{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2025:F7Z7MEBYHGO5TVFPEQSG7WKZZW","short_pith_number":"pith:F7Z7MEBY","schema_version":"1.0","canonical_sha256":"2ff3f61038399dd9d4af24246fd959cd99e16281f0f72519b83f384c18e127a0","source":{"kind":"arxiv","id":"2507.22448","version":1},"attestation_state":"computed","paper":{"title":"Falcon-H1: A Family of Hybrid-Head Language Models Redefining Efficiency and Performance","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":[],"primary_cat":"cs.CL","authors_text":"Abdalgader Abubaker, Billel Mokeddem, Brahim Farhat, Dhia Eddine Rhayem, Giulia Campesan, Guillaume Kunsch, Hakim Hacid, Hamza Yous, Ibrahim Khadraoui, Iheb Chaabane, Ilyas Chahed, Jingwei Zuo, Kacper Piskorski, Leen AlQadi, Maksim Velikanov, Mikhail Lubinets, Mohamed Chami, Mohamed El Amine Seddik, Mugariya Farooq, Ngoc Dung Huynh, Phuc Le Khac, Puneesh Khanna, Ruxandra Cojocaru, Shi Hu, Slim Frikha, Yasser Djilali, Younes Belkada","submitted_at":"2025-07-30T07:55:33Z","abstract_excerpt":"In this report, we introduce Falcon-H1, a new series of large language models (LLMs) featuring hybrid architecture designs optimized for both high performance and efficiency across diverse use cases. Unlike earlier Falcon models built solely on Transformer or Mamba architectures, Falcon-H1 adopts a parallel hybrid approach that combines Transformer-based attention with State Space Models (SSMs), known for superior long-context memory and computational efficiency. We systematically revisited model design, data strategy, and training dynamics, challenging conventional practices in the field. Fal"},"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":"2507.22448","kind":"arxiv","version":1},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.CL","submitted_at":"2025-07-30T07:55:33Z","cross_cats_sorted":[],"title_canon_sha256":"9d52a55bb6b4b305aecc3209497c1911866c7522a873af541357322b7b53e8ee","abstract_canon_sha256":"7278333d3c97637ebffebfbb1af54a5f1735f4458544d7be9ada482cd994e56a"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T11:45:42.505600Z","signature_b64":"+RLZV+erTJCDY5/DHn8n/KBKRLsXGuCDlQdZz9NgBoLNUIqndw8AjIbvThZK0ulvKrURbG0iv6yGPewCrLdoAw==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"2ff3f61038399dd9d4af24246fd959cd99e16281f0f72519b83f384c18e127a0","last_reissued_at":"2026-07-05T11:45:42.505157Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T11:45:42.505157Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Falcon-H1: A Family of Hybrid-Head Language Models Redefining Efficiency and Performance","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":[],"primary_cat":"cs.CL","authors_text":"Abdalgader Abubaker, Billel Mokeddem, Brahim Farhat, Dhia Eddine Rhayem, Giulia Campesan, Guillaume Kunsch, Hakim Hacid, Hamza Yous, Ibrahim Khadraoui, Iheb Chaabane, Ilyas Chahed, Jingwei Zuo, Kacper Piskorski, Leen AlQadi, Maksim Velikanov, Mikhail Lubinets, Mohamed Chami, Mohamed El Amine Seddik, Mugariya Farooq, Ngoc Dung Huynh, Phuc Le Khac, Puneesh Khanna, Ruxandra Cojocaru, Shi Hu, Slim Frikha, Yasser Djilali, Younes Belkada","submitted_at":"2025-07-30T07:55:33Z","abstract_excerpt":"In this report, we introduce Falcon-H1, a new series of large language models (LLMs) featuring hybrid architecture designs optimized for both high performance and efficiency across diverse use cases. Unlike earlier Falcon models built solely on Transformer or Mamba architectures, Falcon-H1 adopts a parallel hybrid approach that combines Transformer-based attention with State Space Models (SSMs), known for superior long-context memory and computational efficiency. We systematically revisited model design, data strategy, and training dynamics, challenging conventional practices in the