{"bundle_type":"pith_open_graph_bundle","bundle_version":"1.0","pith_number":"pith:2026:NOU336TIEQ5T7FJ7QRUKP4J33J","short_pith_number":"pith:NOU336TI","canonical_record":{"source":{"id":"2606.00888","kind":"arxiv","version":1},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.LG","submitted_at":"2026-05-30T20:47:45Z","cross_cats_sorted":["cs.AI"],"title_canon_sha256":"3fece932d0c94c57e628cc0f00a5552eebceee1945d4e57ccf6f4a6fc4b3a54e","abstract_canon_sha256":"be4c3af4b0d04da59005f6268a708d81445159b6aea169881a7e407678669717"},"schema_version":"1.0"},"canonical_sha256":"6ba9bdfa68243b3f953f8468a7f13bda75bdad055500d503cf34835e9d2a6ce8","source":{"kind":"arxiv","id":"2606.00888","version":1},"source_aliases":[{"alias_kind":"arxiv","alias_value":"2606.00888","created_at":"2026-06-02T01:04:08Z"},{"alias_kind":"arxiv_version","alias_value":"2606.00888v1","created_at":"2026-06-02T01:04:08Z"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2606.00888","created_at":"2026-06-02T01:04:08Z"},{"alias_kind":"pith_short_12","alias_value":"NOU336TIEQ5T","created_at":"2026-06-02T01:04:08Z"},{"alias_kind":"pith_short_16","alias_value":"NOU336TIEQ5T7FJ7","created_at":"2026-06-02T01:04:08Z"},{"alias_kind":"pith_short_8","alias_value":"NOU336TI","created_at":"2026-06-02T01:04:08Z"}],"events":[{"event_type":"record_created","subject_pith_number":"pith:2026:NOU336TIEQ5T7FJ7QRUKP4J33J","target":"record","payload":{"canonical_record":{"source":{"id":"2606.00888","kind":"arxiv","version":1},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.LG","submitted_at":"2026-05-30T20:47:45Z","cross_cats_sorted":["cs.AI"],"title_canon_sha256":"3fece932d0c94c57e628cc0f00a5552eebceee1945d4e57ccf6f4a6fc4b3a54e","abstract_canon_sha256":"be4c3af4b0d04da59005f6268a708d81445159b6aea169881a7e407678669717"},"schema_version":"1.0"},"canonical_sha256":"6ba9bdfa68243b3f953f8468a7f13bda75bdad055500d503cf34835e9d2a6ce8","receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-06-02T01:04:08.877394Z","signature_b64":"rV3NvqGtUA/EpsrEpiebkk7S2v5nlq2iAaxit+pMoH5gl19SnKAON53U4SxkUlVuv8L0IOudjndi3nvWNOBfDQ==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"6ba9bdfa68243b3f953f8468a7f13bda75bdad055500d503cf34835e9d2a6ce8","last_reissued_at":"2026-06-02T01:04:08.876999Z","signature_status":"signed_v1","first_computed_at":"2026-06-02T01:04:08.876999Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"source_kind":"arxiv","source_id":"2606.00888","source_version":1,"attestation_state":"computed"},"signer":{"signer_id":"pith.science","signer_type":"pith_registry","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"created_at":"2026-06-02T01:04:08Z","supersedes":[],"prev_event":null,"signature":{"signature_status":"signed_v1","algorithm":"ed25519","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","signature_b64":"dni/ZFLPaZdPftqhao7D1nriSDzjR0C28rD3+KT+qB28P0oPjECKDt0fO/AF+E6BeIvZeM8AsvlGpi4TRshYBw==","signed_message":"open_graph_event_sha256_bytes","signed_at":"2026-06-02T16:32:24.333291Z"},"content_sha256":"bff8cc57b0892d34e89f72019b8352109dca481de1967825a13220c2d2ceae98","schema_version":"1.0","event_id":"sha256:bff8cc57b0892d34e89f72019b8352109dca481de1967825a13220c2d2ceae98"},{"event_type":"graph_snapshot","subject_pith_number":"pith:2026:NOU336TIEQ5T7FJ7QRUKP4J33J","target":"graph","payload":{"graph_snapshot":{"paper":{"title":"Memory-Efficient LLM Training with Dynamic Sparsity: From Stability to Practical Scaling","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.AI"],"primary_cat":"cs.LG","authors_text":"Boqian Wu, Decebal Constantin Mocanu, Elena Mocanu, Maurice van Keulen, Mykola Pechenizkiy, Patrik Okanovic, Qiao Xiao, Tomasz Sternal, Torsten Hoefler","submitted_at":"2026-05-30T20:47:45Z","abstract_excerpt":"Dynamic Sparse Training (DST) offers a promising paradigm for improving