{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2020:VAZB2GETEBBUDDKOZRFLXTXD4S","short_pith_number":"pith:VAZB2GET","schema_version":"1.0","canonical_sha256":"a8321d18932043418d4ecc4abbcee3e49c5e05ca491a0204a8b65cb411d6126e","source":{"kind":"arxiv","id":"2002.09402","version":3},"attestation_state":"computed","paper":{"title":"Addressing Some Limitations of Transformers with Feedback Memory","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":["cs.CL","stat.ML"],"primary_cat":"cs.LG","authors_text":"Angela Fan, Armand Joulin, Edouard Grave, Sainbayar Sukhbaatar, Thibaut Lavril","submitted_at":"2020-02-21T16:37:57Z","abstract_excerpt":"Transformers have been successfully applied to sequential, auto-regressive tasks despite being feedforward networks. Unlike recurrent neural networks, Transformers use attention to capture temporal relations while processing input tokens in parallel. While this parallelization makes them computationally efficient, it restricts the model from fully exploiting the sequential nature of the input. The representation at a given layer can only access representations from lower layers, rather than the higher level representations already available. In this work, we propose the Feedback Transformer ar"},"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":"2002.09402","kind":"arxiv","version":3},"metadata":{"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.LG","submitted_at":"2020-02-21T16:37:57Z","cross_cats_sorted":["cs.CL","stat.ML"],"title_canon_sha256":"a1b4a34fce25efc929c771f31976adaabf8b7b357716830cebb8c9bc1fb7c85d","abstract_canon_sha256":"88eb98b2a4780ec83cf19a72a024e5f5ffc5216fd12c7201a390f2b658b78480"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T02:09:08.746534Z","signature_b64":"C5DUSLZ9c7N4lm2PO7j8AL4phIkfeS6uj+7poxMCMlKThHMMfTRAaKxWGLCX0uuS/AUM24eubn1i6yi1VonMDQ==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"a8321d18932043418d4ecc4abbcee3e49c5e05ca491a0204a8b65cb411d6126e","last_reissued_at":"2026-07-05T02:09:08.746094Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T02:09:08.746094Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Addressing Some Limitations of Transformers with Feedback Memory","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":["cs.CL","stat.ML"],"primary_cat":"cs.LG","authors_text":"Angela Fan, Armand Joulin, Edouard Grave, Sainbayar Sukhbaatar, Thibaut Lavril","submitted_at":"2020-02-21T16:37:57Z","abstract_excerpt":"Transformers have been successfully applied to sequential, auto-regressive tasks despite being feedforward networks. Unlike recurrent neural networks, Transformers use attention to capture temporal relations while processing input tokens in parallel. While this parallelization makes them computationally efficient, it restricts the model from fully exploiting the sequential nature of the input. The representation at a given layer can only access representations from lower layers, rather than the higher level representations already available. In this work, we propose the Feedback Transformer ar"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2002.09402","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/2002.09402/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":"2002.09402","created_at":"2026-07-05T02:09:08.746155+00:00"},{"alias_kind":"arxiv_version","alias_value":"2002.09402v3","created_at":"2026-07-05T02:09:08.746155+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2002.09402","created_at":"2026-07-05T02:09:08.746155+00:00"},{"alias_kind":"pith_short_12","alias_value":"VAZB2GETEBBU","created_at":"2026-07-05T02:09:08.746155+00:00"},{"alias_kind":"pith_short_16","alias_value":"VAZB2GETEBBUDDKO","created_at":"2026-07-05T02:09:08.746155+00:00"},{"alias_kind":"pith_short_8","alias_value":"VAZB2GET","created_at":"2026-07-05T02:09:08.746155+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":5,"internal_anchor_count":0,"sample":[{"citing_arxiv_id":"2606.06479","citing_title":"Pretraining Recurrent Networks without Recurrence","ref_index":29,"is_internal_anchor":false},{"citing_arxiv_id":"2605.26797","citing_title":"Latent Recurrent Transformer: Architecture Exploration, Training Strategies, and Scaling Behavior","ref_index":3,"is_internal_anchor":false},{"citing_arxiv_id":"2502.05171","citing_title":"Scaling up Test-Time Compute with Latent Reasoning: A Recurrent Depth Approach","ref_index":53,"is_internal_anchor":false},{"citing_arxiv_id":"2604.21215","citing_title":"The Recurrent Transformer: Greater Effective Depth and Efficient Decoding","ref_index":57,"is_internal_anchor":false},{"citing_arxiv_id":"2604.18580","citing_title":"Sessa: Selective State Space Attention","ref_index":17,"is_internal_anchor":false}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/VAZB2GETEBBUDDKOZRFLXTXD4S","json":"https://pith.science/pith/VAZB2GETEBBUDDKOZRFLXTXD4S.json","graph_json":"https://pith.science/api/pith-number/VAZB2GETEBBUDDKOZRFLXTXD4S/graph.json","events_json":"https://pith.science/api/pith-number/VAZB2GETEBBUDDKOZRFLXTXD4S/events.json","paper":"https://pith.science/paper/VAZB2GET"},"agent_actions":{"view_html":"https://pith.science/pith/VAZB2GETEBBUDDKOZRFLXTXD4S","download_json":"https://pith.science/pith/VAZB2GETEBBUDDKOZRFLXTXD4S.json","view_paper":"https://pith.science/paper/VAZB2GET","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2002.09402&json=true","fetch_graph":"https://pith.science/api/pith-number/VAZB2GETEBBUDDKOZRFLXTXD4S/graph.json","fetch_events":"https://pith.science/api/pith-number/VAZB2GETEBBUDDKOZRFLXTXD4S/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/VAZB2GETEBBUDDKOZRFLXTXD4S/action/timestamp_anchor","attest_storage":"https://pith.science/pith/VAZB2GETEBBUDDKOZRFLXTXD4S/action/storage_attestation","attest_author":"https://pith.science/pith/VAZB2GETEBBUDDKOZRFLXTXD4S/action/author_attestation","sign_citation":"https://pith.science/pith/VAZB2GETEBBUDDKOZRFLXTXD4S/action/citation_signature","submit_replication":"https://pith.science/pith/VAZB2GETEBBUDDKOZRFLXTXD4S/action/replication_record"}},"created_at":"2026-07-05T02:09:08.746155+00:00","updated_at":"2026-07-05T02:09:08.746155+00:00"}