field. Fal"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2507.22448","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/2507.22448/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":"2507.22448","created_at":"2026-07-05T11:45:42.505218+00:00"},{"alias_kind":"arxiv_version","alias_value":"2507.22448v1","created_at":"2026-07-05T11:45:42.505218+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2507.22448","created_at":"2026-07-05T11:45:42.505218+00:00"},{"alias_kind":"pith_short_12","alias_value":"F7Z7MEBYHGO5","created_at":"2026-07-05T11:45:42.505218+00:00"},{"alias_kind":"pith_short_16","alias_value":"F7Z7MEBYHGO5TVFP","created_at":"2026-07-05T11:45:42.505218+00:00"},{"alias_kind":"pith_short_8","alias_value":"F7Z7MEBY","created_at":"2026-07-05T11:45:42.505218+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":12,"internal_anchor_count":0,"sample":[{"citing_arxiv_id":"2606.03014","citing_title":"MOSAIC: Efficient Mixture-of-Agent Scheduling via Adaptive Aggregation and Inference Concurrency","ref_index":8,"is_internal_anchor":false},{"citing_arxiv_id":"2606.02332","citing_title":"Forget Attention: Importance-Aware Attention Is All You Need","ref_index":8,"is_internal_anchor":false},{"citing_arxiv_id":"2606.30562","citing_title":"Morphing into Hybrid Attention Models","ref_index":72,"is_internal_anchor":false},{"citing_arxiv_id":"2605.17653","citing_title":"LLMForge: Multi-Backend Hardware-Aware Neural Architecture Search with Infinite-Head Attention for Edge Language Models","ref_index":39,"is_internal_anchor":false},{"citing_arxiv_id":"2509.05276","citing_title":"SpikingBrain: Spiking Brain-inspired Large Models","ref_index":45,"is_internal_anchor":false},{"citing_arxiv_id":"2510.20505","citing_title":"RELOOP: Recursive Retrieval with Multi-Hop Reasoner and Planners for Heterogeneous QA","ref_index":35,"is_internal_anchor":false},{"citing_arxiv_id":"2510.26692","citing_title":"Kimi Linear: An Expressive, Efficient Attention Architecture","ref_index":129,"is_internal_anchor":false},{"citing_arxiv_id":"2604.01168","citing_title":"S0 Tuning: Zero-Overhead Adaptation of Hybrid Recurrent-Attention Models","ref_index":4,"is_internal_anchor":false},{"citing_arxiv_id":"2605.05662","citing_title":"XL-SafetyBench: A Country-Grounded Cross-Cultural Benchmark for LLM Safety and Cultural Sensitivity","ref_index":60,"is_internal_anchor":false},{"citing_arxiv_id":"2604.22575","citing_title":"SpikingBrain2.0: Brain-Inspired Foundation Models for Efficient Long-Context and Cross-Platform Inference","ref_index":41,"is_internal_anchor":false},{"citing_arxiv_id":"2605.01106","citing_title":"Component-Aware Self-Speculative Decoding in Hybrid Language Models","ref_index":11,"is_internal_anchor":false},{"citing_arxiv_id":"2604.19877","citing_title":"Super Apriel: One Checkpoint, Many Speeds","ref_index":72,"is_internal_anchor":false}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/F7Z7MEBYHGO5TVFPEQSG7WKZZW","json":"https://pith.science/pith/F7Z7MEBYHGO5TVFPEQSG7WKZZW.json","graph_json":"https://pith.science/api/pith-number/F7Z7MEBYHGO5TVFPEQSG7WKZZW/graph.json","events_json":"https://pith.science/api/pith-number/F7Z7MEBYHGO5TVFPEQSG7WKZZW/events.json","paper":"https://pith.science/paper/F7Z7MEBY"},"agent_actions":{"view_html":"https://pith.science/pith/F7Z7MEBYHGO5TVFPEQSG7WKZZW","download_json":"https://pith.science/pith/F7Z7MEBYHGO5TVFPEQSG7WKZZW.json","view_paper":"https://pith.science/paper/F7Z7MEBY","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2507.22448&json=true","fetch_graph":"https://pith.science/api/pith-number/F7Z7MEBYHGO5TVFPEQSG7WKZZW/graph.json","fetch_events":"https://pith.science/api/pith-number/F7Z7MEBYHGO5TVFPEQSG7WKZZW/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/F7Z7MEBYHGO5TVFPEQSG7WKZZW/action/timestamp_anchor","attest_storage":"https://pith.science/pith/F7Z7MEBYHGO5TVFPEQSG7WKZZW/action/storage_attestation","attest_author":"https://pith.science/pith/F7Z7MEBYHGO5TVFPEQSG7WKZZW/action/author_attestation","sign_citation":"https://pith.science/pith/F7Z7MEBYHGO5TVFPEQSG7WKZZW/action/citation_signature","submit_replication":"https://pith.science/pith/F7Z7MEBYHGO5TVFPEQSG7WKZZW/action/replication_record"}},"created_at":"2026-07-05T11:45:42.505218+00:00","updated_at":"2026-07-05T11:45:42.505218+00:00"}