{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2024:XTWMO3QFAEF7HUV5UTAI245LVM","short_pith_number":"pith:XTWMO3QF","schema_version":"1.0","canonical_sha256":"bcecc76e05010bf3d2bda4c08d73abab18802fc2f9e930acf2c4a6587a29e8b5","source":{"kind":"arxiv","id":"2406.10260","version":2},"attestation_state":"computed","paper":{"title":"Flextron: Many-in-One Flexible Large Language Model","license":"http://creativecommons.org/publicdomain/zero/1.0/","headline":"","cross_cats":["cs.LG"],"primary_cat":"cs.CL","authors_text":"Greg Heinrich, Hongxu Yin, Jan Kautz, Pavlo Molchanov, Ruisi Cai, Saurav Muralidharan, Zhangyang Wang","submitted_at":"2024-06-11T01:16:10Z","abstract_excerpt":"Training modern LLMs is extremely resource intensive, and customizing them for various deployment scenarios characterized by limited compute and memory resources through repeated training is impractical. In this paper, we introduce Flextron, a network architecture and post-training model optimization framework supporting flexible model deployment. The Flextron architecture utilizes a nested elastic structure to rapidly adapt to specific user-defined latency and accuracy targets during inference with no additional fine-tuning required. It is also input-adaptive, and can automatically route toke"},"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":"2406.10260","kind":"arxiv","version":2},"metadata":{"license":"http://creativecommons.org/publicdomain/zero/1.0/","primary_cat":"cs.CL","submitted_at":"2024-06-11T01:16:10Z","cross_cats_sorted":["cs.LG"],"title_canon_sha256":"af93ba26fe7f86a4be86c82abc63a3d3c7e0ed141ec5a4cc349b8d719e12041b","abstract_canon_sha256":"73c5aca65410c91a15bfcbcb6b77554acd994be03ef18a105403c51ac9b16dd0"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T09:00:05.251124Z","signature_b64":"AKTxEFAFgGxqK297ZqftDvcscMBra+a4oXe12rhoMHfRY/qx0VgmqqaSHSYYS/hSaTBr4dYpH8ITm43x9xyGAg==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"bcecc76e05010bf3d2bda4c08d73abab18802fc2f9e930acf2c4a6587a29e8b5","last_reissued_at":"2026-07-05T09:00:05.250661Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T09:00:05.250661Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Flextron: Many-in-One Flexible Large Language Model","license":"http://creativecommons.org/publicdomain/zero/1.0/","headline":"","cross_cats":["cs.LG"],"primary_cat":"cs.CL","authors_text":"Greg Heinrich, Hongxu Yin, Jan Kautz, Pavlo Molchanov, Ruisi Cai, Saurav Muralidharan, Zhangyang Wang","submitted_at":"2024-06-11T01:16:10Z","abstract_excerpt":"Training modern LLMs is extremely resource intensive, and customizing them for various deployment scenarios characterized by limited compute and memory resources through repeated training is impractical. In this paper, we introduce Flextron, a network architecture and post-training model optimization framework supporting flexible model deployment. The Flextron architecture utilizes a nested elastic structure to rapidly adapt to specific user-defined latency and accuracy targets during inference with no additional fine-tuning required. It is also input-adaptive, and can automatically route toke"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2406.10260","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/2406.10260/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":"2406.10260","created_at":"2026-07-05T09:00:05.250718+00:00"},{"alias_kind":"arxiv_version","alias_value":"2406.10260v2","created_at":"2026-07-05T09:00:05.250718+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2406.10260","created_at":"2026-07-05T09:00:05.250718+00:00"},{"alias_kind":"pith_short_12","alias_value":"XTWMO3QFAEF7","created_at":"2026-07-05T09:00:05.250718+00:00"},{"alias_kind":"pith_short_16","alias_value":"XTWMO3QFAEF7HUV5","created_at":"2026-07-05T09:00:05.250718+00:00"},{"alias_kind":"pith_short_8","alias_value":"XTWMO3QF","created_at":"2026-07-05T09:00:05.250718+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":3,"internal_anchor_count":1,"sample":[{"citing_arxiv_id":"2607.05722","citing_title":"Nemotron-Labs-Diffusion: A Tri-Mode Language Model Unifying Autoregressive, Diffusion, and Self-Speculation Decoding","ref_index":68,"is_internal_anchor":true},{"citing_arxiv_id":"2605.12491","citing_title":"Elastic Attention Cores for Scalable Vision Transformers","ref_index":36,"is_internal_anchor":false},{"citing_arxiv_id":"2605.07182","citing_title":"Star Elastic: Many-in-One Reasoning LLMs with Efficient Budget Control","ref_index":8,"is_internal_anchor":false}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/XTWMO3QFAEF7HUV5UTAI245LVM","json":"https://pith.science/pith/XTWMO3QFAEF7HUV5UTAI245LVM.json","graph_json":"https://pith.science/api/pith-number/XTWMO3QFAEF7HUV5UTAI245LVM/graph.json","events_json":"https://pith.science/api/pith-number/XTWMO3QFAEF7HUV5UTAI245LVM/events.json","paper":"https://pith.science/paper/XTWMO3QF"},"agent_actions":{"view_html":"https://pith.science/pith/XTWMO3QFAEF7HUV5UTAI245LVM","download_json":"https://pith.science/pith/XTWMO3QFAEF7HUV5UTAI245LVM.json","view_paper":"https://pith.science/paper/XTWMO3QF","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2406.10260&json=true","fetch_graph":"https://pith.science/api/pith-number/XTWMO3QFAEF7HUV5UTAI245LVM/graph.json","fetch_events":"https://pith.science/api/pith-number/XTWMO3QFAEF7HUV5UTAI245LVM/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/XTWMO3QFAEF7HUV5UTAI245LVM/action/timestamp_anchor","attest_storage":"https://pith.science/pith/XTWMO3QFAEF7HUV5UTAI245LVM/action/storage_attestation","attest_author":"https://pith.science/pith/XTWMO3QFAEF7HUV5UTAI245LVM/action/author_attestation","sign_citation":"https://pith.science/pith/XTWMO3QFAEF7HUV5UTAI245LVM/action/citation_signature","submit_replication":"https://pith.science/pith/XTWMO3QFAEF7HUV5UTAI245LVM/action/replication_record"}},"created_at":"2026-07-05T09:00:05.250718+00:00","updated_at":"2026-07-05T09:00:05.250718+00:00"}