{"state_type":"pith_open_graph_state","state_version":"1.0","pith_number":"pith:2026:4ZAE2CQ5P2OAQA7JZEBMVDQNF5","merge_version":"pith-open-graph-merge-v1","event_count":2,"valid_event_count":2,"invalid_event_count":0,"equivocation_count":0,"current":{"canonical_record":{"metadata":{"abstract_canon_sha256":"9a236153d0ed19a973f09d5e91d5efffac6932927864cff9ce38cb74bc5e31b8","cross_cats_sorted":["cs.AI"],"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.LG","submitted_at":"2026-05-14T18:03:16Z","title_canon_sha256":"4f1ed45308da2ef4127da20d11a5a70c05ad7e972ddcc4c497973587c1bbc514"},"schema_version":"1.0","source":{"id":"2605.15290","kind":"arxiv","version":1}},"source_aliases":[{"alias_kind":"arxiv","alias_value":"2605.15290","created_at":"2026-05-20T00:00:50Z"},{"alias_kind":"arxiv_version","alias_value":"2605.15290v1","created_at":"2026-05-20T00:00:50Z"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2605.15290","created_at":"2026-05-20T00:00:50Z"},{"alias_kind":"pith_short_12","alias_value":"4ZAE2CQ5P2OA","created_at":"2026-05-20T00:00:50Z"},{"alias_kind":"pith_short_16","alias_value":"4ZAE2CQ5P2OAQA7J","created_at":"2026-05-20T00:00:50Z"},{"alias_kind":"pith_short_8","alias_value":"4ZAE2CQ5","created_at":"2026-05-20T00:00:50Z"}],"graph_snapshots":[{"event_id":"sha256:d025fd83a9c4b5d5b2d80437f4d763539185b42c2ee8b577b87ac37a5022c475","target":"graph","created_at":"2026-05-20T00:00:50Z","signer":{"key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","signer_id":"pith.science","signer_type":"pith_registry"},"payload":{"graph_snapshot":{"author_claims":{"count":0,"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57","strong_count":0},"builder_version":"pith-number-builder-2026-05-17-v1","claims":{"count":4,"items":[{"attestation":"unclaimed","claim_id":"C1","kind":"strongest_claim","source":"verdict.strongest_claim","status":"machine_extracted","text":"We demonstrate the efficacy of our theoretical derivations by showing learning rate transfer across the GQA repetition hyperparameter as well as experiments regarding transfer over weight decay."},{"attestation":"unclaimed","claim_id":"C2","kind":"weakest_assumption","source":"verdict.weakest_assumption","status":"machine_extracted","text":"The modified spectral norm preserves the valid scaling law of network weights when weight matrices are not full rank; this premise is invoked to enable the GQA derivation and is stated as the key technical step after promoting spectral conditions to a definition."},{"attestation":"unclaimed","claim_id":"C3","kind":"one_line_summary","source":"verdict.one_line_summary","status":"machine_extracted","text":"Derives μP scalings for GQA via promoted spectral-norm definition of feature learning and a modified norm preserving scaling laws for non-full-rank matrices, with experiments showing learning-rate transfer."},{"attestation":"unclaimed","claim_id":"C4","kind":"headline","source":"verdict.pith_extraction.headline","status":"machine_extracted","text":"A modified spectral norm for non-full-rank matrices lets maximal update parameterization apply to grouped-query attention."}],"snapshot_sha256":"9b97e1d39f8720e1de5e39e19a35524effc84db90983169c75a8dc14a689d81f"},"formal_canon":{"evidence_count":1,"snapshot_sha256":"a3c123c89a3bcfc1e6766430c4cd2dc48121a8cd654203f36a38c2dc2bd07741"},"integrity":{"available":true,"clean":true,"detectors_run":[{"findings_count":0,"name":"doi_title_agreement","ran_at":"2026-05-19T17:01:18.316486Z","status":"completed","version":"1.0.0"},{"findings_count":0,"name":"doi_compliance","ran_at":"2026-05-19T16:45:40.492995Z","status":"completed","version":"1.0.0"},{"findings_count":0,"name":"claim_evidence","ran_at":"2026-05-19T14:41:54.239751Z","status":"completed","version":"1.0.0"},{"findings_count":0,"name":"ai_meta_artifact","ran_at":"2026-05-19T13:33:22.786996Z","status":"skipped","version":"1.0.0"}],"endpoint":"/pith/2605.15290/integrity.json","findings":[],"snapshot_sha256":"96c1b083a4d1493f805e127ba21ec7c2e55e2f77d7ff72c04c27d138b4c58472","summary":{"advisory":0,"by_detector":{},"critical":0,"informational":0}},"paper":{"abstract_excerpt":"Hyperparameter transfer across model architectures dramatically reduces the amount of compute necessary for tuning large language models (LLMs). The maximal update parameterization ({\\mu}P) ensures transfer through principled mathematical analysis but can be challenging to derive for new model architectures. Building on the spectral feature-learning view of Yang et al. (2023a), we make two advances. First, we promote spectral norm conditions on the weights from a heuristic to the definition of feature learning, and as a consequence arrive at the Complete-P depth and weight-decay scalings witho","authors_text":"Alexander Moreno, Daria Soboleva, Eric Xing, Huijuan Wang, Joel Hestness, Kyle R. Chickering, Mengxi