{"as_of":"2026-08-23T01:06:00Z","caps":{"database_statements":6,"inbound":100,"outbound":100},"context_digest":"sha256:238fcdebd3cf2954488b8644dc3b9509b1de553fd889ab8f0b907763b553ae01","coverage":[{"denominator":31,"lane":"reference_resolution","note":"Typed states for the displayed outbound observations.","records_observed":31,"source":"paper_references, paper_reference_links","source_observed_at":"2026-08-12T00:46:33.917982Z","state":"measured"},{"denominator":31,"lane":"standing_notices","note":"One-hop event checks from named stored sources.","records_observed":31,"source":"scholarly_work_events, retraction_status_cache","source_observed_at":"2026-08-22T06:32:14.747728+00:00","state":"measured"},{"denominator":0,"lane":"inbound_itemization","note":"Pith citing papers itemized under the disclosed page cap.","records_observed":0,"source":"paper_references, paper_reference_links","source_observed_at":null,"state":"measured"},{"denominator":1,"lane":"external_citation_measurements","note":"A source-named dated measurement, never combined with another source.","records_observed":0,"source":"cited_works","source_observed_at":null,"state":"measured"}],"external_citation_measurements":[],"inbound":[],"links":{"evidence":"/evidence","html":"/paper/2608.07974/citation-record","integrity":"/paper/2608.07974/integrity","json":"/paper/2608.07974/citation-record.json","paper":"/paper/2608.07974"},"outbound":[{"citation":{"cited_paper":null,"cited_work":{"arxiv_id":null,"doi":null,"metadata_source":null,"pith_arxiv_id":null,"snapshot_observed_at":"2026-08-12T00:46:33.770748Z","title":"End-to-end test-time training for long context,","venue":null,"work_id":null,"year":2025},"citing_paper":{"arxiv_id":"2608.07974","last_updated":"2026-08-08T07:19:21Z","snapshot_observed_at":"2026-08-18T00:09:50.943155Z","submitted_at":"2026-08-08T07:19:21Z","title":"ZeroLock: Concurrent Memory-Efficient LLM Training via Modular Update Decoupling","version":1},"reference_index":1,"source":"pdf_text","source_observed_at":"2026-08-12T00:46:33.770748Z"},"links":{"citing_paper":"/paper/2608.07974"},"observation_digest":"sha256:cbcce22e15a33cdd21ac1a2d1b013006dd42c8b7785c96689c3b682326153873","observation_id":"5a7863c9-3613-4aae-b2ad-00b46b11bbab","resolution":{"observed_at":"2026-08-12T00:46:33.770748Z","resolver_source":null,"status":"unresolved"},"standing_notice":{"events":[],"reason":"canonical_work_link_unavailable","source_receipts":[],"state":"unavailable"}},{"citation":{"cited_paper":null,"cited_work":{"arxiv_id":null,"doi":null,"metadata_source":"raw_reference","pith_arxiv_id":null,"snapshot_observed_at":"2026-08-12T00:46:34.668124Z","title":"Cross-subject eeg signals-based emotion recognition using contrastive learning,","venue":null,"work_id":"9f29bdee-1fb2-439f-adf5-218c8550579c","year":2025},"citing_paper":{"arxiv_id":"2608.07974","last_updated":"2026-08-08T07:19:21Z","snapshot_observed_at":"2026-08-18T00:09:50.943155Z","submitted_at":"2026-08-08T07:19:21Z","title":"ZeroLock: Concurrent Memory-Efficient LLM Training via Modular Update Decoupling","version":1},"reference_index":2,"source":"pdf_text","source_observed_at":"2026-08-12T00:46:33.776013Z"},"links":{"citing_paper":"/paper/2608.07974"},"observation_digest":"sha256:59db8bb5e168627412ddfdb62277b9c4560efb2a4158e14af55b2f8cedeb3d77","observation_id":"7f679fd1-231b-4403-bd6d-15d17782c553","resolution":{"observed_at":"2026-08-12T00:46:34.672962Z","resolver_source":"raw_fallback","status":"verified_fuzzy"},"standing_notice":{"events":[],"observation":"No event found in the named queried sources as of 2026-08-22T06:32:14.747728+00:00.","reason":null,"source_receipts":[{"observed_at":"2026-08-22T06:32:14.747728+00:00","source":"crossref"},{"observed_at":"2026-08-22T06:32:06.552537+00:00","source":"retraction_watch"}],"state":"measured"}},{"citation":{"cited_paper":null,"cited_work":{"arxiv_id":null,"doi":null,"metadata_source":"raw_reference","pith_arxiv_id":null,"snapshot_observed_at":"2026-08-12T00:46:34.653572Z","title":"GPipe: Efficient training of giant neural networks using pipeline parallelism,","venue":null,"work_id":"4b020aca-fa75-486e-a7d6-f9d17de4ab35","year":2019},"citing_paper":{"arxiv_id":"2608.07974","last_updated":"2026-08-08T07:19:21Z","snapshot_observed_at":"2026-08-18T00:09:50.943155Z","submitted_at":"2026-08-08T07:19:21Z","title":"ZeroLock: Concurrent Memory-Efficient LLM Training via Modular Update Decoupling","version":1},"reference_index":3,"source":"pdf_text","source_observed_at":"2026-08-12T00:46:33.780928Z"},"links":{"citing_paper":"/paper/2608.07974"},"observation_digest":"sha256:c73fd8e49930b8116b21dc05432c3845339f2ed3b7b79d320464f674dcaa173d","observation_id":"8fb9bf48-7fd4-4cf9-8916-23b06228ecdc","resolution":{"observed_at":"2026-08-12T00:46:34.658561Z","resolver_source":"raw_fallback","status":"verified_fuzzy"},"standing_notice":{"events":[],"observation":"No event found in the named queried sources as of 2026-08-22T06:32:14.747728+00:00.","reason":null,"source_receipts":[{"observed_at":"2026-08-22T06:32:14.747728+00:00","source":"crossref"},{"observed_at":"2026-08-22T06:32:06.552537+00:00","source":"retraction_watch"}],"state":"measured"}},{"citation":{"cited_paper":null,"cited_work":{"arxiv_id":null,"doi":null,"metadata_source":"raw_reference","pith_arxiv_id":null,"snapshot_observed_at":"2026-08-12T00:46:34.638848Z","title":"PipeDream: Generalized pipeline parallelism for DNN