{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2026:W6JTPEFJ3J6U2XHHV2GSNDRVOU","short_pith_number":"pith:W6JTPEFJ","schema_version":"1.0","canonical_sha256":"b7933790a9da7d4d5ce7ae8d268e35750c297a97a6207b759461042a6ea0f112","source":{"kind":"arxiv","id":"2605.21803","version":1},"attestation_state":"computed","paper":{"title":"Same Architecture, Different Capacity: Optimizer-Induced Spectral Scaling Laws","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":[],"primary_cat":"cs.LG","authors_text":"Brandon Reagen, Nandan Kumar Jha","submitted_at":"2026-05-20T23:00:34Z","abstract_excerpt":"Scaling laws have made language-model performance predictable from model size, data, and compute, but they typically treat the optimizer as a fixed training detail. We show that this assumption misses a fundamental axis of representation scaling: how effectively the optimizer converts added FFN width into utilized spectral capacity. Using eigenspectra of feed-forward network representations, measured through soft and hard spectral-ranks, we find that \\emph{the same Transformer architecture realizes markedly different spectral scaling laws when trained with different optimizers}. Holding archit"},"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":"2605.21803","kind":"arxiv","version":1},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.LG","submitted_at":"2026-05-20T23:00:34Z","cross_cats_sorted":[],"title_canon_sha256":"b9b2cde9cb9354eda187880de97b4015e0c307f25da8e37d1f7ad889e4d6297d","abstract_canon_sha256":"a312ce39f7ad166de44e2162e7c2aad4b8ed8c7624f2ef0c1f2a232a89c163dd"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-05-22T01:03:33.150831Z","signature_b64":"sHkHXorEU15nKvQWE4DqpoSsZ2QgZ3ZKD7OcJ6WYccCjdPF4Q1kkvQgy3JVOrKRTlw/KSHPLli/A6xhwXLlqBQ==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"b7933790a9da7d4d5ce7ae8d268e35750c297a97a6207b759461042a6ea0f112","last_reissued_at":"2026-05-22T01:03:33.150309Z","signature_status":"signed_v1","first_computed_at":"2026-05-22T01:03:33.150309Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Same Architecture, Different Capacity: Optimizer-Induced Spectral Scaling Laws","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":[],"primary_cat":"cs.LG","authors_text":"Brandon Reagen, Nandan Kumar Jha","submitted_at":"2026-05-20T23:00:34Z","abstract_excerpt":"Scaling laws have made language-model performance predictable from model size, data, and compute, but they typically treat the optimizer as a fixed training detail. We show that this assumption misses a fundamental axis of representation scaling: how effectively the optimizer converts added FFN width into utilized spectral capacity. Using eigenspectra of feed-forward network representations, measured through soft and hard spectral-ranks, we find that \\emph{the same Transformer architecture realizes markedly different spectral scaling laws when trained with different optimizers}. Holding archit"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2605.21803","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/2605.21803/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":"2605.21803","created_at":"2026-05-22T01:03:33.150402+00:00"},{"alias_kind":"arxiv_version","alias_value":"2605.21803v1","created_at":"2026-05-22T01:03:33.150402+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2605.21803","created_at":"2026-05-22T01:03:33.150402+00:00"},{"alias_kind":"pith_short_12","alias_value":"W6JTPEFJ3J6U","created_at":"2026-05-22T01:03:33.150402+00:00"},{"alias_kind":"pith_short_16","alias_value":"W6JTPEFJ3J6U2XHH","created_at":"2026-05-22T01:03:33.150402+00:00"},{"alias_kind":"pith_short_8","alias_value":"W6JTPEFJ","created_at":"2026-05-22T01:03:33.150402+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/W6JTPEFJ3J6U2XHHV2GSNDRVOU","json":"https://pith.science/pith/W6JTPEFJ3J6U2XHHV2GSNDRVOU.json","graph_json":"https://pith.science/api/pith-number/W6JTPEFJ3J6U2XHHV2GSNDRVOU/graph.json","events_json":"https://pith.science/api/pith-number/W6JTPEFJ3J6U2XHHV2GSNDRVOU/events.json","paper":"https://pith.science/paper/W6JTPEFJ"},"agent_actions":{"view_html":"https://pith.science/pith/W6JTPEFJ3J6U2XHHV2GSNDRVOU","download_json":"https://pith.science/pith/W6JTPEFJ3J6U2XHHV2GSNDRVOU.json","view_paper":"https://pith.science/paper/W6JTPEFJ","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2605.21803&json=true","fetch_graph":"https://pith.science/api/pith-number/W6JTPEFJ3J6U2XHHV2GSNDRVOU/graph.json","fetch_events":"https://pith.science/api/pith-number/W6JTPEFJ3J6U2XHHV2GSNDRVOU/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/W6JTPEFJ3J6U2XHHV2GSNDRVOU/action/timestamp_anchor","attest_storage":"https://pith.science/pith/W6JTPEFJ3J6U2XHHV2GSNDRVOU/action/storage_attestation","attest_author":"https://pith.science/pith/W6JTPEFJ3J6U2XHHV2GSNDRVOU/action/author_attestation","sign_citation":"https://pith.science/pith/W6JTPEFJ3J6U2XHHV2GSNDRVOU/action/citation_signature","submit_replication":"https://pith.science/pith/W6JTPEFJ3J6U2XHHV2GSNDRVOU/action/replication_record"}},"created_at":"2026-05-22T01:03:33.150402+00:00","updated_at":"2026-05-22T01:03:33.150402+00:00"}