{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2024:QFQTTTINDXZHAJOC5PPRSVNENX","short_pith_number":"pith:QFQTTTIN","schema_version":"1.0","canonical_sha256":"816139cd0d1df27025c2ebdf1955a46de9b80ca6de307223227865c684910e0b","source":{"kind":"arxiv","id":"2403.01590","version":2},"attestation_state":"computed","paper":{"title":"The Hidden Attention of Mamba Models","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":[],"primary_cat":"cs.LG","authors_text":"Ameen Ali, Itamar Zimerman, Lior Wolf","submitted_at":"2024-03-03T18:58:21Z","abstract_excerpt":"The Mamba layer offers an efficient selective state space model (SSM) that is highly effective in modeling multiple domains, including NLP, long-range sequence processing, and computer vision. Selective SSMs are viewed as dual models, in which one trains in parallel on the entire sequence via an IO-aware parallel scan, and deploys in an autoregressive manner. We add a third view and show that such models can be viewed as attention-driven models. This new perspective enables us to empirically and theoretically compare the underlying mechanisms to that of the self-attention layers in transformer"},"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":"2403.01590","kind":"arxiv","version":2},"metadata":{"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.LG","submitted_at":"2024-03-03T18:58:21Z","cross_cats_sorted":[],"title_canon_sha256":"ef402acbe24be3841c89ab14aecee6dc5606bc09e49258533df4b965d14e4e62","abstract_canon_sha256":"ed2bb35937b3fb3d317e4100238b7e5a4b7b27e66cb276ca575441226f46846d"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T08:02:51.496425Z","signature_b64":"8QmP5ROWjfothIpnu7u3LN2AWNQRFAJcF+LFWlbw6mQ57VBnfe8qSr4z2bHBYhLPHfQBGQpyUkqqP8DN3wzBAA==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"816139cd0d1df27025c2ebdf1955a46de9b80ca6de307223227865c684910e0b","last_reissued_at":"2026-07-05T08:02:51.495990Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T08:02:51.495990Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"The Hidden Attention of Mamba Models","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":[],"primary_cat":"cs.LG","authors_text":"Ameen Ali, Itamar Zimerman, Lior Wolf","submitted_at":"2024-03-03T18:58:21Z","abstract_excerpt":"The Mamba layer offers an efficient selective state space model (SSM) that is highly effective in modeling multiple domains, including NLP, long-range sequence processing, and computer vision. Selective SSMs are viewed as dual models, in which one trains in parallel on the entire sequence via an IO-aware parallel scan, and deploys in an autoregressive manner. We add a third view and show that such models can be viewed as attention-driven models. This new perspective enables us to empirically and theoretically compare the underlying mechanisms to that of the self-attention layers in transformer"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2403.01590","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/2403.01590/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":"2403.01590","created_at":"2026-07-05T08:02:51.496051+00:00"},{"alias_kind":"arxiv_version","alias_value":"2403.01590v2","created_at":"2026-07-05T08:02:51.496051+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2403.01590","created_at":"2026-07-05T08:02:51.496051+00:00"},{"alias_kind":"pith_short_12","alias_value":"QFQTTTINDXZH","created_at":"2026-07-05T08:02:51.496051+00:00"},{"alias_kind":"pith_short_16","alias_value":"QFQTTTINDXZHAJOC","created_at":"2026-07-05T08:02:51.496051+00:00"},{"alias_kind":"pith_short_8","alias_value":"QFQTTTIN","created_at":"2026-07-05T08:02:51.496051+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":5,"internal_anchor_count":0,"sample":[{"citing_arxiv_id":"2606.22325","citing_title":"All Routes Lead to Collapse","ref_index":5,"is_internal_anchor":false},{"citing_arxiv_id":"2606.10932","citing_title":"Density Field State Space Models: 1-Bit Distillation, Efficient Inference, and Knowledge Organization in Mamba-2","ref_index":1,"is_internal_anchor":false},{"citing_arxiv_id":"2503.18970","citing_title":"Advancing Intelligent Sequence Modeling: Evolution, Trade-offs, and Applications of State- Space Architectures from S4 to Mamba","ref_index":59,"is_internal_anchor":false},{"citing_arxiv_id":"2403.19887","citing_title":"Jamba: A Hybrid Transformer-Mamba Language Model","ref_index":1,"is_internal_anchor":false},{"citing_arxiv_id":"2405.21060","citing_title":"Transformers are SSMs: Generalized Models and Efficient Algorithms Through Structured State Space Duality","ref_index":4,"is_internal_anchor":false}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/QFQTTTINDXZHAJOC5PPRSVNENX","json":"https://pith.science/pith/QFQTTTINDXZHAJOC5PPRSVNENX.json","graph_json":"https://pith.science/api/pith-number/QFQTTTINDXZHAJOC5PPRSVNENX/graph.json","events_json":"https://pith.science/api/pith-number/QFQTTTINDXZHAJOC5PPRSVNENX/events.json","paper":"https://pith.science/paper/QFQTTTIN"},"agent_actions":{"view_html":"https://pith.science/pith/QFQTTTINDXZHAJOC5PPRSVNENX","download_json":"https://pith.science/pith/QFQTTTINDXZHAJOC5PPRSVNENX.json","view_paper":"https://pith.science/paper/QFQTTTIN","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2403.01590&json=true","fetch_graph":"https://pith.science/api/pith-number/QFQTTTINDXZHAJOC5PPRSVNENX/graph.json","fetch_events":"https://pith.science/api/pith-number/QFQTTTINDXZHAJOC5PPRSVNENX/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/QFQTTTINDXZHAJOC5PPRSVNENX/action/timestamp_anchor","attest_storage":"https://pith.science/pith/QFQTTTINDXZHAJOC5PPRSVNENX/action/storage_attestation","attest_author":"https://pith.science/pith/QFQTTTINDXZHAJOC5PPRSVNENX/action/author_attestation","sign_citation":"https://pith.science/pith/QFQTTTINDXZHAJOC5PPRSVNENX/action/citation_signature","submit_replication":"https://pith.science/pith/QFQTTTINDXZHAJOC5PPRSVNENX/action/replication_record"}},"created_at":"2026-07-05T08:02:51.496051+00:00","updated_at":"2026-07-05T08:02:51.496051+00:00"}