{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2026:GREATM223BKA4U5SON2LFQB4B4","short_pith_number":"pith:GREATM22","schema_version":"1.0","canonical_sha256":"344809b35ad8540e53b27374b2c03c0f1eab0d470070fbf096c065f2b5c2f664","source":{"kind":"arxiv","id":"2605.15871","version":1},"attestation_state":"computed","paper":{"title":"Agentic Discovery of Neural Architectures: AIRA-Compose and AIRA-Design","license":"http://creativecommons.org/licenses/by-nc-sa/4.0/","headline":"","cross_cats":[],"primary_cat":"cs.AI","authors_text":"Alberto Pepe, Anton Protopopov, Bilge Acun, Carole-Jean Wu, Chien-Yu Lin, Despoina Magka, Yannan Nellie Wu, Yoram Bachrach","submitted_at":"2026-05-15T11:40:41Z","abstract_excerpt":"Toward recursive self-improvement, we investigate LLM agents autonomously designing foundation models beyond standard Transformers. We introduce a dual-framework approach: AIRA-Compose for high-level architecture search, and AIRA-Design for low-level mechanistic implementation. AIRA-Compose uses 11 agents to explore fundamental computational primitives under a 24-hour budget. Agents evaluate million-parameter candidates, extrapolating top designs to 350M, 1B, and 3B scales. This yields 14 architectures across two families: AIRAformers (Transformer-based) and AIRAhybrids (Transformer-Mamba). Pr"},"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":"2605.15871","kind":"arxiv","version":1},"metadata":{"license":"http://creativecommons.org/licenses/by-nc-sa/4.0/","primary_cat":"cs.AI","submitted_at":"2026-05-15T11:40:41Z","cross_cats_sorted":[],"title_canon_sha256":"bfa53c849e1d4bd42c0b7f2bb5e238fbf452cc2e4cf23e72b7519052e55b50c5","abstract_canon_sha256":"fbe1f48c0dabc4c61baf8f2a9a6770a81b37aa67212cc49afb9aeaacf8731610"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-05-20T00:01:22.966620Z","signature_b64":"QSw8eEuyf0qR1gqLAbC3AO1sEfb+WX3bJX1BzJcmjoCDbQ3XB3A5MesK+BZ/II9mxFJ4VUPSalzovj4TBpkhBQ==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"344809b35ad8540e53b27374b2c03c0f1eab0d470070fbf096c065f2b5c2f664","last_reissued_at":"2026-05-20T00:01:22.965866Z","signature_status":"signed_v1","first_computed_at":"2026-05-20T00:01:22.965866Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Agentic Discovery of Neural Architectures: AIRA-Compose and AIRA-Design","license":"http://creativecommons.org/licenses/by-nc-sa/4.0/","headline":"","cross_cats":[],"primary_cat":"cs.AI","authors_text":"Alberto Pepe, Anton Protopopov, Bilge Acun, Carole-Jean Wu, Chien-Yu Lin, Despoina Magka, Yannan Nellie Wu, Yoram Bachrach","submitted_at":"2026-05-15T11:40:41Z","abstract_excerpt":"Toward recursive self-improvement, we investigate LLM agents autonomously designing foundation models beyond standard Transformers. We introduce a dual-framework approach: AIRA-Compose for high-level architecture search, and AIRA-Design for low-level mechanistic implementation. AIRA-Compose uses 11 agents to explore fundamental computational primitives under a 24-hour budget. Agents evaluate million-parameter candidates, extrapolating top designs to 350M, 1B, and 3B scales. This yields 14 architectures across two families: AIRAformers (Transformer-based) and AIRAhybrids (Transformer-Mamba). Pr"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2605.15871","kind":"arxiv","version":1},"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/2605.15871/integrity.json","findings":[],"available":true,"detectors_run":[{"name":"ai_meta_artifact","ran_at":"2026-05-19T17:33:48.693543Z","status":"skipped","version":"1.0.0","findings_count":0},{"name":"claim_evidence","ran_at":"2026-05-19T17:01:55.806193Z","status":"completed","version":"1.0.0","findings_count":0}],"snapshot_sha256":"87bb14899f9ee16a01f1381771a771517f12ceced512c71f46e6cd9c8d2d8d1c"},"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":"2605.15871","created_at":"2026-05-20T00:01:22.965997+00:00"},{"alias_kind":"arxiv_version","alias_value":"2605.15871v1","created_at":"2026-05-20T00:01:22.965997+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2605.15871","created_at":"2026-05-20T00:01:22.965997+00:00"},{"alias_kind":"pith_short_12","alias_value":"GREATM223BKA","created_at":"2026-05-20T00:01:22.965997+00:00"},{"alias_kind":"pith_short_16","alias_value":"GREATM223BKA4U5S","created_at":"2026-05-20T00:01:22.965997+00:00"},{"alias_kind":"pith_short_8","alias_value":"GREATM22","created_at":"2026-05-20T00:01:22.965997+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":0,"internal_anchor_count":0,"sample":[]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/GREATM223BKA4U5SON2LFQB4B4","json":"https://pith.science/pith/GREATM223BKA4U5SON2LFQB4B4.json","graph_json":"https://pith.science/api/pith-number/GREATM223BKA4U5SON2LFQB4B4/graph.json","events_json":"https://pith.science/api/pith-number/GREATM223BKA4U5SON2LFQB4B4/events.json","paper":"https://pith.science/paper/GREATM22"},"agent_actions":{"view_html":"https://pith.science/pith/GREATM223BKA4U5SON2LFQB4B4","download_json":"https://pith.science/pith/GREATM223BKA4U5SON2LFQB4B4.json","view_paper":"https://pith.science/paper/GREATM22","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2605.15871&json=true","fetch_graph":"https://pith.science/api/pith-number/GREATM223BKA4U5SON2LFQB4B4/graph.json","fetch_events":"https://pith.science/api/pith-number/GREATM223BKA4U5SON2LFQB4B4/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/GREATM223BKA4U5SON2LFQB4B4/action/timestamp_anchor","attest_storage":"https://pith.science/pith/GREATM223BKA4U5SON2LFQB4B4/action/storage_attestation","attest_author":"https://pith.science/pith/GREATM223BKA4U5SON2LFQB4B4/action/author_attestation","sign_citation":"https://pith.science/pith/GREATM223BKA4U5SON2LFQB4B4/action/citation_signature","submit_replication":"https://pith.science/pith/GREATM223BKA4U5SON2LFQB4B4/action/replication_record"}},"created_at":"2026-05-20T00:01:22.965997+00:00","updated_at":"2026-05-20T00:01:22.965997+00:00"}