{"state_type":"pith_open_graph_state","state_version":"1.0","pith_number":"pith:2026:QMCAWIL5AXFP4Y6GTLMEHQUJT2","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":"84a5948012050bc97e76c01cda4ff86e78a223497cc5590b26437950090b5b33","cross_cats_sorted":[],"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.LG","submitted_at":"2026-04-10T12:17:18Z","title_canon_sha256":"f59b06e55ceb072b5bc53af46f96114d78f57fc9ea1d72a33a834856408e498c"},"schema_version":"1.0","source":{"id":"2604.09258","kind":"arxiv","version":2}},"source_aliases":[{"alias_kind":"arxiv","alias_value":"2604.09258","created_at":"2026-05-28T01:04:39Z"},{"alias_kind":"arxiv_version","alias_value":"2604.09258v2","created_at":"2026-05-28T01:04:39Z"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2604.09258","created_at":"2026-05-28T01:04:39Z"},{"alias_kind":"pith_short_12","alias_value":"QMCAWIL5AXFP","created_at":"2026-05-28T01:04:39Z"},{"alias_kind":"pith_short_16","alias_value":"QMCAWIL5AXFP4Y6G","created_at":"2026-05-28T01:04:39Z"},{"alias_kind":"pith_short_8","alias_value":"QMCAWIL5","created_at":"2026-05-28T01:04:39Z"}],"graph_snapshots":[{"event_id":"sha256:5853c9e12ae29a6bf0231a06fba46dce6e52932bf425633a9f29bf0f25be30fd","target":"graph","created_at":"2026-05-28T01:04:39Z","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":"Nexus significantly boosts downstream performance, despite achieving the same pretraining loss. Notably, on the 3B model, Nexus reduces the out-of-distribution loss by 0.012 and yields up to a 15.0% accuracy improvement on complex reasoning tasks (e.g., GSM8k)."},{"attestation":"unclaimed","claim_id":"C2","kind":"weakest_assumption","source":"verdict.weakest_assumption","status":"machine_extracted","text":"The geometric closeness of task-specific minima is intrinsically linked to downstream generalization, and that maximizing gradient similarity during optimization produces this closeness."},{"attestation":"unclaimed","claim_id":"C3","kind":"one_line_summary","source":"verdict.one_line_summary","status":"machine_extracted","text":"Nexus optimizer improves LLM downstream performance by converging to common minima across data sources despite identical pretraining loss."},{"attestation":"unclaimed","claim_id":"C4","kind":"headline","source":"verdict.pith_extraction.headline","status":"machine_extracted","text":"Converging to common minima across data sources during pretraining improves downstream generalization even at identical loss values."}],"snapshot_sha256":"a584887b95800e03e044bfd266dd9f91be2c62b2cc57f44f5fd665782b17d358"},"formal_canon":{"evidence_count":2,"snapshot_sha256":"061188b7996e67bdbad99c3626f52832ce173c1e1b7cc37dd2c29826c484c4aa"},"integrity":{"available":true,"clean":true,"detectors_run":[],"endpoint":"/pith/2604.09258/integrity.json","findings":[],"snapshot_sha256":"c28c3603d3b5d939e8dc4c7e95fa8dfce3d595e45f758748cecf8e644a296938","summary":{"advisory":0,"by_detector":{},"critical":0,"informational":0}},"paper":{"abstract_excerpt":"The foundational capabilities of large language models are acquired during pretraining on internet-scale, highly heterogeneous data mixtures. In this work, we investigate an interesting geometric question regarding the converged state of pretraining: Does the model converge to a common minimizer across all data sources (e.g., \\cref{fig:cwa_illustration:close}), or merely a minimizer of the summed loss (e.g., \\cref{fig:cwa_illustration:distant})? We hypothesize that the geometric \"closeness\" of task-specific minima