the training and inference efficiency of deep neural networks; however, we find that in large language model training, DST can suffer from optimization instability, manifested as loss spikes after topology updates. In this work, we show that the naive use of standard Adam-based optimizers leads to a cold-start issue for newly regrown parameters, resulting in excessively large updates and disrupted training dynamics. To address this issue, we propose Sparse Memory-Efficient Training (SMET), which stabilizes DST with optimiz"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2606.00888","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/2606.00888/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"},"verdict_id":null},"signer":{"signer_id":"pith.science","signer_type":"pith_registry","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"created_at":"2026-06-02T01:04:08Z","supersedes":[],"prev_event":null,"signature":{"signature_status":"signed_v1","algorithm":"ed25519","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","signature_b64":"ACBQACBln/Ep8uCrY1clt1d+vVNVHzoD/JlgDXvAv/CrWLjYW1zn6WY/yeWDjub/O8l/4OCk796Vm3Y2UeZbBw==","signed_message":"open_graph_event_sha256_bytes","signed_at":"2026-06-02T16:32:24.333694Z"},"content_sha256":"aa2074b514f28b808a182848efa29b83754ad21518ac14fe39f21f8a2ad42a0d","schema_version":"1.0","event_id":"sha256:aa2074b514f28b808a182848efa29b83754ad21518ac14fe39f21f8a2ad42a0d"}],"timestamp_proofs":[],"mirror_hints":[{"mirror_type":"https","name":"Pith Resolver","base_url":"https://pith.science","bundle_url":"https://pith.science/pith/NOU336TIEQ5T7FJ7QRUKP4J33J/bundle.json","state_url":"https://pith.science/pith/NOU336TIEQ5T7FJ7QRUKP4J33J/state.json","well_known_bundle_url":"https://pith.science/.well-known/pith/NOU336TIEQ5T7FJ7QRUKP4J33J/bundle.json","status":"primary"}],"public_keys":[{"key_id":"pith-v1-2026-05","algorithm":"ed25519","format":"raw","public_key_b64":"stVStoiQhXFxp4s2pdzPNoqVNBMojDU/fJ2db5S3CbM=","public_key_hex":"b2d552b68890857171a78b36a5dccf368a953413288c353f7c9d9d6f94b709b3","fingerprint_sha256_b32_first128bits":"RVFV5Z2OI2J3ZUO7ERDEBCYNKS","fingerprint_sha256_hex":"8d4b5ee74e4693bcd1df2446408b0d54","rotates_at":null,"url":"https://pith.science/pith-signing-key.json","notes":"Pith uses this Ed25519 key to sign canonical record SHA-256 digests. Verify with: ed25519_verify(public_key, message=canonical_sha256_bytes, signature=base64decode(signature_b64))."}],"merge_version":"pith-open-graph-merge-v1","built_at":"2026-06-02T16:32:24Z","links":{"resolver":"https://pith.science/pith/NOU336TIEQ5T7FJ7QRUKP4J33J","bundle":"https://pith.science/pith/NOU336TIEQ5T7FJ7QRUKP4J33J/bundle.json","state":"https://pith.science/pith/NOU336TIEQ5T7FJ7QRUKP4J33J/state.json","well_known_bundle":"https://pith.science/.well-known/pith/NOU336TIEQ5T7FJ7QRUKP4J33J/bundle.json"},"state":{"state_type":"pith_open_graph_state","state_version":"1.0","pith_number":"pith:2026:NOU336TIEQ5T7FJ7QRUKP4J33J","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":"be4c3af4b0d04da59005f6268a708d81445159b6aea169881a7e407678669717","cross_cats_sorted":["cs.AI"],"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.LG","submitted_at":"2026-05-30T20:47:45Z","title_canon_sha256":"3fece932d0c94c57e628cc0f00a5552eebceee1945d4e57ccf6f4a6fc4b3a54e"},"schema_version":"1.0","source":{"id":"2606.00888","kind":"arxiv","version":1}},"source_aliases":[{"alias_kind":"arxiv","alias_value":"2606.00888","created_at":"2026-06-02T01:04:08Z"},{"alias_kind":"arxiv_version","alias_value":"2606.00888v1","created_at":"2026-06-02T01:04:08Z"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2606.00888","created_at":"2026-06-02T01:04:08Z"},{"alias_kind":"pith_short_12","alias_value":"NOU336TIEQ5T","created_at":"2026-06-02T01:04:08Z"},{"alias_kind":"pith_short_16","alias