training,","venue":null,"work_id":"d1ef5ce6-fb03-4b57-8151-c1ea95d1670a","year":2019},"citing_paper":{"arxiv_id":"2608.07974","last_updated":"2026-08-08T07:19:21Z","snapshot_observed_at":"2026-08-18T00:09:50.943155Z","submitted_at":"2026-08-08T07:19:21Z","title":"ZeroLock: Concurrent Memory-Efficient LLM Training via Modular Update Decoupling","version":1},"reference_index":4,"source":"pdf_text","source_observed_at":"2026-08-12T00:46:33.785583Z"},"links":{"citing_paper":"/paper/2608.07974"},"observation_digest":"sha256:87e13c6113e853521217eb25ca307f7f0489b2a5efbcec91da35719f53a39e4b","observation_id":"358fe6c8-a330-4435-928d-914abd502bea","resolution":{"observed_at":"2026-08-12T00:46:34.643962Z","resolver_source":"raw_fallback","status":"verified_fuzzy"},"standing_notice":{"events":[],"observation":"No event found in the named queried sources as of 2026-08-22T06:32:14.747728+00:00.","reason":null,"source_receipts":[{"observed_at":"2026-08-22T06:32:14.747728+00:00","source":"crossref"},{"observed_at":"2026-08-22T06:32:06.552537+00:00","source":"retraction_watch"}],"state":"measured"}},{"citation":{"cited_paper":null,"cited_work":{"arxiv_id":null,"doi":null,"metadata_source":"raw_reference","pith_arxiv_id":null,"snapshot_observed_at":"2026-08-12T00:46:34.624227Z","title":"Confidant: Customizing transformer-based llms via collaborative training on mobile devices,","venue":null,"work_id":"acf8e785-87c4-40b9-92c3-6f371b41cb34","year":2025},"citing_paper":{"arxiv_id":"2608.07974","last_updated":"2026-08-08T07:19:21Z","snapshot_observed_at":"2026-08-18T00:09:50.943155Z","submitted_at":"2026-08-08T07:19:21Z","title":"ZeroLock: Concurrent Memory-Efficient LLM Training via Modular Update Decoupling","version":1},"reference_index":5,"source":"pdf_text","source_observed_at":"2026-08-12T00:46:33.791192Z"},"links":{"citing_paper":"/paper/2608.07974"},"observation_digest":"sha256:423fdf085f0546ed8830f98c929a30f5b4c837b0d40281738aca5aa7144c1b4d","observation_id":"b1c1270b-d892-4ef8-8202-3b10ffda93c8","resolution":{"observed_at":"2026-08-12T00:46:34.629303Z","resolver_source":"raw_fallback","status":"verified_fuzzy"},"standing_notice":{"events":[],"observation":"No event found in the named queried sources as of 2026-08-22T06:32:14.747728+00:00.","reason":null,"source_receipts":[{"observed_at":"2026-08-22T06:32:14.747728+00:00","source":"crossref"},{"observed_at":"2026-08-22T06:32:06.552537+00:00","source":"retraction_watch"}],"state":"measured"}},{"citation":{"cited_paper":null,"cited_work":{"arxiv_id":null,"doi":null,"metadata_source":"raw_reference","pith_arxiv_id":null,"snapshot_observed_at":"2026-08-12T00:46:34.610151Z","title":"Memory-efficient pipeline-parallel DNN training,","venue":null,"work_id":"a6bf9806-b563-43ae-87ec-439c957abf62","year":2021},"citing_paper":{"arxiv_id":"2608.07974","last_updated":"2026-08-08T07:19:21Z","snapshot_observed_at":"2026-08-18T00:09:50.943155Z","submitted_at":"2026-08-08T07:19:21Z","title":"ZeroLock: Concurrent Memory-Efficient LLM Training via Modular Update Decoupling","version":1},"reference_index":6,"source":"pdf_text","source_observed_at":"2026-08-12T00:46:33.796123Z"},"links":{"citing_paper":"/paper/2608.07974"},"observation_digest":"sha256:4b63702bfdf1aa893072ad96b3d50808a9b66f99a83ed19db4c4601548538a69","observation_id":"c210203a-2b00-4f29-a1c5-7fb39533454a","resolution":{"observed_at":"2026-08-12T00:46:34.614684Z","resolver_source":"raw_fallback","status":"verified_fuzzy"},"standing_notice":{"events":[],"observation":"No event found in the named queried sources as of 2026-08-22T06:32:14.747728+00:00.","reason":null,"source_receipts":[{"observed_at":"2026-08-22T06:32:14.747728+00:00","source":"crossref"},{"observed_at":"2026-08-22T06:32:06.552537+00:00","source":"retraction_watch"}],"state":"measured"}},{"citation":{"cited_paper":null,"cited_work":{"arxiv_id":null,"doi":null,"metadata_source":"raw_reference","pith_arxiv_id":null,"snapshot_observed_at":"2026-08-12T00:46:34.595568Z","title":"DAPPLE: A pipelined data parallel approach for training large models,","venue":null,"work_id":"dc007500-c0f4-4042-b60a-832b91f59580","year":2021},"citing_paper":{"arxiv_id":"2608.07974","last_updated":"2026-08-08T07:19:21Z","snapshot_observed_at":"2026-08-18T00:09:50.943155Z","submitted_at":"2026-08-08T07:19:21Z","title":"ZeroLock: Concurrent Memory-Efficient LLM Training via Modular Update Decoupling","version":1},"reference_index":7,"source":"pdf_text","source_observed_at":"2026-08-12T00:46:33.801538Z"},"links":{"citing_paper":"/paper/2608.07974"},"observation_digest":"sha256:7566c956049efc86775f5ea48ee10bfe7b4b0de4243c406112b2fa0d4599af24","observation_id":"f8e54f4a-740f-4742-8eb3-debc8a7ab215","resolution":{"observed_at":"2026-08-12T00:46:34.600541Z","resolver_source":"raw_fallback","status":"verified_fuzzy"},"standing_notice":{"events":[],"observation":"No event found in the named queried sources as of 2026-08-22T06:32:14.747728+00:00.","reason":null,"source_receipts":[{"observed_at":"2026-08-22T06:32:14.747728+00:00","source":"crossref"},{"observed_at":"2026-08-22T06:32:06.552537+00:00","source":"retraction_watch"}],"state":"measured"}},{"citation":{"cited_paper":null,"cited_work":{"arxiv_id":null,"doi":null,"metadata_source":"raw_reference","pith_arxiv_id":null,"snapshot_observed_at":"2026-08-12T00:46:34.580342Z","title":"Efficient large-scale language model training on GPU clusters using Megatron-LM,","venue":null,"work_id":"a5e4916a-a2bc-4cac-9c3d-342be9f07e24","year":2021},"citing_paper":{"arxiv_id":"2608.07974","last_updated":"2026-08-08T07:19:21Z","snapshot_observed_at":"2026-08-18T00:09:50.943155Z","submitted_at":"2026-08-08T07:19:21Z","title":"ZeroLock: Concurrent Memory-Efficient LLM Training via Modular Update Decoupling","version":1},"reference_index":8,"source":"pdf_text","source_observed_at":"2026-08-12T00:46:33.806303Z"},"links":{"citing_paper":"/paper/2608.07974"},"observation_digest":"sha256:ca1630350b434a46c14e0e634f385737fe4b2b3f8d2cf67e3a789c06f1ad4854","observation_id":"55562b2e-781e-48f5-9d9e-7e25f6d21962","resolution":{"observed_at":"2026-08-12T00:46:34.585318Z","resolver_source":"raw_fallback","status":"verified_fuzzy"},"standing_notice":{"events":[],"observation":"No event found in the named queried sources as of 2026-08-22T06:32:14.747728+00:00.","reason":null,"source_receipts":[{"observed_at":"2026-08-22T06:32:14.747728+00:00","source":"crossref"},{"observed_at":"2026-08-22T06:32:06.552537+00:00","source":"retraction_watch"}],"state":"measured"}},{"citation":{"cited_paper":null,"cited_work":{"arxiv_id":null,"doi":null,"metadata_source":"raw_reference","pith_arxiv_id":null,"snapshot_observed_at":"2026-08-12T00:46:34.565009Z","title":"Chimera: Efficiently training large-scale neural networks with bidirectional pipelines,","venue":null,"work_id":"63c14644-9151-4105-9569-5042ee125afb","year":2021},"citing_paper":{"arxiv_id":"2608.07974","last_updated":"2026-08-08T07:19:21Z","snapshot_observed_at":"2026-08-18T00:09:50.943155Z","submitted_at":"2026-08-08T07:19:21Z","title":"ZeroLock: Concurrent Memory-Efficient LLM Training via Modular Update Decoupling","version":1},"reference_index":9,"source":"pdf_text","source_observed_at":"2026-08-12T00:46:33.811024Z"},"links":{"citing_paper":"/paper/2608.07974"},"observation_digest":"sha256:7a109625579709af28509c20eab14210dd5cfe532e0e81622208920e339c2c0b","observation_id":"5c0641d4-7a4d-424f-b6ab-d455d7a8d7a6","resolution":{"observed_at":"2026-08-12T00:46:34.570046Z","resolver_source":"raw_fallback","status":"verified_fuzzy"},"standing_notice":{"events":[],"observation":"No event found in the named queried sources as of 2026-08-22T06:32:14.747728+00:00.","reason":null,"source_receipts":[{"observed_at":"2026-08-22T06:32:14.747728+00:00","source":"crossref"},{"observed_at":"2026-08-22T06:32:06.552537+00:00","source":"retraction_watch"}],"state":"measured"}},{"citation":{"cited_paper":null,"cited_work":{"arxiv_id":null,"doi":null,"metadata_source":"raw_reference","pith_arxiv_id":null,"snapshot_observed_at":"2026-08-12T00:46:34.547193Z","title":"Zero bubble (almost) pipeline parallelism,","venue":null,"work_id":"f32de4c1-3c53-4804-ba8d-8da7f6620451","year":2024},"citing_paper":{"arxiv_id":"2608.07974","last_updated":"2026-08-08T07:19:21Z","snapshot_observed_at":"2026-08-18T00:09:50.943155Z","submitted_at":"2026-08-08T07:19:21Z","title":"ZeroLock: Concurrent Memory-Efficient LLM Training via Modular Update Decoupling","version":1},"reference_index":10,"source":"pdf_text","source_observed_at":"2026-08-12T00:46:33.815379Z"},"links":{"citing_paper":"/paper/2608.07974"},"observation_digest":"sha256:f159e154824128d7b34a3a29a6f3cc000b2159d6ca6bd275c6a27f1dc9d79881","observation_id":"0803ad90-64a3-41a2-b0b7-4cfff52ec579","resolution":{"observed_at":"2026-08-12T00:46:34.554267Z","resolver_source":"raw_fallback","status":"verified_fuzzy"},"standing_notice":{"events":[],"observation":"No event found in the named queried sources as of 2026-08-22T06:32:14.747728+00:00.","reason":null,"source_receipts":[{"observed_at":"2026-08-22T06:32:14.747728+00:00","source":"crossref"},{"observed_at":"2026-08-22T06:32:06.552537+00:00","source":"retraction_watch"}],"state":"measured"}},{"citation":{"cited_paper":null,"cited_work":{"arxiv_id":null,"doi":null,"metadata_source":"raw_reference","pith_arxiv_id":null,"snapshot_observed_at":"2026-08-12T00:46:34.531741Z","title":"Universal checkpointing: A flexible and efficient distributed checkpointing system for large-scale DNN training with reconfigurable parallelism,","venue":null,"work_id":"c5bb20fd-9a13-4de3-8a2d-4729f53bcbc3","year":2025},"citing_paper":{"arxiv_id":"2608.07974","last_updated":"2026-08-08T07:19:21Z","snapshot_observed_at":"2026-08-18T00:09:50.943155Z","submitted_at":"2026-08-08T07:19:21Z","title":"ZeroLock: Concurrent Memory-Efficient LLM Training via Modular Update Decoupling","version":1},"reference_index":11,"source":"pdf_text","source_observed_at":"2026-08-12T00:46:33.819892Z"},"links":{"citing_paper":"/paper/2608.07974"},"observation_digest":"sha256:bdfb051ce13f7a4b60d1076f274270085f1faadfa2890317c719d2a03800a688","observation_id":"f7160761-c329-4055-b4d7-e815eac16991","resolution":{"observed_at":"2026-08-12T00:46:34.536594Z","resolver_source":"raw_fallback","status":"verified_fuzzy"},"standing_notice":{"events":[],"observation":"No event found in the named queried sources as of 2026-08-22T06:32:14.747728+00:00.","reason":null,"source_receipts":[{"observed_at":"2026-08-22T06:32:14.747728+00:00","source":"crossref"},{"observed_at":"2026-08-22T06:32:06.552537+00:00","source":"retraction_watch"}],"state":"measured"}},{"citation":{"cited_paper":null,"cited_work":{"arxiv_id":null,"doi":null,"metadata_source":"raw_reference","pith_arxiv_id":null,"snapshot_observed_at":"2026-08-12T00:46:34.517815Z","title":"Sparse checkpointing for fast and reliable MoE training,","venue":null,"work_id":"3382fed8-ba8f-4c64-a9a5-ce2b464166cb","year":2026},"citing_paper":{"arxiv_id":"2608.07974","last_updated":"2026-08-08T07:19:21Z","snapshot_observed_at":"2026-08-18T00:09:50.943155Z","submitted_at":"2026-08-08T07:19:21Z","title":"ZeroLock: