{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2021:QC4DQG7IHMENRHFCK2YEWWTMXW","short_pith_number":"pith:QC4DQG7I","schema_version":"1.0","canonical_sha256":"80b8381be83b08d89ca256b04b5a6cbd893eb9d2daeb5606c9ef80ff82e2e469","source":{"kind":"arxiv","id":"2106.06295","version":2},"attestation_state":"computed","paper":{"title":"Going Beyond Linear Transformers with Recurrent Fast Weight Programmers","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":[],"primary_cat":"cs.LG","authors_text":"Imanol Schlag, J\\\"urgen Schmidhuber, Kazuki Irie, R\\'obert Csord\\'as","submitted_at":"2021-06-11T10:32:11Z","abstract_excerpt":"Transformers with linearised attention (''linear Transformers'') have demonstrated the practical scalability and effectiveness of outer product-based Fast Weight Programmers (FWPs) from the '90s. However, the original FWP formulation is more general than the one of linear Transformers: a slow neural network (NN) continually reprograms the weights of a fast NN with arbitrary architecture. In existing linear Transformers, both NNs are feedforward and consist of a single layer. Here we explore new variations by adding recurrence to the slow and fast nets. We evaluate our novel recurrent FWPs (RFW"},"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":"2106.06295","kind":"arxiv","version":2},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.LG","submitted_at":"2021-06-11T10:32:11Z","cross_cats_sorted":[],"title_canon_sha256":"94756e57933b546416a827083ad7e80a8e5cbcadb6ad82560a764eed3f9b7dff","abstract_canon_sha256":"0604d7fe446bce801b5f399ef439b62acbeb7543fef7b08d3a74a43496bceed9"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T03:26:18.174515Z","signature_b64":"Plc/F4XjvZoX6+I4SrlGSjYmHU0Pf1TRV2rHIO09MOtw2JNrcUF51/IKWAM45G1T0QaBsWVk8v6WXaFToGEfAQ==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"80b8381be83b08d89ca256b04b5a6cbd893eb9d2daeb5606c9ef80ff82e2e469","last_reissued_at":"2026-07-05T03:26:18.173999Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T03:26:18.173999Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Going Beyond Linear Transformers with Recurrent Fast Weight Programmers","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":[],"primary_cat":"cs.LG","authors_text":"Imanol Schlag, J\\\"urgen Schmidhuber, Kazuki Irie, R\\'obert Csord\\'as","submitted_at":"2021-06-11T10:32:11Z","abstract_excerpt":"Transformers with linearised attention (''linear Transformers'') have demonstrated the practical scalability and effectiveness of outer product-based Fast Weight Programmers (FWPs) from the '90s. However, the original FWP formulation is more general than the one of linear Transformers: a slow neural network (NN) continually reprograms the weights of a fast NN with arbitrary architecture. In existing linear Transformers, both NNs are feedforward and consist of a single layer. Here we explore new variations by adding recurrence to the slow and fast nets. We evaluate our novel recurrent FWPs (RFW"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2106.06295","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/2106.06295/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":"2106.06295","created_at":"2026-07-05T03:26:18.174060+00:00"},{"alias_kind":"arxiv_version","alias_value":"2106.06295v2","created_at":"2026-07-05T03:26:18.174060+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2106.06295","created_at":"2026-07-05T03:26:18.174060+00:00"},{"alias_kind":"pith_short_12","alias_value":"QC4DQG7IHMEN","created_at":"2026-07-05T03:26:18.174060+00:00"},{"alias_kind":"pith_short_16","alias_value":"QC4DQG7IHMENRHFC","created_at":"2026-07-05T03:26:18.174060+00:00"},{"alias_kind":"pith_short_8","alias_value":"QC4DQG7I","created_at":"2026-07-05T03:26:18.174060+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":1,"internal_anchor_count":1,"sample":[{"citing_arxiv_id":"2608.10288","citing_title":"Power law graph attention: exact generalization of scaled dot-product attention, empirical collapse at inference","ref_index":19,"is_internal_anchor":true}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/QC4DQG7IHMENRHFCK2YEWWTMXW","json":"https://pith.science/pith/QC4DQG7IHMENRHFCK2YEWWTMXW.json","graph_json":"https://pith.science/api/pith-number/QC4DQG7IHMENRHFCK2YEWWTMXW/graph.json","events_json":"https://pith.science/api/pith-number/QC4DQG7IHMENRHFCK2YEWWTMXW/events.json","paper":"https://pith.science/paper/QC4DQG7I"},"agent_actions":{"view_html":"https://pith.science/pith/QC4DQG7IHMENRHFCK2YEWWTMXW","download_json":"https://pith.science/pith/QC4DQG7IHMENRHFCK2YEWWTMXW.json","view_paper":"https://pith.science/paper/QC4DQG7I","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2106.06295&json=true","fetch_graph":"https://pith.science/api/pith-number/QC4DQG7IHMENRHFCK2YEWWTMXW/graph.json","fetch_events":"https://pith.science/api/pith-number/QC4DQG7IHMENRHFCK2YEWWTMXW/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/QC4DQG7IHMENRHFCK2YEWWTMXW/action/timestamp_anchor","attest_storage":"https://pith.science/pith/QC4DQG7IHMENRHFCK2YEWWTMXW/action/storage_attestation","attest_author":"https://pith.science/pith/QC4DQG7IHMENRHFCK2YEWWTMXW/action/author_attestation","sign_citation":"https://pith.science/pith/QC4DQG7IHMENRHFCK2YEWWTMXW/action/citation_signature","submit_replication":"https://pith.science/pith/QC4DQG7IHMENRHFCK2YEWWTMXW/action/replication_record"}},"created_at":"2026-07-05T03:26:18.174060+00:00","updated_at":"2026-07-05T03:26:18.174060+00:00"}