_value":"NOU336TIEQ5T7FJ7","created_at":"2026-06-02T01:04:08Z"},{"alias_kind":"pith_short_8","alias_value":"NOU336TI","created_at":"2026-06-02T01:04:08Z"}],"graph_snapshots":[{"event_id":"sha256:aa2074b514f28b808a182848efa29b83754ad21518ac14fe39f21f8a2ad42a0d","target":"graph","created_at":"2026-06-02T01:04:08Z","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":[],"endpoint":"/pith/2606.00888/integrity.json","findings":[],"snapshot_sha256":"c28c3603d3b5d939e8dc4c7e95fa8dfce3d595e45f758748cecf8e644a296938","summary":{"advisory":0,"by_detector":{},"critical":0,"informational":0}},"paper":{"abstract_excerpt":"Dynamic Sparse Training (DST) offers a promising paradigm for improving the training and inference efficiency of deep neural networks; however, we find that in large language model training, DST can suffer from optimization instability, manifested as loss spikes after topology updates. In this work, we show that the naive use of standard Adam-based optimizers leads to a cold-start issue for newly regrown parameters, resulting in excessively large updates and disrupted training dynamics. To address this issue, we propose Sparse Memory-Efficient Training (SMET), which stabilizes DST with optimiz","authors_text":"Boqian Wu, Decebal Constantin Mocanu, Elena Mocanu, Maurice van Keulen, Mykola Pechenizkiy, Patrik Okanovic, Qiao Xiao, Tomasz Sternal, Torsten Hoefler","cross_cats":["cs.AI"],"headline":"","license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.LG","submitted_at":"2026-05-30T20:47:45Z","title":"Memory-Efficient LLM Training with Dynamic Sparsity: From Stability to Practical Scaling"},"references":{"count":0,"internal_anchors":0,"resolved_work":0,"sample":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2606.00888","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:bff8cc57b0892d34e89f72019b8352109dca481de1967825a13220c2d2ceae98","target":"record","created_at":"2026-06-02T01:04:08Z","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":"be4c3af4b0d04da59005f6268a708d81445159b6aea169881a7e407678669717","cross_cats_sorted":["cs.AI"],"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.LG","submitted_at":"2026-05-30T20:47:45Z","title_canon_sha256":"3fece932d0c94c57e628cc0f00a5552eebceee1945d4e57ccf6f4a6fc4b3a54e"},"schema_version":"1.0","source":{"id":"2606.00888","kind":"arxiv","version":1}},"canonical_sha256":"6ba9bdfa68243b3f953f8468a7f13bda75bdad055500d503cf34835e9d2a6ce8","receipt":{"algorithm":"ed25519","builder_version":"pith-number-builder-2026-05-17-v1","canonical_sha256":"6ba9bdfa68243b3f953f8468a7f13bda75bdad055500d503cf34835e9d2a6ce8","first_computed_at":"2026-06-02T01:04:08.876999Z","key_id":"pith-v1-2026-05","kind":"pith_receipt","last_reissued_at":"2026-06-02T01:04:08.876999Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","receipt_version":"0.3","signature_b64":"rV3NvqGtUA/EpsrEpiebkk7S2v5nlq2iAaxit+pMoH5gl19SnKAON53U4SxkUlVuv8L0IOudjndi3nvWNOBfDQ==","signature_status":"signed_v1","signed_at":"2026-06-02T01:04:08.877394Z","signed_message":"canonical_sha256_bytes"},"source_id":"2606.00888","source_kind":"arxiv","source_version":1}}},"equivocations":[],"invalid_events":[],"applied_event_ids":["sha256:bff8cc57b0892d34e89f72019b8352109dca481de1967825a13220c2d2ceae98","sha256:aa2074b514f28b808a182848efa29b83754ad21518ac14fe39f21f8a2ad42a0d"],"state_sha256":"78ee485071aa4eb87c01cf898d91ffee83c36f87ff0a028db24e258a72fc5d00"},"bundle_signature":{"signature_status":"signed_v1","algorithm":"ed25519","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","signature_b64":"xOJq2l9MwgAhInssff9gEJg+izc+U0LxXxeCwm0KCXxUdFgJyfPjoj8a8vyh2QVljXcmHmfxvxs0YPvquMsHDg==","signed_message":"bundle_sha256_bytes","signed_at":"2026-06-02T16:32:24.335838Z","bundle_sha256":"569d11fd8f7b2f52863be954ea71b060517608140ead8a90ba304f6f64a5b498"}}