Concurrent Memory-Efficient LLM Training via Modular Update Decoupling","version":1},"reference_index":12,"source":"pdf_text","source_observed_at":"2026-08-12T00:46:33.824292Z"},"links":{"citing_paper":"/paper/2608.07974"},"observation_digest":"sha256:d81db5bad54011c975fde475b4fa137c5c5cde1ddc91a6199e1de343d50f6280","observation_id":"137e6c87-ab0f-4c1c-a8fa-ca2f164aecbf","resolution":{"observed_at":"2026-08-12T00:46:34.522521Z","resolver_source":"raw_fallback","status":"verified_fuzzy"},"standing_notice":{"events":[],"observation":"No event found in the named queried sources as of 2026-08-22T06:32:14.747728+00:00.","reason":null,"source_receipts":[{"observed_at":"2026-08-22T06:32:14.747728+00:00","source":"crossref"},{"observed_at":"2026-08-22T06:32:06.552537+00:00","source":"retraction_watch"}],"state":"measured"}},{"citation":{"cited_paper":null,"cited_work":{"arxiv_id":null,"doi":null,"metadata_source":"raw_reference","pith_arxiv_id":null,"snapshot_observed_at":"2026-08-12T00:46:34.503290Z","title":"Attack of the bubbles: Straggler-resilient pipeline parallelism for large model training,","venue":null,"work_id":"733ea29a-ba81-421c-a42d-277dd7582dae","year":2026},"citing_paper":{"arxiv_id":"2608.07974","last_updated":"2026-08-08T07:19:21Z","snapshot_observed_at":"2026-08-18T00:09:50.943155Z","submitted_at":"2026-08-08T07:19:21Z","title":"ZeroLock: Concurrent Memory-Efficient LLM Training via Modular Update Decoupling","version":1},"reference_index":13,"source":"pdf_text","source_observed_at":"2026-08-12T00:46:33.829173Z"},"links":{"citing_paper":"/paper/2608.07974"},"observation_digest":"sha256:c4b0057d12b76635a55cb2433a235ead5f23691275a12815f1592b40cc770320","observation_id":"a12ebc72-7b7b-442b-9701-0ca6066ea048","resolution":{"observed_at":"2026-08-12T00:46:34.508234Z","resolver_source":"raw_fallback","status":"verified_fuzzy"},"standing_notice":{"events":[],"observation":"No event found in the named queried sources as of 2026-08-22T06:32:14.747728+00:00.","reason":null,"source_receipts":[{"observed_at":"2026-08-22T06:32:14.747728+00:00","source":"crossref"},{"observed_at":"2026-08-22T06:32:06.552537+00:00","source":"retraction_watch"}],"state":"measured"}},{"citation":{"cited_paper":null,"cited_work":{"arxiv_id":"2509.19855","doi":null,"metadata_source":"raw_reference","pith_arxiv_id":null,"snapshot_observed_at":"2026-08-12T00:46:34.187883Z","title":"CollaPipe: Adaptive segment-optimized pipeline parallelism for collabo- rative LLM training in heterogeneous edge networks,","venue":null,"work_id":"086d7b00-3a90-42ee-a578-c2b8809de5a9","year":2025},"citing_paper":{"arxiv_id":"2608.07974","last_updated":"2026-08-08T07:19:21Z","snapshot_observed_at":"2026-08-18T00:09:50.943155Z","submitted_at":"2026-08-08T07:19:21Z","title":"ZeroLock: Concurrent Memory-Efficient LLM Training via Modular Update Decoupling","version":1},"reference_index":14,"source":"pdf_text","source_observed_at":"2026-08-12T00:46:33.833780Z"},"links":{"citing_paper":"/paper/2608.07974"},"observation_digest":"sha256:29c9cfaf529c00dcd581c2b4e831a9d1083ffde4559b3699213b4469e8851973","observation_id":"b4806688-c2e1-4e6b-99bf-d130af39f4e9","resolution":{"observed_at":"2026-08-12T00:46:34.195442Z","resolver_source":"raw_fallback","status":"verified_exact"},"standing_notice":{"events":[],"observation":"No event found in the named queried sources as of 2026-08-22T06:32:14.747728+00:00.","reason":null,"source_receipts":[{"observed_at":"2026-08-22T06:32:14.747728+00:00","source":"crossref"},{"observed_at":"2026-08-22T06:32:06.552537+00:00","source":"retraction_watch"}],"state":"measured"}},{"citation":{"cited_paper":null,"cited_work":{"arxiv_id":null,"doi":null,"metadata_source":"raw_reference","pith_arxiv_id":null,"snapshot_observed_at":"2026-08-12T00:46:34.488816Z","title":"SWARM parallelism: Training large models can be surprisingly communication- efficient,","venue":null,"work_id":"bd3dabb7-b266-4ed4-9144-8607df7e9c7d","year":2023},"citing_paper":{"arxiv_id":"2608.07974","last_updated":"2026-08-08T07:19:21Z","snapshot_observed_at":"2026-08-18T00:09:50.943155Z","submitted_at":"2026-08-08T07:19:21Z","title":"ZeroLock: Concurrent Memory-Efficient LLM Training via Modular Update Decoupling","version":1},"reference_index":15,"source":"pdf_text","source_observed_at":"2026-08-12T00:46:33.838574Z"},"links":{"citing_paper":"/paper/2608.07974"},"observation_digest":"sha256:2b03c23627467eaa7b7406e1553bcbb2ed5c452b7610d7493cc7fd466c41ccd0","observation_id":"66e15773-bc30-4426-85e4-2c440a04b878","resolution":{"observed_at":"2026-08-12T00:46:34.493719Z","resolver_source":"raw_fallback","status":"verified_fuzzy"},"standing_notice":{"events":[],"observation":"No event found in the named queried sources as of 2026-08-22T06:32:14.747728+00:00.","reason":null,"source_receipts":[{"observed_at":"2026-08-22T06:32:14.747728+00:00","source":"crossref"},{"observed_at":"2026-08-22T06:32:06.552537+00:00","source":"retraction_watch"}],"state":"measured"}},{"citation":{"cited_paper":null,"cited_work":{"arxiv_id":null,"doi":null,"metadata_source":"raw_reference","pith_arxiv_id":null,"snapshot_observed_at":"2026-08-12T00:46:34.474163Z","title":"Petals: Collaborative infer- ence and fine-tuning of large models,","venue":null,"work_id":"f5d77b5e-337a-4afb-af5a-2786c63f28fe","year":2023},"citing_paper":{"arxiv_id":"2608.07974","last_updated":"2026-08-08T07:19:21Z","snapshot_observed_at":"2026-08-18T00:09:50.943155Z","submitted_at":"2026-08-08T07:19:21Z","title":"ZeroLock: