{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2024:ZE45AUMCL2LBE7XPAR5APDYJR3","short_pith_number":"pith:ZE45AUMC","schema_version":"1.0","canonical_sha256":"c939d051825e96127eef047a078f098ef85734138f41444ba62b015b36b28d74","source":{"kind":"arxiv","id":"2403.17844","version":2},"attestation_state":"computed","paper":{"title":"Mechanistic Design and Scaling of Hybrid Architectures","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":[],"primary_cat":"cs.LG","authors_text":"Armin W Thomas, Bj\\\"orn Deiseroth, Brian Hie, Ce Zhang, Christopher R\\'e, Eric Nguyen, Kristian Kersting, Michael Poli, Pragaash Ponnusamy, Stefano Ermon, Stefano Massaroli, Taiji Suzuki","submitted_at":"2024-03-26T16:33:12Z","abstract_excerpt":"The development of deep learning architectures is a resource-demanding process, due to a vast design space, long prototyping times, and high compute costs associated with at-scale model training and evaluation. We set out to simplify this process by grounding it in an end-to-end mechanistic architecture design (MAD) pipeline, encompassing small-scale capability unit tests predictive of scaling laws. Through a suite of synthetic token manipulation tasks such as compression and recall, designed to probe capabilities, we identify and test new hybrid architectures constructed from a variety of com"},"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":"2403.17844","kind":"arxiv","version":2},"metadata":{"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.LG","submitted_at":"2024-03-26T16:33:12Z","cross_cats_sorted":[],"title_canon_sha256":"7da585dcf076a9e74ad090f8aeb1635090f495f6984c6a00b015d48e3e622c10","abstract_canon_sha256":"3bcfd42de5d5dacc60d4e6eb8beea8344fe4004dc54c147db1a7ef6e3a9c6b10"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T08:56:47.057813Z","signature_b64":"0XSrRz6SkJrAxLLRSwiQ6a79S6rAvVp+1yQDIk+ryqKHwbkbEBE1MFI9Ja3ZMcr5+fHP9Isw55xrH/0aSkgtAA==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"c939d051825e96127eef047a078f098ef85734138f41444ba62b015b36b28d74","last_reissued_at":"2026-07-05T08:56:47.057370Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T08:56:47.057370Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Mechanistic Design and Scaling of Hybrid Architectures","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":[],"primary_cat":"cs.LG","authors_text":"Armin W Thomas, Bj\\\"orn Deiseroth, Brian Hie, Ce Zhang, Christopher R\\'e, Eric Nguyen, Kristian Kersting, Michael Poli, Pragaash Ponnusamy, Stefano Ermon, Stefano Massaroli, Taiji Suzuki","submitted_at":"2024-03-26T16:33:12Z","abstract_excerpt":"The development of deep learning architectures is a resource-demanding process, due to a vast design space, long prototyping times, and high compute costs associated with at-scale model training and evaluation. We set out to simplify this process by grounding it in an end-to-end mechanistic architecture design (MAD) pipeline, encompassing small-scale capability unit tests predictive of scaling laws. Through a suite of synthetic token manipulation tasks such as compression and recall, designed to probe capabilities, we identify and test new hybrid architectures constructed from a variety of com"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2403.17844","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/2403.17844/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":"2403.17844","created_at":"2026-07-05T08:56:47.057428+00:00"},{"alias_kind":"arxiv_version","alias_value":"2403.17844v2","created_at":"2026-07-05T08:56:47.057428+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2403.17844","created_at":"2026-07-05T08:56:47.057428+00:00"},{"alias_kind":"pith_short_12","alias_value":"ZE45AUMCL2LB","created_at":"2026-07-05T08:56:47.057428+00:00"},{"alias_kind":"pith_short_16","alias_value":"ZE45AUMCL2LBE7XP","created_at":"2026-07-05T08:56:47.057428+00:00"},{"alias_kind":"pith_short_8","alias_value":"ZE45AUMC","created_at":"2026-07-05T08:56:47.057428+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":11,"internal_anchor_count":0,"sample":[{"citing_arxiv_id":"2606.24320","citing_title":"ZONOS2 Technical Report","ref_index":77,"is_internal_anchor":false},{"citing_arxiv_id":"2606.21228","citing_title":"Sakana Fugu Technical Report","ref_index":223,"is_internal_anchor":false},{"citing_arxiv_id":"2606.03825","citing_title":"Dynamic Short Convolutions Improve Transformers","ref_index":57,"is_internal_anchor":false},{"citing_arxiv_id":"2605.31163","citing_title":"Memory by Design: Probabilistic Sequence Layers","ref_index":24,"is_internal_anchor":false},{"citing_arxiv_id":"2606.24320","citing_title":"ZONOS2 Technical Report","ref_index":77,"is_internal_anchor":false},{"citing_arxiv_id":"2605.21070","citing_title":"Towards Understanding Self-Pretraining for Sequence Classification","ref_index":92,"is_internal_anchor":false},{"citing_arxiv_id":"2605.19150","citing_title":"Flash PD-SSM: Memory-Optimized Structured Sparse State-Space Models","ref_index":41,"is_internal_anchor":false},{"citing_arxiv_id":"2510.04800","citing_title":"Hybrid Architectures for Language Models: Systematic Analysis and Design Insights","ref_index":40,"is_internal_anchor":false},{"citing_arxiv_id":"2509.19349","citing_title":"ShinkaEvolve: Towards Open-Ended And Sample-Efficient Program Evolution","ref_index":189,"is_internal_anchor":false},{"citing_arxiv_id":"2605.08451","citing_title":"RubiConv -- Efficient Boundary-Respecting Convolutions","ref_index":16,"is_internal_anchor":false},{"citing_arxiv_id":"2605.05365","citing_title":"ZAYA1-8B Technical Report","ref_index":51,"is_internal_anchor":false}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/ZE45AUMCL2LBE7XPAR5APDYJR3","json":"https://pith.science/pith/ZE45AUMCL2LBE7XPAR5APDYJR3.json","graph_json":"https://pith.science/api/pith-number/ZE45AUMCL2LBE7XPAR5APDYJR3/graph.json","events_json":"https://pith.science/api/pith-number/ZE45AUMCL2LBE7XPAR5APDYJR3/events.json","paper":"https://pith.science/paper/ZE45AUMC"},"agent_actions":{"view_html":"https://pith.science/pith/ZE45AUMCL2LBE7XPAR5APDYJR3","download_json":"https://pith.science/pith/ZE45AUMCL2LBE7XPAR5APDYJR3.json","view_paper":"https://pith.science/paper/ZE45AUMC","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2403.17844&json=true","fetch_graph":"https://pith.science/api/pith-number/ZE45AUMCL2LBE7XPAR5APDYJR3/graph.json","fetch_events":"https://pith.science/api/pith-number/ZE45AUMCL2LBE7XPAR5APDYJR3/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/ZE45AUMCL2LBE7XPAR5APDYJR3/action/timestamp_anchor","attest_storage":"https://pith.science/pith/ZE45AUMCL2LBE7XPAR5APDYJR3/action/storage_attestation","attest_author":"https://pith.science/pith/ZE45AUMCL2LBE7XPAR5APDYJR3/action/author_attestation","sign_citation":"https://pith.science/pith/ZE45AUMCL2LBE7XPAR5APDYJR3/action/citation_signature","submit_replication":"https://pith.science/pith/ZE45AUMCL2LBE7XPAR5APDYJR3/action/replication_record"}},"created_at":"2026-07-05T08:56:47.057428+00:00","updated_at":"2026-07-05T08:56:47.057428+00:00"}