MemAgent: Reshaping Long-Context LLM with Multi-Conv RL-based Memory Agent","ref_index":40,"is_internal_anchor":false},{"citing_arxiv_id":"2604.23798","citing_title":"ELSA: Exact Linear-Scan Attention for Fast and Memory-Light Vision Transformers","ref_index":13,"is_internal_anchor":false}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/CW3IKBEHPUL5Y3UDVZ6ILT7M2C","json":"https://pith.science/pith/CW3IKBEHPUL5Y3UDVZ6ILT7M2C.json","graph_json":"https://pith.science/api/pith-number/CW3IKBEHPUL5Y3UDVZ6ILT7M2C/graph.json","events_json":"https://pith.science/api/pith-number/CW3IKBEHPUL5Y3UDVZ6ILT7M2C/events.json","paper":"https://pith.science/paper/CW3IKBEH"},"agent_actions":{"view_html":"https://pith.science/pith/CW3IKBEHPUL5Y3UDVZ6ILT7M2C","download_json":"https://pith.science/pith/CW3IKBEHPUL5Y3UDVZ6ILT7M2C.json","view_paper":"https://pith.science/paper/CW3IKBEH","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2405.13956&json=true","fetch_graph":"https://pith.science/api/pith-number/CW3IKBEHPUL5Y3UDVZ6ILT7M2C/graph.json","fetch_events":"https://pith.science/api/pith-number/CW3IKBEHPUL5Y3UDVZ6ILT7M2C/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/CW3IKBEHPUL5Y3UDVZ6ILT7M2C/action/timestamp_anchor","attest_storage":"https://pith.science/pith/CW3IKBEHPUL5Y3UDVZ6ILT7M2C/action/storage_attestation","attest_author":"https://pith.science/pith/CW3IKBEHPUL5Y3UDVZ6ILT7M2C/action/author_attestation","sign_citation":"https://pith.science/pith/CW3IKBEHPUL5Y3UDVZ6ILT7M2C/action/citation_signature","submit_replication":"https://pith.science/pith/CW3IKBEHPUL5Y3UDVZ6ILT7M2C/action/replication_record"}},"created_at":"2026-07-05T08:24:06.148845+00:00","updated_at":"2026-07-05T08:24:06.148845+00:00"}