Concurrent Memory-Efficient LLM Training via Modular Update Decoupling","version":1},"reference_index":16,"source":"pdf_text","source_observed_at":"2026-08-12T00:46:33.843338Z"},"links":{"citing_paper":"/paper/2608.07974"},"observation_digest":"sha256:bd359f0d7cfa12d34f17522673edc9fc5de5ce580e4c0fca067fc34dc7ce41c5","observation_id":"dfa104de-ecbd-4799-9b99-5b76c6eba5c1","resolution":{"observed_at":"2026-08-12T00:46:34.479123Z","resolver_source":"raw_fallback","status":"verified_fuzzy"},"standing_notice":{"events":[],"observation":"No event found in the named queried sources as of 2026-08-22T06:32:14.747728+00:00.","reason":null,"source_receipts":[{"observed_at":"2026-08-22T06:32:14.747728+00:00","source":"crossref"},{"observed_at":"2026-08-22T06:32:06.552537+00:00","source":"retraction_watch"}],"state":"measured"}},{"citation":{"cited_paper":null,"cited_work":{"arxiv_id":null,"doi":null,"metadata_source":"raw_reference","pith_arxiv_id":null,"snapshot_observed_at":"2026-08-12T00:46:34.458289Z","title":"Decoupled parallel backpropagation with convergence guarantee,","venue":null,"work_id":"beafcbb9-5dbe-48ec-8dd4-8a9fa59c3645","year":2018},"citing_paper":{"arxiv_id":"2608.07974","last_updated":"2026-08-08T07:19:21Z","snapshot_observed_at":"2026-08-18T00:09:50.943155Z","submitted_at":"2026-08-08T07:19:21Z","title":"ZeroLock: Concurrent Memory-Efficient LLM Training via Modular Update Decoupling","version":1},"reference_index":17,"source":"pdf_text","source_observed_at":"2026-08-12T00:46:33.847818Z"},"links":{"citing_paper":"/paper/2608.07974"},"observation_digest":"sha256:d9ff06027a8b7e44a6e340be788850b51624da3ee74db0ec6470dc0bfda4219b","observation_id":"dd5b1f40-690a-41e9-a12a-892798b2133b","resolution":{"observed_at":"2026-08-12T00:46:34.463654Z","resolver_source":"raw_fallback","status":"verified_fuzzy"},"standing_notice":{"events":[],"observation":"No event found in the named queried sources as of 2026-08-22T06:32:14.747728+00:00.","reason":null,"source_receipts":[{"observed_at":"2026-08-22T06:32:14.747728+00:00","source":"crossref"},{"observed_at":"2026-08-22T06:32:06.552537+00:00","source":"retraction_watch"}],"state":"measured"}},{"citation":{"cited_paper":null,"cited_work":{"arxiv_id":null,"doi":null,"metadata_source":"raw_reference","pith_arxiv_id":null,"snapshot_observed_at":"2026-08-12T00:46:34.442933Z","title":"Depth-progressive monotonic learning without global backpropagation,","venue":null,"work_id":"e9e007b7-7b7d-4480-9748-d732027bd359","year":2026},"citing_paper":{"arxiv_id":"2608.07974","last_updated":"2026-08-08T07:19:21Z","snapshot_observed_at":"2026-08-18T00:09:50.943155Z","submitted_at":"2026-08-08T07:19:21Z","title":"ZeroLock: Concurrent Memory-Efficient LLM Training via Modular Update Decoupling","version":1},"reference_index":18,"source":"pdf_text","source_observed_at":"2026-08-12T00:46:33.852499Z"},"links":{"citing_paper":"/paper/2608.07974"},"observation_digest":"sha256:f94bb61f2f138748405688a08a1bf19b93a69f3bcd8c81056ca104b8ce1c3bff","observation_id":"0689e8ca-d738-45b1-adb3-981ce3602829","resolution":{"observed_at":"2026-08-12T00:46:34.447949Z","resolver_source":"raw_fallback","status":"verified_fuzzy"},"standing_notice":{"events":[],"observation":"No event found in the named queried sources as of 2026-08-22T06:32:14.747728+00:00.","reason":null,"source_receipts":[{"observed_at":"2026-08-22T06:32:14.747728+00:00","source":"crossref"},{"observed_at":"2026-08-22T06:32:06.552537+00:00","source":"retraction_watch"}],"state":"measured"}},{"citation":{"cited_paper":null,"cited_work":{"arxiv_id":"0426.63642","doi":null,"metadata_source":"raw_reference","pith_arxiv_id":null,"snapshot_observed_at":"2026-08-12T00:46:34.087952Z","title":"Beyond- backpropagation training: Methods, applications, and perspectives,","venue":null,"work_id":"ef4a7921-5560-4bed-b404-e2dc764809cf","year":2026},"citing_paper":{"arxiv_id":"2608.07974","last_updated":"2026-08-08T07:19:21Z","snapshot_observed_at":"2026-08-18T00:09:50.943155Z","submitted_at":"2026-08-08T07:19:21Z","title":"ZeroLock: Concurrent Memory-Efficient LLM Training via Modular Update Decoupling","version":1},"reference_index":19,"source":"pdf_text","source_observed_at":"2026-08-12T00:46:33.857508Z"},"links":{"citing_paper":"/paper/2608.07974"},"observation_digest":"sha256:d7c8087955e9357203568a865fee2ae3498ef9f431fed325f26dddc58d215296","observation_id":"6930df10-4428-4666-a38f-ff7cea643f24","resolution":{"observed_at":"2026-08-12T00:46:34.099602Z","resolver_source":"raw_fallback","status":"metadata_mismatch"},"standing_notice":{"events":[],"observation":"No event found in the named queried sources as of 2026-08-22T06:32:14.747728+00:00.","reason":null,"source_receipts":[{"observed_at":"2026-08-22T06:32:14.747728+00:00","source":"crossref"},{"observed_at":"2026-08-22T06:32:06.552537+00:00","source":"retraction_watch"}],"state":"measured"}},{"citation":{"cited_paper":null,"cited_work":{"arxiv_id":null,"doi":null,"metadata_source":"raw_reference","pith_arxiv_id":null,"snapshot_observed_at":"2026-08-12T00:46:34.427050Z","title":"Direct feedback alignment provides learning in deep neural networks,","venue":null,"work_id":"f243d55b-4f09-4ed7-b0fe-bc8668a35d69","year":2016},"citing_paper":{"arxiv_id":"2608.07974","last_updated":"2026-08-08T07:19:21Z","snapshot_observed_at":"2026-08-18T00:09:50.943155Z","submitted_at":"2026-08-08T07:19:21Z","title":"ZeroLock: Concurrent Memory-Efficient LLM Training via Modular Update Decoupling","version":1},"reference_index":20,"source":"pdf_text","source_observed_at":"2026-08-12T00:46:33.862508Z"},"links":{"citing_paper":"/paper/2608.07974"},"observation_digest":"sha256:9ddde642763aa772fbdb1383428dbb7abf616d05f575b01a449ea466315559f7","observation_id":"c593da75-0bfd-4d38-b9e6-3b3263e8a51d","resolution":{"observed_at":"2026-08-12T00:46:34.431998Z","resolver_source":"raw_fallback","status":"verified_fuzzy"},"standing_notice":{"events":[],"observation":"No event found in the named queried sources as of 2026-08-22T06:32:14.747728+00:00.","reason":null,"source_receipts":[{"observed_at":"2026-08-22T06:32:14.747728+00:00","source":"crossref"},{"observed_at":"2026-08-22T06:32:06.552537+00:00","source":"retraction_watch"}],"state":"measured"}},{"citation":{"cited_paper":null,"cited_work":{"arxiv_id":null,"doi":null,"metadata_source":"raw_reference","pith_arxiv_id":null,"snapshot_observed_at":"2026-08-12T00:46:34.411740Z","title":"Deep learning without weight transport,","venue":null,"work_id":"1f1b62d1-5dbf-4d9e-967c-7c7430b3c949","year":2019},"citing_paper":{"arxiv_id":"2608.07974","last_updated":"2026-08-08T07:19:21Z","snapshot_observed_at":"2026-08-18T00:09:50.943155Z","submitted_at":"2026-08-08T07:19:21Z","title":"ZeroLock: Concurrent Memory-Efficient LLM Training via Modular Update Decoupling","version":1},"reference_index":21,"source":"pdf_text","source_observed_at":"2026-08-12T00:46:33.867604Z"},"links":{"citing_paper":"/paper/2608.07974"},"observation_digest":"sha256:0f8c002e15c908a5018dca03941ca57fcc5112de62a2c47a9cf7d9e3816ee72e","observation_id":"454a38ba-a87b-4f56-8b6b-6ddcf6290282","resolution":{"observed_at":"2026-08-12T00:46:34.416765Z","resolver_source":"raw_fallback","status":"verified_fuzzy"},"standing_notice":{"events":[],"observation":"No event found in the named queried sources as of 2026-08-22T06:32:14.747728+00:00.","reason":null,"source_receipts":[{"observed_at":"2026-08-22T06:32:14.747728+00:00","source":"crossref"},{"observed_at":"2026-08-22T06:32:06.552537+00:00","source":"retraction_watch"}],"state":"measured"}},{"citation":{"cited_paper":null,"cited_work":{"arxiv_id":null,"doi":null,"metadata_source":"raw_reference","pith_arxiv_id":null,"snapshot_observed_at":"2026-08-12T00:46:34.396452Z","title":"Fine-tuning language models with just forward passes,","venue":null,"work_id":"28588760-755a-4236-86ad-31a7e67364c0","year":2023},"citing_paper":{"arxiv_id":"2608.07974","last_updated":"2026-08-08T07:19:21Z","snapshot_observed_at":"2026-08-18T00:09:50.943155Z","submitted_at":"2026-08-08T07:19:21Z","title":"ZeroLock: Concurrent Memory-Efficient LLM Training via Modular Update Decoupling","version":1},"reference_index":22,"source":"pdf_text","source_observed_at":"2026-08-12T00:46:33.872400Z"},"links":{"citing_paper":"/paper/2608.07974"},"observation_digest":"sha256:72673414e3344e741a94126d6a1197048856b200935022b51b5c875ed6da1e19","observation_id":"a431d332-d24a-40c9-b0eb-d72772b83c4d","resolution":{"observed_at":"2026-08-12T00:46:34.401637Z","resolver_source":"raw_fallback","status":"verified_fuzzy"},"standing_notice":{"events":[],"observation":"No event found in the named queried sources as of 2026-08-22T06:32:14.747728+00:00.","reason":null,"source_receipts":[{"observed_at":"2026-08-22T06:32:14.747728+00:00","source":"crossref"},{"observed_at":"2026-08-22T06:32:06.552537+00:00","source":"retraction_watch"}],"state":"measured"}},{"citation":{"cited_paper":null,"cited_work":{"arxiv_id":null,"doi":null,"metadata_source":"raw_reference","pith_arxiv_id":null,"snapshot_observed_at":"2026-08-12T00:46:34.380065Z","title":"Curvzo: Adaptive curvature-guided sparse zeroth-order optimization for efficient llm fine-tuning,","venue":null,"work_id":"ed30a4bd-2c2b-460e-a11a-2e8ccb967d43","year":2026},"citing_paper":{"arxiv_id":"2608.07974","last_updated":"2026-08-08T07:19:21Z","snapshot_observed_at":"2026-08-18T00:09:50.943155Z","submitted_at":"2026-08-08T07:19:21Z","title":"ZeroLock: Concurrent Memory-Efficient LLM Training via Modular Update Decoupling","version":1},"reference_index":23,"source":"pdf_text","source_observed_at":"2026-08-12T00:46:33.877627Z"},"links":{"citing_paper":"/paper/2608.07974"},"observation_digest":"sha256:dd697c0f0cfbfd10d6803113f76156e88b2f2f5fb00c73b3afeaa3759fd8904b","observation_id":"1429a096-00f3-48d7-bdf8-df9dad67dba7","resolution":{"observed_at":"2026-08-12T00:46:34.385676Z","resolver_source":"raw_fallback","status":"verified_fuzzy"},"standing_notice":{"events":[],"observation":"No event found in the named queried sources as of 2026-08-22T06:32:14.747728+00:00.","reason":null,"source_receipts":[{"observed_at":"2026-08-22T06:32:14.747728+00:00","source":"crossref"},{"observed_at":"2026-08-22T06:32:06.552537+00:00","source":"retraction_watch"}],"state":"measured"}},{"citation":{"cited_paper":null,"cited_work":{"arxiv_id":null,"doi":null,"metadata_source":"raw_reference","pith_arxiv_id":null,"snapshot_observed_at":"2026-08-12T00:46:34.364175Z","title":"Noprop: Training neural networks without back-propagation or forward-propagation,","venue":null,"work_id":"6a203111-c361-44a5-9499-cb1a31b2bf66","year":2026},"citing_paper":{"arxiv_id":"2608.07974","last_updated":"2026-08-08T07:19:21Z","snapshot_observed_at":"2026-08-18T00:09:50.943155Z","submitted_at":"2026-08-08T07:19:21Z","title":"ZeroLock: Concurrent Memory-Efficient LLM Training via Modular Update