{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2026:KQGQ3HUFVD5QJXTDZURY2OGC6I","short_pith_number":"pith:KQGQ3HUF","schema_version":"1.0","canonical_sha256":"540d0d9e85a8fb04de63cd238d38c2f2182086c8d712ff6ee20cb40ffaa9e74d","source":{"kind":"arxiv","id":"2604.07663","version":2},"attestation_state":"computed","paper":{"title":"SAGE: Sign-Adaptive Gradient for Memory-Efficient LLM Optimization","license":"http://creativecommons.org/licenses/by/4.0/","headline":"SAGE replaces AdamW in hybrid LLM optimizers by adding a bounded per-dimension scale that tames embedding-layer gradients, cutting memory while raising perplexity performance.","cross_cats":[],"primary_cat":"cs.LG","authors_text":"Hyun-Tae Kim, Wooin Lee","submitted_at":"2026-04-09T00:07:38Z","abstract_excerpt":"The AdamW optimizer, while standard for LLM pretraining, is a critical memory bottleneck, consuming optimizer states equivalent to twice the model's size. Although light-state optimizers like SinkGD attempt to address this issue, we identify the embedding layer dilemma: these methods fail to handle the sparse, high-variance gradients inherent to embeddings, forcing a hybrid design that reverts to AdamW and partially negates the memory gains. We propose SAGE (Sign Adaptive GradiEnt), a novel optimizer that resolves this dilemma by replacing AdamW in this hybrid structure. SAGE combines a Lion-s"},"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":true,"formal_links_present":true},"canonical_record":{"source":{"id":"2604.07663","kind":"arxiv","version":2},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.LG","submitted_at":"2026-04-09T00:07:38Z","cross_cats_sorted":[],"title_canon_sha256":"aa16abe577c9cc3b4f2638308385ac68083532f098b0c3f4f4a2f80d4b295b19","abstract_canon_sha256":"5a8f77f162041ef977b19efbf3b0e41f621029f50f0f7f23836c3f9a532a3e23"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-06-26T01:15:51.382405Z","signature_b64":"YEFqOUzepbftRQ3FGXbPtW4Ax9j3Nh+O8QJweaC/e5JHlrIBeDnpIdAygtVWg7LTAAFaBhh3Y+EKldUvNG+uCA==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"540d0d9e85a8fb04de63cd238d38c2f2182086c8d712ff6ee20cb40ffaa9e74d","last_reissued_at":"2026-06-26T01:15:51.382011Z","signature_status":"signed_v1","first_computed_at":"2026-06-26T01:15:51.382011Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"SAGE: Sign-Adaptive Gradient for Memory-Efficient LLM Optimization","license":"http://creativecommons.org/licenses/by/4.0/","headline":"SAGE replaces AdamW in hybrid LLM optimizers by adding a bounded per-dimension scale that tames embedding-layer gradients, cutting memory while raising perplexity performance.","cross_cats":[],"primary_cat":"cs.LG","authors_text":"Hyun-Tae Kim, Wooin Lee","submitted_at":"2026-04-09T00:07:38Z","abstract_excerpt":"The AdamW optimizer, while standard for LLM pretraining, is a critical memory bottleneck, consuming optimizer states equivalent to twice the model's size. Although light-state optimizers like SinkGD attempt to address this issue, we identify the embedding layer dilemma: these methods fail to handle the sparse, high-variance gradients inherent to embeddings, forcing a hybrid design that reverts to AdamW and partially negates the memory gains. We propose SAGE (Sign Adaptive GradiEnt), a novel optimizer that resolves this dilemma by replacing AdamW in this hybrid structure. SAGE combines a Lion-s"},"claims":{"count":4,"items":[{"kind":"strongest_claim","text":"On Llama models up to 1.3B parameters, our SAGE-based hybrid achieves new state-of-the-art perplexity, outperforming all baselines, including SinkGD hybrid, while significantly reducing optimizer state memory.","source":"verdict.strongest_claim","status":"machine_extracted","claim_id":"C1","attestation":"unclaimed"},{"kind":"weakest_assumption","text":"That the high-variance, sparse gradients in the embedding layer can be sufficiently controlled by a single per-dimension scale provably bounded by 1.0 without reintroducing the instability that forced prior methods back to AdamW.","source":"verdict.weakest_assumption","status":"machine_extracted","claim_id":"C2","attestation":"unclaimed"},{"kind":"one_line_summary","text":"SAGE replaces AdamW in memory-efficient LLM hybrids with a Lion-style sign update plus a provably bounded O(d) adaptive scale, delivering SOTA perplexity on Llama-1.3B while cutting optimizer-state memory.","source":"verdict.one_line_summary","status":"machine_extracted","claim_id":"C3","attestation":"unclaimed"},{"kind":"headline","text":"SAGE replaces AdamW in hybrid LLM optimizers by adding a bounded per-dimension scale that tames embedding-layer gradients, cutting memory while raising perplexity