{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2023:AKIELWMENCYHUTMBXGOM3USMRE","short_pith_number":"pith:AKIELWME","schema_version":"1.0","canonical_sha256":"029045d98468b07a4d81b99ccdd24c891ee91a820f7d9c97107a753e1b5cded2","source":{"kind":"arxiv","id":"2312.17244","version":2},"attestation_state":"computed","paper":{"title":"The LLM Surgeon","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.CL"],"primary_cat":"cs.LG","authors_text":"Markus Nagel, Mart van Baalen, Tijmen Blankevoort, Tycho F.A. van der Ouderaa, Yuki M. Asano","submitted_at":"2023-12-28T18:59:09Z","abstract_excerpt":"State-of-the-art language models are becoming increasingly large in an effort to achieve the highest performance on large corpora of available textual data. However, the sheer size of the Transformer architectures makes it difficult to deploy models within computational, environmental or device-specific constraints. We explore data-driven compression of existing pretrained models as an alternative to training smaller models from scratch. To do so, we scale Kronecker-factored curvature approximations of the target loss landscape to large language models. In doing so, we can compute both the dyn"},"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":"2312.17244","kind":"arxiv","version":2},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.LG","submitted_at":"2023-12-28T18:59:09Z","cross_cats_sorted":["cs.CL"],"title_canon_sha256":"11593e2ed29055530401cb197a9cfea16152296949e34557d09c24a1652f1143","abstract_canon_sha256":"e03a444627a2e5bb911e7f6ef311bd3f207e2dbdfbe20de745cc8319837efb68"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T07:58:46.576359Z","signature_b64":"Fgp8jLLPZVZp+12PCET4WXiFv+NJdMH36D+NieDEq0bqueWIk4byAch8xN+uBEuGZwJMJuSozOzj7+glgB5lDg==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"029045d98468b07a4d81b99ccdd24c891ee91a820f7d9c97107a753e1b5cded2","last_reissued_at":"2026-07-05T07:58:46.575809Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T07:58:46.575809Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"The LLM Surgeon","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.CL"],"primary_cat":"cs.LG","authors_text":"Markus Nagel, Mart van Baalen, Tijmen Blankevoort, Tycho F.A. van der Ouderaa, Yuki M. Asano","submitted_at":"2023-12-28T18:59:09Z","abstract_excerpt":"State-of-the-art language models are becoming increasingly large in an effort to achieve the highest performance on large corpora of available textual data. However, the sheer size of the Transformer architectures makes it difficult to deploy models within computational, environmental or device-specific constraints. We explore data-driven compression of existing pretrained models as an alternative to training smaller models from scratch. To do so, we scale Kronecker-factored curvature approximations of the target loss landscape to large language models. In doing so, we can compute both the dyn"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2312.17244","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/2312.17244/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":"2312.17244","created_at":"2026-07-05T07:58:46.575871+00:00"},{"alias_kind":"arxiv_version","alias_value":"2312.17244v2","created_at":"2026-07-05T07:58:46.575871+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2312.17244","created_at":"2026-07-05T07:58:46.575871+00:00"},{"alias_kind":"pith_short_12","alias_value":"AKIELWMENCYH","created_at":"2026-07-05T07:58:46.575871+00:00"},{"alias_kind":"pith_short_16","alias_value":"AKIELWMENCYHUTMB","created_at":"2026-07-05T07:58:46.575871+00:00"},{"alias_kind":"pith_short_8","alias_value":"AKIELWME","created_at":"2026-07-05T07:58:46.575871+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":4,"internal_anchor_count":1,"sample":[{"citing_arxiv_id":"2607.07964","citing_title":"KronQ: LLM Quantization via Kronecker-Factored Hessian","ref_index":35,"is_internal_anchor":true},{"citing_arxiv_id":"2606.01666","citing_title":"DOT-MoE: Differentiable Optimal Transport for MoEfication","ref_index":17,"is_internal_anchor":false},{"citing_arxiv_id":"2606.00573","citing_title":"LASER: Loss-Aware Singular-value Decomposition and Rank Allocation for Efficient Low-Precision Vision-Language Models","ref_index":43,"is_internal_anchor":false},{"citing_arxiv_id":"2508.07407","citing_title":"A Comprehensive Survey of Self-Evolving AI Agents: A New Paradigm Bridging Foundation Models and Lifelong Agentic Systems","ref_index":98,"is_internal_anchor":false}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/AKIELWMENCYHUTMBXGOM3USMRE","json":"https://pith.science/pith/AKIELWMENCYHUTMBXGOM3USMRE.json","graph_json":"https://pith.science/api/pith-number/AKIELWMENCYHUTMBXGOM3USMRE/graph.json","events_json":"https://pith.science/api/pith-number/AKIELWMENCYHUTMBXGOM3USMRE/events.json","paper":"https://pith.science/paper/AKIELWME"},"agent_actions":{"view_html":"https://pith.science/pith/AKIELWMENCYHUTMBXGOM3USMRE","download_json":"https://pith.science/pith/AKIELWMENCYHUTMBXGOM3USMRE.json","view_paper":"https://pith.science/paper/AKIELWME","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2312.17244&json=true","fetch_graph":"https://pith.science/api/pith-number/AKIELWMENCYHUTMBXGOM3USMRE/graph.json","fetch_events":"https://pith.science/api/pith-number/AKIELWMENCYHUTMBXGOM3USMRE/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/AKIELWMENCYHUTMBXGOM3USMRE/action/timestamp_anchor","attest_storage":"https://pith.science/pith/AKIELWMENCYHUTMBXGOM3USMRE/action/storage_attestation","attest_author":"https://pith.science/pith/AKIELWMENCYHUTMBXGOM3USMRE/action/author_attestation","sign_citation":"https://pith.science/pith/AKIELWMENCYHUTMBXGOM3USMRE/action/citation_signature","submit_replication":"https://pith.science/pith/AKIELWMENCYHUTMBXGOM3USMRE/action/replication_record"}},"created_at":"2026-07-05T07:58:46.575871+00:00","updated_at":"2026-07-05T07:58:46.575871+00:00"}