is intrinsically linked to downstream generalization. We reveal that standard op","authors_text":"Huanran Chen, Huaqing Zhang, Jun Zhu, Ke Shen, Xiao Li, Yinpeng Dong","cross_cats":[],"headline":"Converging to common minima across data sources during pretraining improves downstream generalization even at identical loss values.","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.LG","submitted_at":"2026-04-10T12:17:18Z","title":"Nexus: Same Pretraining Loss, Better Downstream Generalization via Common Minima"},"references":{"count":0,"internal_anchors":0,"resolved_work":0,"sample":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2604.09258","kind":"arxiv","version":2},"verdict":{"created_at":"2026-05-10T17:50:26.241619Z","id":"d3ba602d-47e6-4b65-9d4a-d0b8d58839f6","model_set":{"reader":"grok-4.3"},"one_line_summary":"Nexus optimizer improves LLM downstream performance by converging to common minima across data sources despite identical pretraining loss.","pipeline_version":"pith-pipeline@v0.9.0","pith_extraction_headline":"Converging to common minima across data sources during pretraining improves downstream generalization even at identical loss values.","strongest_claim":"Nexus significantly boosts downstream performance, despite achieving the same pretraining loss. Notably, on the 3B model, Nexus reduces the out-of-distribution loss by 0.012 and yields up to a 15.0% accuracy improvement on complex reasoning tasks (e.g., GSM8k).","weakest_assumption":"The geometric closeness of task-specific minima is intrinsically linked to downstream generalization, and that maximizing gradient similarity during optimization produces this closeness."}},"verdict_id":"d3ba602d-47e6-4b65-9d4a-d0b8d58839f6"}}],"author_attestations":[],"timestamp_anchors":[],"storage_attestations":[],"citation_signatures":[],"replication_records":[],"corrections":[],"mirror_hints":[],"record_created":{"event_id":"sha256:d40d8c651772bd2bacf28e008bdf30162947f58d5d43cff6ab1d51fb23c5682c","target":"record","created_at":"2026-05-28T01:04:39Z","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":"84a5948012050bc97e76c01cda4ff86e78a223497cc5590b26437950090b5b33","cross_cats_sorted":[],"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.LG","submitted_at":"2026-04-10T12:17:18Z","title_canon_sha256":"f59b06e55ceb072b5bc53af46f96114d78f57fc9ea1d72a33a834856408e498c"},"schema_version":"1.0","source":{"id":"2604.09258","kind":"arxiv","version":2}},"canonical_sha256":"83040b217d05cafe63c69ad843c2899e9fff1b2b8c52e61e1b0ba5ea911e6b9a","receipt":{"algorithm":"ed25519","builder_version":"pith-number-builder-2026-05-17-v1","canonical_sha256":"83040b217d05cafe63c69ad843c2899e9fff1b2b8c52e61e1b0ba5ea911e6b9a","first_computed_at":"2026-05-28T01:04:39.854864Z","key_id":"pith-v1-2026-05","kind":"pith_receipt","last_reissued_at":"2026-05-28T01:04:39.854864Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","receipt_version":"0.3","signature_b64":"I0RiaycFY+HivsYTP/WyrRRXj6I0mkCpUSXQDSjvcOFrtW2BNmPqM9cPPMdGFz/1pMcKHwRVYN20gPIYAKhFBA==","signature_status":"signed_v1","signed_at":"2026-05-28T01:04:39.855575Z","signed_message":"canonical_sha256_bytes"},"source_id":"2604.09258","source_kind":"arxiv","source_version":2}}},"equivocations":[],"invalid_events":[],"applied_event_ids":["sha256:d40d8c651772bd2bacf28e008bdf30162947f58d5d43cff6ab1d51fb23c5682c","sha256:5853c9e12ae29a6bf0231a06fba46dce6e52932bf425633a9f29bf0f25be30fd"],"state_sha256":"5cdc8fa00b02878da29bf7ce74db029dc2e5fcb9224f7a26137f4c9efbb8ed11"}