{"state_type":"pith_open_graph_state","state_version":"1.0","pith_number":"pith:2026:CSWEHIKURWZBJ3RNXZXGBXMI46","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":"06889f7640dac651dfb09aedea84e5d876436f42bd8cf88b9f34be3fae857fbb","cross_cats_sorted":["cs.NA","math.NA"],"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.LG","submitted_at":"2026-05-17T14:20:51Z","title_canon_sha256":"6a942f85a6ca1abc8ada0e8a577afedf6d003d37849b7aed7cad297e90443139"},"schema_version":"1.0","source":{"id":"2605.17471","kind":"arxiv","version":1}},"source_aliases":[{"alias_kind":"arxiv","alias_value":"2605.17471","created_at":"2026-05-20T00:04:40Z"},{"alias_kind":"arxiv_version","alias_value":"2605.17471v1","created_at":"2026-05-20T00:04:40Z"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2605.17471","created_at":"2026-05-20T00:04:40Z"},{"alias_kind":"pith_short_12","alias_value":"CSWEHIKURWZB","created_at":"2026-05-20T00:04:40Z"},{"alias_kind":"pith_short_16","alias_value":"CSWEHIKURWZBJ3RN","created_at":"2026-05-20T00:04:40Z"},{"alias_kind":"pith_short_8","alias_value":"CSWEHIKU","created_at":"2026-05-20T00:04:40Z"}],"graph_snapshots":[{"event_id":"sha256:4b8e8e5b30a294283474593da65af6e3f6a7a61c04afe778b637490fdb3886e5","target":"graph","created_at":"2026-05-20T00:04:40Z","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":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"formal_canon":{"evidence_count":0,"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"integrity":{"available":true,"clean":true,"detectors_run":[{"findings_count":0,"name":"claim_evidence","ran_at":"2026-05-19T21:41:57.697233Z","status":"completed","version":"1.0.0"},{"findings_count":0,"name":"ai_meta_artifact","ran_at":"2026-05-19T21:33:23.654040Z","status":"skipped","version":"1.0.0"}],"endpoint":"/pith/2605.17471/integrity.json","findings":[],"snapshot_sha256":"91ecb93a39a4ec77857f2f637d83a5935afed225a46e974cb663fb4ddc55c8cc","summary":{"advisory":0,"by_detector":{},"critical":0,"informational":0}},"paper":{"abstract_excerpt":"Quantization-aware training (QAT) is widely adopted to quantize language models by training full-precision weights using gradients from the quantized model. The main bottleneck is its slow convergence and early performance plateau, particularly below 4-bit-widths. While this problem has been observed in prior work, its precise cause remains unclear. In this paper, we analyze the convergence of QAT by estimating the spectrum of the loss-surface Hessians. We find that the weights converge to flat regions around saddle points, where a large fraction of the Hessian eigenvalues are both positive an","authors_text":"Changsheng Zhao, Dongyue Li, Harshit Khaitan, Hongyang R. Zhang, Kai Yi, Raghuraman Krishnamoorthi, Steven Li, Zechun Liu, Zhenshuo Zhang","cross_cats":["cs.NA","math.NA"],"headline":"","license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.LG","submitted_at":"2026-05-17T14:20:51Z","title":"WinQ: Accelerating Quantization-Aware Training of Language Models Around Saddle Points"},"references":{"count":0,"internal_anchors":0,"resolved_work":0,"sample":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2605.17471","kind":"arxiv","version":1},"verdict":{"created_at":null,"id":null,"model_set":{},"one_line_summary":"","pipeline_version":null,"pith_extraction_headline":"","strongest_claim":"","weakest_assumption":""}},"verdict_id":null}}],"author_attestations":[],"timestamp_anchors":[],"storage_attestations":[],"citation_signatures":[],"replication_records":[],"corrections":[],"mirror_hints":[],"record_created":{"event_id":"sha256:ded85c6f9a23fa85ac0f56e2025d0b4d55032c718a9d698f598dfaac69237514","target":"record","created_at":"2026-05-20T00:04:40Z","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":"06889f7640dac651dfb09aedea84e5d876436f42bd8cf88b9f34be3fae857fbb","cross_cats_sorted":["cs.NA","math.NA"],"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.LG","submitted_at":"2026-05-17T14:20:51Z","title_canon_sha256":"6a942f85a6ca1abc8ada0e8a577afedf6d003d37849b7aed7cad297e90443139"},"schema_version":"1.0","source":{"id":"2605.17471","kind":"arxiv","version":1}},"canonical_sha256":"14ac43a1548db214ee2dbe6e60dd88e797b50293624b6ffb65434b0895808b6c","receipt":{"algorithm":"ed25519","builder_version":"pith-number-builder-2026-05-17-v1","canonical_sha256":"14ac43a1548db214ee2dbe6e60dd88e797b50293624b6ffb65434b0895808b6c","first_computed_at":"2026-05-20T00:04:40.697667Z","key_id":"pith-v1-2026-05","kind":"pith_receipt","last_reissued_at":"2026-05-20T00:04:40.697667Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","receipt_version":"0.3","signature_b64":"1MocaXbtrvxxV5PmhzP4x/s97tECPGtezv0HCXnKXDiIBbjEmoe3CyVi9uD2vnLnBXSCSG/q0609E6RF7JmTCg==","signature_status":"signed_v1","signed_at":"2026-05-20T00:04:40.698441Z","signed_message":"canonical_sha256_bytes"},"source_id":"2605.17471","source_kind":"arxiv","source_version":1}}},"equivocations":[],"invalid_events":[],"applied_event_ids":["sha256:ded85c6f9a23fa85ac0f56e2025d0b4d55032c718a9d698f598dfaac69237514","sha256:4b8e8e5b30a294283474593da65af6e3f6a7a61c04afe778b637490fdb3886e5"],"state_sha256":"87c541dda73bfd221bf216a3daf50718e00eaab69a37ec67f32583fbca543c8e"}