performance.","source":"verdict.pith_extraction.headline","status":"machine_extracted","claim_id":"C4","attestation":"unclaimed"}],"snapshot_sha256":"9959ef4638bbf38b963afe1a2259eda47df1e90fb591bc7e06d3b07a3a633033"},"source":{"id":"2604.07663","kind":"arxiv","version":2},"verdict":{"id":"b0946ce6-2077-4e74-877c-38f3fb057507","model_set":{"reader":"grok-4.3"},"created_at":"2026-05-10T18:33:19.243230Z","strongest_claim":"On Llama models up to 1.3B parameters, our SAGE-based hybrid achieves new state-of-the-art perplexity, outperforming all baselines, including SinkGD hybrid, while significantly reducing optimizer state memory.","one_line_summary":"SAGE replaces AdamW in memory-efficient LLM hybrids with a Lion-style sign update plus a provably bounded O(d) adaptive scale, delivering SOTA perplexity on Llama-1.3B while cutting optimizer-state memory.","pipeline_version":"pith-pipeline@v0.9.0","weakest_assumption":"That the high-variance, sparse gradients in the embedding layer can be sufficiently controlled by a single per-dimension scale provably bounded by 1.0 without reintroducing the instability that forced prior methods back to AdamW.","pith_extraction_headline":"SAGE replaces AdamW in hybrid LLM optimizers by adding a bounded per-dimension scale that tames embedding-layer gradients, cutting memory while raising perplexity performance."},"integrity":{"clean":true,"summary":{"advisory":0,"critical":0,"by_detector":{},"informational":0},"endpoint":"/pith/2604.07663/integrity.json","findings":[],"available":true,"detectors_run":[],"snapshot_sha256":"c28c3603d3b5d939e8dc4c7e95fa8dfce3d595e45f758748cecf8e644a296938"},"references":{"count":3,"sample":[{"doi":"","year":2016,"title":"Layer Normalization","work_id":"20a2d720-0046-4c7c-bcd6-327ec8143f69","ref_index":1,"cited_arxiv_id":"1607.06450","is_internal_anchor":true},{"doi":"","year":2024,"title":"DAtts L ⊙ 1√ h softmax′ gAtt s L√ h !# Y K L , DY K L =","work_id":"fe712d95-0ed1-4d1a-94aa-b36a44a56919","ref_index":2,"cited_arxiv_id":"","is_internal_anchor":false},{"doi":"","year":2048,"title":"Opt.\" refers to optimizer state memory, while","work_id":"68f80759-6f11-4808-8e11-3247f4a03dd9","ref_index":3,"cited_arxiv_id":"","is_internal_anchor":false}],"resolved_work":3,"snapshot_sha256":"2577a907f5d50e501c1dd2a2c777f4dce9b45a3f6abf1b802efbb7184d046c09","internal_anchors":1},"formal_canon":{"evidence_count":1,"snapshot_sha256":"d2e5cd0875624164da59059db0f1aca195c89aacecf5a84d78bc5af301a8de37"},"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":"2604.07663","created_at":"2026-06-26T01:15:51.382064+00:00"},{"alias_kind":"arxiv_version","alias_value":"2604.07663v2","created_at":"2026-06-26T01:15:51.382064+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2604.07663","created_at":"2026-06-26T01:15:51.382064+00:00"},{"alias_kind":"pith_short_12","alias_value":"KQGQ3HUFVD5Q","created_at":"2026-06-26T01:15:51.382064+00:00"},{"alias_kind":"pith_short_16","alias_value":"KQGQ3HUFVD5QJXTD","created_at":"2026-06-26T01:15:51.382064+00:00"},{"alias_kind":"pith_short_8","alias_value":"KQGQ3HUF","created_at":"2026-06-26T01:15:51.382064+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":0,"internal_anchor_count":0,"sample":[]},"formal_canon":{"evidence_count":1,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/KQGQ3HUFVD5QJXTDZURY2OGC6I","json":"https://pith.science/pith/KQGQ3HUFVD5QJXTDZURY2OGC6I.json","graph_json":"https://pith.science/api/pith-number/KQGQ3HUFVD5QJXTDZURY2OGC6I/graph.json","events_json":"https://pith.science/api/pith-number/KQGQ3HUFVD5QJXTDZURY2OGC6I/events.json","paper":"https://pith.science/paper/KQGQ3HUF"},"agent_actions":{"view_html":"https://pith.science/pith/KQGQ3HUFVD5QJXTDZURY2OGC6I","download_json":"https://pith.science/pith/KQGQ3HUFVD5QJXTDZURY2OGC6I.json","view_paper":"https://pith.science/paper/KQGQ3HUF","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2604.07663&json=true","fetch_graph":"https://pith.science/api/pith-number/KQGQ3HUFVD5QJXTDZURY2OGC6I/graph.json","fetch_events":"https://pith.science/api/pith-number/KQGQ3HUFVD5QJXTDZURY2OGC6I/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/KQGQ3HUFVD5QJXTDZURY2OGC6I/action/timestamp_anchor","attest_storage":"https://pith.science/pith/KQGQ3HUFVD5QJXTDZURY2OGC6I/action/storage_attestation","attest_author":"https://pith.science/pith/KQGQ3HUFVD5QJXTDZURY2OGC6I/action/author_attestation","sign_citation":"https://pith.science/pith/KQGQ3HUFVD5QJXTDZURY2OGC6I/action/citation_signature","submit_replication":"https://pith.science/pith/KQGQ3HUFVD5QJXTDZURY2OGC6I/action/replication_record"}},"created_at":"2026-06-26T01:15:51.382064+00:00","updated_at":"2026-06-26T01:15:51.382064+00:00"}