Wu, Muhao Chen, Xuezhe Ma, Zhengzhong Liu","cross_cats":["cs.AI"],"headline":"A modified spectral norm for non-full-rank matrices lets maximal update parameterization apply to grouped-query attention.","license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.LG","submitted_at":"2026-05-14T18:03:16Z","title":"GQA-{\\mu}P: The maximal parameterization update for grouped query attention"},"references":{"count":30,"internal_anchors":8,"resolved_work":30,"sample":[{"cited_arxiv_id":"2305.13245","doi":"","is_internal_anchor":true,"ref_index":1,"title":"GQA: Training Generalized Multi-Query Transformer Models from Multi-Head Checkpoints","work_id":"b73ad5b2-e553-4c71-b0c9-67e67ba7b158","year":null},{"cited_arxiv_id":"","doi":"","is_internal_anchor":false,"ref_index":2,"title":"Why do we need weight decay in modern deep learning? ArXiv, abs/2310.04415","work_id":"c97a49ec-f435-4fcd-812a-ac270ff17004","year":null},{"cited_arxiv_id":"","doi":"","is_internal_anchor":false,"ref_index":3,"title":"Power lines: Scaling laws for weight decay and batch size in llm pre-training","work_id":"e2555467-0d85-4bbb-8e73-ad73ad65dc98","year":null},{"cited_arxiv_id":"","doi":"","is_internal_anchor":false,"ref_index":4,"title":"Cerebras-gpt: Open compute- optimal language models trained on the cerebras wafer- scale cluster","work_id":"146665a8-e124-465b-aa7f-84693124b019","year":null},{"cited_arxiv_id":"","doi":"","is_internal_anchor":false,"ref_index":5,"title":"Don’t be lazy: CompleteP enables compute- efficient deep transformers, January 2026","work_id":"85f11780-ed20-4881-8d31-bb0834b58027","year":null}],"snapshot_sha256":"21846ac245695c053d53eb777e3260b88e0070308a0f70598da4853a8be9574e"},"source":{"id":"2605.15290","kind":"arxiv","version":1},"verdict":{"created_at":"2026-05-19T16:31:30.644410Z","id":"8284b6d5-b2d5-44f0-a2fd-d986c81ac348","model_set":{"reader":"grok-4.3"},"one_line_summary":"Derives μP scalings for GQA via promoted spectral-norm definition of feature learning and a modified norm preserving scaling laws for non-full-rank matrices, with experiments showing learning-rate transfer.","pipeline_version":"pith-pipeline@v0.9.0","pith_extraction_headline":"A modified spectral norm for non-full-rank matrices lets maximal update parameterization apply to grouped-query attention.","strongest_claim":"We demonstrate the efficacy of our theoretical derivations by showing learning rate transfer across the GQA repetition hyperparameter as well as experiments regarding transfer over weight decay.","weakest_assumption":"The modified spectral norm preserves the valid scaling law of network weights when weight matrices are not full rank; this premise is invoked to enable the GQA derivation and is stated as the key technical step after promoting spectral conditions to a definition."}},"verdict_id":"8284b6d5-b2d5-44f0-a2fd-d986c81ac348"}}],"author_attestations":[],"timestamp_anchors":[],"storage_attestations":[],"citation_signatures":[],"replication_records":[],"corrections":[],"mirror_hints":[],"record_created":{"event_id":"sha256:91cdf7c3e3f4fe5b85b269b17451277718330d16c5657a2e53ab844cb76ea952","target":"record","created_at":"2026-05-20T00:00:50Z","signer":{"key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","signer_id":"pith.science","signer_type":"pith_registry"},"payload":{"attestation_state":"computed","canonical_record":{"metadata":{"abstract_canon_sha256":"9a236153d0ed19a973f09d5e91d5efffac6932927864cff9ce38cb74bc5e31b8","cross_cats_sorted":["cs.AI"],"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.LG","submitted_at":"2026-05-14T18:03:16Z","title_canon_sha256":"4f1ed45308da2ef4127da20d11a5a70c05ad7e972ddcc4c497973587c1bbc514"},"schema_version":"1.0","source":{"id":"2605.15290","kind":"arxiv","version":1}},"canonical_sha256":"e6404d0a1d7e9c0803e9c902ca8e0d2f488b932ee2a429751a1c5bc4e859073d","receipt":{"algorithm":"ed25519","builder_version":"pith-number-builder-2026-05-17-v1","canonical_sha256":"e6404d0a1d7e9c0803e9c902ca8e0d2f488b932ee2a429751a1c5bc4e859073d","first_computed_at":"2026-05-20T00:00:50.909862Z","key_id":"pith-v1-2026-05","kind":"pith_receipt","last_reissued_at":"2026-05-20T00:00:50.909862Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","receipt_version":"0.3","signature_b64":"tixJv1DgirHYqWRkH6BceRFB8xadRtwxS8BXCAEJwpjkQaxEn+LCvlbqQ9HKvn/R/QsGJINSIiwrLJCx1aFIDQ==","signature_status":"signed_v1","signed_at":"2026-05-20T00:00:50.910674Z","signed_message":"canonical_sha256_bytes"},"source_id":"2605.15290","source_kind":"arxiv","source_version":1}}},"equivocations":[],"invalid_events":[],"applied_event_ids":["sha256:91cdf7c3e3f4fe5b85b269b17451277718330d16c5657a2e53ab844cb76ea952","sha256:d025fd83a9c4b5d5b2d80437f4d763539185b42c2ee8b577b87ac37a5022c475"],"state_sha256":"dc4b792b86bfedbae791a8883d808819ebdeba7dae9c896bc56eb214f77dc499"}