Decoupling","version":1},"reference_index":24,"source":"pdf_text","source_observed_at":"2026-08-12T00:46:33.882852Z"},"links":{"citing_paper":"/paper/2608.07974"},"observation_digest":"sha256:e4f7c146a6abafea846b8cefd50a0691e7529fc2c024a273c86624d5678f931e","observation_id":"7b2b594b-8bfb-400e-b380-3dd39c4f580e","resolution":{"observed_at":"2026-08-12T00:46:34.369090Z","resolver_source":"raw_fallback","status":"verified_fuzzy"},"standing_notice":{"events":[],"observation":"No event found in the named queried sources as of 2026-08-22T06:32:14.747728+00:00.","reason":null,"source_receipts":[{"observed_at":"2026-08-22T06:32:14.747728+00:00","source":"crossref"},{"observed_at":"2026-08-22T06:32:06.552537+00:00","source":"retraction_watch"}],"state":"measured"}},{"citation":{"cited_paper":null,"cited_work":{"arxiv_id":null,"doi":null,"metadata_source":"raw_reference","pith_arxiv_id":null,"snapshot_observed_at":"2026-08-12T00:46:34.348702Z","title":"Predictive coding approximates backprop along arbitrary computation graphs,","venue":null,"work_id":"285dcacd-0c83-4a03-b62e-71f66dd73cfd","year":2022},"citing_paper":{"arxiv_id":"2608.07974","last_updated":"2026-08-08T07:19:21Z","snapshot_observed_at":"2026-08-18T00:09:50.943155Z","submitted_at":"2026-08-08T07:19:21Z","title":"ZeroLock: Concurrent Memory-Efficient LLM Training via Modular Update Decoupling","version":1},"reference_index":25,"source":"pdf_text","source_observed_at":"2026-08-12T00:46:33.887561Z"},"links":{"citing_paper":"/paper/2608.07974"},"observation_digest":"sha256:07d9974309d8234d1feab8175515a5695e6d8887bc627ae60e221b019ba59419","observation_id":"8d820b95-65e2-477e-a9f7-ed06e5db332b","resolution":{"observed_at":"2026-08-12T00:46:34.353685Z","resolver_source":"raw_fallback","status":"verified_fuzzy"},"standing_notice":{"events":[],"observation":"No event found in the named queried sources as of 2026-08-22T06:32:14.747728+00:00.","reason":null,"source_receipts":[{"observed_at":"2026-08-22T06:32:14.747728+00:00","source":"crossref"},{"observed_at":"2026-08-22T06:32:06.552537+00:00","source":"retraction_watch"}],"state":"measured"}},{"citation":{"cited_paper":{"arxiv_id":"2411.12780","last_updated":"2024-11-19T08:09:18Z","snapshot_observed_at":"2026-08-14T08:22:57.930092Z","submitted_at":"2024-11-19T08:09:18Z","title":"Faster Multi-GPU Training with PPLL: A Pipeline Parallelism Framework Leveraging Local Learning","version":1},"cited_work":{"arxiv_id":null,"doi":null,"metadata_source":null,"pith_arxiv_id":"2411.12780","snapshot_observed_at":"2026-08-12T00:46:33.892664Z","title":"Faster multi-GPU training with PPLL: A pipeline parallelism framework leveraging local learning,","venue":null,"work_id":null,"year":2024},"citing_paper":{"arxiv_id":"2608.07974","last_updated":"2026-08-08T07:19:21Z","snapshot_observed_at":"2026-08-18T00:09:50.943155Z","submitted_at":"2026-08-08T07:19:21Z","title":"ZeroLock: Concurrent Memory-Efficient LLM Training via Modular Update Decoupling","version":1},"reference_index":26,"source":"pdf_text","source_observed_at":"2026-08-12T00:46:33.892664Z"},"links":{"cited_paper":"/paper/2411.12780","citing_paper":"/paper/2608.07974"},"observation_digest":"sha256:fb305d7e2c2c83661ffbeef7b4f07894cf56a08d0d91b0acad75296bc2bc5d71","observation_id":"740843f9-c526-482d-b412-87ce05d8d282","resolution":{"observed_at":"2026-08-12T00:46:33.892664Z","resolver_source":null,"status":"unresolved"},"standing_notice":{"events":[],"reason":"canonical_work_link_unavailable","source_receipts":[],"state":"unavailable"}},{"citation":{"cited_paper":null,"cited_work":{"arxiv_id":null,"doi":null,"metadata_source":"raw_reference","pith_arxiv_id":null,"snapshot_observed_at":"2026-08-12T00:46:34.332726Z","title":"Idle no more: Boosting distributed pipeline training via FluidPipe,","venue":null,"work_id":"d6562f9b-c96f-444a-8d5b-a1a7a1274647","year":2025},"citing_paper":{"arxiv_id":"2608.07974","last_updated":"2026-08-08T07:19:21Z","snapshot_observed_at":"2026-08-18T00:09:50.943155Z","submitted_at":"2026-08-08T07:19:21Z","title":"ZeroLock: Concurrent Memory-Efficient LLM Training via Modular Update Decoupling","version":1},"reference_index":27,"source":"pdf_text","source_observed_at":"2026-08-12T00:46:33.897990Z"},"links":{"citing_paper":"/paper/2608.07974"},"observation_digest":"sha256:ede55a1876a8551675c8f2193f27d7ad43fd020e701295fa640d8a82fe3c4bca","observation_id":"01934c66-198a-4fd8-a292-44df2455137b","resolution":{"observed_at":"2026-08-12T00:46:34.337937Z","resolver_source":"raw_fallback","status":"verified_fuzzy"},"standing_notice":{"events":[],"observation":"No event found in the named queried sources as of 2026-08-22T06:32:14.747728+00:00.","reason":null,"source_receipts":[{"observed_at":"2026-08-22T06:32:14.747728+00:00","source":"crossref"},{"observed_at":"2026-08-22T06:32:06.552537+00:00","source":"retraction_watch"}],"state":"measured"}},{"citation":{"cited_paper":null,"cited_work":{"arxiv_id":null,"doi":null,"metadata_source":"raw_reference","pith_arxiv_id":null,"snapshot_observed_at":"2026-08-12T00:46:34.316780Z","title":"Scpl: Enhancing neural network training throughput with decoupled local losses and model parallelism,","venue":null,"work_id":"983890bd-2efc-472c-b9d8-369e3c3e6b8c","year":2026},"citing_paper":{"arxiv_id":"2608.07974","last_updated":"2026-08-08T07:19:21Z","snapshot_observed_at":"2026-08-18T00:09:50.943155Z","submitted_at":"2026-08-08T07:19:21Z","title":"ZeroLock: Concurrent Memory-Efficient LLM Training via Modular Update