{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2024:T5JOZFCOF2KJBDWBNNMPADZWFO","short_pith_number":"pith:T5JOZFCO","schema_version":"1.0","canonical_sha256":"9f52ec944e2e94908ec16b58f00f362b8bb029e0e83783e7b1f931f4676d3e4b","source":{"kind":"arxiv","id":"2404.09173","version":3},"attestation_state":"computed","paper":{"title":"TransformerFAM: Feedback attention is working memory","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.AI","cs.CL"],"primary_cat":"cs.LG","authors_text":"Dongseong Hwang, Khe Chai Sim, Pedro Moreno Mengibar, Weiran Wang, Zhuoyuan Huo","submitted_at":"2024-04-14T07:43:45Z","abstract_excerpt":"While Transformers have revolutionized deep learning, their quadratic attention complexity hinders their ability to process infinitely long inputs. We propose Feedback Attention Memory (FAM), a novel Transformer architecture that leverages a feedback loop to enable the network to attend to its own latent representations. This design fosters the emergence of working memory within the Transformer, allowing it to process indefinitely long sequences. TransformerFAM requires no additional weights, enabling seamless integration with pre-trained models. Our experiments show that TransformerFAM signif"},"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":"2404.09173","kind":"arxiv","version":3},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.LG","submitted_at":"2024-04-14T07:43:45Z","cross_cats_sorted":["cs.AI","cs.CL"],"title_canon_sha256":"c93c52189e4bd5f98fe3d9febfd61720ec47077044d8b4340096eedb6f37e269","abstract_canon_sha256":"8e5eda486ed3d465f7f1833235725b588f67d25efa4290618f692636522e92a6"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T08:16:26.615785Z","signature_b64":"qNdh+aqCwiZyhEfmnj+ZOZ7PiEtj2Nm3eGz9GPrZpHJZpJs2QixpxCHTwBHFbQAh6tNy4KWJQzKryQi8b3hGCw==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"9f52ec944e2e94908ec16b58f00f362b8bb029e0e83783e7b1f931f4676d3e4b","last_reissued_at":"2026-07-05T08:16:26.615383Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T08:16:26.615383Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"TransformerFAM: Feedback attention is working memory","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.AI","cs.CL"],"primary_cat":"cs.LG","authors_text":"Dongseong Hwang, Khe Chai Sim, Pedro Moreno Mengibar, Weiran Wang, Zhuoyuan Huo","submitted_at":"2024-04-14T07:43:45Z","abstract_excerpt":"While Transformers have revolutionized deep learning, their quadratic attention complexity hinders their ability to process infinitely long inputs. We propose Feedback Attention Memory (FAM), a novel Transformer architecture that leverages a feedback loop to enable the network to attend to its own latent representations. This design fosters the emergence of working memory within the Transformer, allowing it to process indefinitely long sequences. TransformerFAM requires no additional weights, enabling seamless integration with pre-trained models. Our experiments show that TransformerFAM signif"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2404.09173","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/2404.09173/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":"2404.09173","created_at":"2026-07-05T08:16:26.615441+00:00"},{"alias_kind":"arxiv_version","alias_value":"2404.09173v3","created_at":"2026-07-05T08:16:26.615441+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2404.09173","created_at":"2026-07-05T08:16:26.615441+00:00"},{"alias_kind":"pith_short_12","alias_value":"T5JOZFCOF2KJ","created_at":"2026-07-05T08:16:26.615441+00:00"},{"alias_kind":"pith_short_16","alias_value":"T5JOZFCOF2KJBDWB","created_at":"2026-07-05T08:16:26.615441+00:00"},{"alias_kind":"pith_short_8","alias_value":"T5JOZFCO","created_at":"2026-07-05T08:16:26.615441+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":4,"internal_anchor_count":0,"sample":[{"citing_arxiv_id":"2605.26797","citing_title":"Latent Recurrent Transformer: Architecture Exploration, Training Strategies, and Scaling Behavior","ref_index":9,"is_internal_anchor":false},{"citing_arxiv_id":"2404.07143","citing_title":"Leave No Context Behind: Efficient Infinite Context Transformers with Infini-attention","ref_index":15,"is_internal_anchor":false},{"citing_arxiv_id":"2604.21215","citing_title":"The Recurrent Transformer: Greater Effective Depth and Efficient Decoding","ref_index":61,"is_internal_anchor":false},{"citing_arxiv_id":"2604.18580","citing_title":"Sessa: Selective State Space Attention","ref_index":20,"is_internal_anchor":false}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/T5JOZFCOF2KJBDWBNNMPADZWFO","json":"https://pith.science/pith/T5JOZFCOF2KJBDWBNNMPADZWFO.json","graph_json":"https://pith.science/api/pith-number/T5JOZFCOF2KJBDWBNNMPADZWFO/graph.json","events_json":"https://pith.science/api/pith-number/T5JOZFCOF2KJBDWBNNMPADZWFO/events.json","paper":"https://pith.science/paper/T5JOZFCO"},"agent_actions":{"view_html":"https://pith.science/pith/T5JOZFCOF2KJBDWBNNMPADZWFO","download_json":"https://pith.science/pith/T5JOZFCOF2KJBDWBNNMPADZWFO.json","view_paper":"https://pith.science/paper/T5JOZFCO","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2404.09173&json=true","fetch_graph":"https://pith.science/api/pith-number/T5JOZFCOF2KJBDWBNNMPADZWFO/graph.json","fetch_events":"https://pith.science/api/pith-number/T5JOZFCOF2KJBDWBNNMPADZWFO/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/T5JOZFCOF2KJBDWBNNMPADZWFO/action/timestamp_anchor","attest_storage":"https://pith.science/pith/T5JOZFCOF2KJBDWBNNMPADZWFO/action/storage_attestation","attest_author":"https://pith.science/pith/T5JOZFCOF2KJBDWBNNMPADZWFO/action/author_attestation","sign_citation":"https://pith.science/pith/T5JOZFCOF2KJBDWBNNMPADZWFO/action/citation_signature","submit_replication":"https://pith.science/pith/T5JOZFCOF2KJBDWBNNMPADZWFO/action/replication_record"}},"created_at":"2026-07-05T08:16:26.615441+00:00","updated_at":"2026-07-05T08:16:26.615441+00:00"}