Decoupling","version":1},"reference_index":28,"source":"pdf_text","source_observed_at":"2026-08-12T00:46:33.902896Z"},"links":{"citing_paper":"/paper/2608.07974"},"observation_digest":"sha256:fd30aa6628489e2c89c87ab80fd55786a378dbe394abc9248243795fdaa0166f","observation_id":"88c69243-1253-4f94-9b27-3d9a7853a1d7","resolution":{"observed_at":"2026-08-12T00:46:34.321723Z","resolver_source":"raw_fallback","status":"verified_fuzzy"},"standing_notice":{"events":[],"observation":"No event found in the named queried sources as of 2026-08-22T06:32:14.747728+00:00.","reason":null,"source_receipts":[{"observed_at":"2026-08-22T06:32:14.747728+00:00","source":"crossref"},{"observed_at":"2026-08-22T06:32:06.552537+00:00","source":"retraction_watch"}],"state":"measured"}},{"citation":{"cited_paper":{"arxiv_id":"2605.04913","last_updated":"2026-06-08T07:11:37Z","snapshot_observed_at":"2026-08-12T12:42:19.763135Z","submitted_at":"2026-05-06T13:41:23Z","title":"Rethinking Local Learning: A Cheaper and Faster Recipe for LLM Post-Training","version":4},"cited_work":{"arxiv_id":null,"doi":null,"metadata_source":null,"pith_arxiv_id":"2605.04913","snapshot_observed_at":"2026-08-12T00:46:33.907700Z","title":"Rethinking local learning: A cheaper and faster recipe for LLM post-training,","venue":null,"work_id":null,"year":2026},"citing_paper":{"arxiv_id":"2608.07974","last_updated":"2026-08-08T07:19:21Z","snapshot_observed_at":"2026-08-18T00:09:50.943155Z","submitted_at":"2026-08-08T07:19:21Z","title":"ZeroLock: Concurrent Memory-Efficient LLM Training via Modular Update Decoupling","version":1},"reference_index":29,"source":"pdf_text","source_observed_at":"2026-08-12T00:46:33.907700Z"},"links":{"cited_paper":"/paper/2605.04913","citing_paper":"/paper/2608.07974"},"observation_digest":"sha256:25a1fd17a472f91475d00f65e5552514351b32646f6dd18d5c06678dae2796ef","observation_id":"0f8d825f-b61a-4c73-932b-2f94f343c122","resolution":{"observed_at":"2026-08-12T00:46:33.907700Z","resolver_source":null,"status":"unresolved"},"standing_notice":{"events":[],"reason":"canonical_work_link_unavailable","source_receipts":[],"state":"unavailable"}},{"citation":{"cited_paper":null,"cited_work":{"arxiv_id":null,"doi":null,"metadata_source":"raw_reference","pith_arxiv_id":null,"snapshot_observed_at":"2026-08-12T00:46:34.301657Z","title":"LoRA: Low-rank adaptation of large language models,","venue":null,"work_id":"54790b43-36e0-48b1-ace8-9d26daa4b380","year":2022},"citing_paper":{"arxiv_id":"2608.07974","last_updated":"2026-08-08T07:19:21Z","snapshot_observed_at":"2026-08-18T00:09:50.943155Z","submitted_at":"2026-08-08T07:19:21Z","title":"ZeroLock: Concurrent Memory-Efficient LLM Training via Modular Update Decoupling","version":1},"reference_index":30,"source":"pdf_text","source_observed_at":"2026-08-12T00:46:33.913050Z"},"links":{"citing_paper":"/paper/2608.07974"},"observation_digest":"sha256:ad651ae75f8c19226fb06ab2d080da9b09bc2177ea0477dbb112e7d60656d11c","observation_id":"98bad4f2-a3be-4f38-90ea-7f29c49922f4","resolution":{"observed_at":"2026-08-12T00:46:34.306538Z","resolver_source":"raw_fallback","status":"verified_fuzzy"},"standing_notice":{"events":[],"observation":"No event found in the named queried sources as of 2026-08-22T06:32:14.747728+00:00.","reason":null,"source_receipts":[{"observed_at":"2026-08-22T06:32:14.747728+00:00","source":"crossref"},{"observed_at":"2026-08-22T06:32:06.552537+00:00","source":"retraction_watch"}],"state":"measured"}},{"citation":{"cited_paper":null,"cited_work":{"arxiv_id":null,"doi":null,"metadata_source":null,"pith_arxiv_id":null,"snapshot_observed_at":"2026-08-12T00:46:33.917982Z","title":"Convergence analysis of split federated learning on heterogeneous data,","venue":null,"work_id":null,"year":2024},"citing_paper":{"arxiv_id":"2608.07974","last_updated":"2026-08-08T07:19:21Z","snapshot_observed_at":"2026-08-18T00:09:50.943155Z","submitted_at":"2026-08-08T07:19:21Z","title":"ZeroLock: Concurrent Memory-Efficient LLM Training via Modular Update Decoupling","version":1},"reference_index":31,"source":"pdf_text","source_observed_at":"2026-08-12T00:46:33.917982Z"},"links":{"citing_paper":"/paper/2608.07974"},"observation_digest":"sha256:dff50db9f831ea7249358ec4679705dc6c7390bd0634309699a5b50393e64eff","observation_id":"fb406d83-7019-478c-80da-0cf32e67f85a","resolution":{"observed_at":"2026-08-12T00:46:33.917982Z","resolver_source":null,"status":"unresolved"},"standing_notice":{"events":[],"reason":"canonical_work_link_unavailable","source_receipts":[],"state":"unavailable"}}],"paper":{"arxiv_id":"2608.07974","last_updated":"2026-08-08T07:19:21Z","latest_version":1,"primary_category":"cs.LG","snapshot_observed_at":"2026-08-18T00:09:50.943155Z","submitted_at":"2026-08-08T07:19:21Z","title":"ZeroLock: Concurrent Memory-Efficient LLM Training via Modular Update Decoupling"},"reference_resolution":{"displayed":31,"state_counts":{"malformed_identifier":0,"metadata_mismatch":1,"parse_uncertain":0,"unresolved":4,"verified_exact":1,"verified_fuzzy":25},"total_outbound_references":31},"refusal":"A citation records a reference. It does not transfer a finding from one paper to another.","schema":"pith.paper-citation-record.v1","standing_sources":[{"observed_at":"2026-08-22T06:32:14.747728+00:00","source":"crossref"},{"observed_at":"2026-08-22T06:32:06.552537+00:00","source":"retraction_watch"}],"thesis":"As of 23 August 2026, this Paper Citation Record lists 31 of 31 outbound references and 0 inbound Pith citation observations for arXiv:2608.07974."}