{"state_type":"pith_open_graph_state","state_version":"1.0","pith_number":"pith:2026:YKUE67AD6VUSXZ3BOHXVIOSZXB","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":"998e7d6fc4ec61e701d48cae1b3adbda0d3f59e97e83ef7002eba2969d30a6e7","cross_cats_sorted":[],"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.DC","submitted_at":"2026-07-20T07:57:34Z","title_canon_sha256":"dc371da43d7bcf12d62cd3e75667e8a06c1257af02b88e29d4f17399c85d0342"},"schema_version":"1.0","source":{"id":"2607.17644","kind":"arxiv","version":1}},"source_aliases":[{"alias_kind":"arxiv","alias_value":"2607.17644","created_at":"2026-07-21T01:21:56Z"},{"alias_kind":"arxiv_version","alias_value":"2607.17644v1","created_at":"2026-07-21T01:21:56Z"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2607.17644","created_at":"2026-07-21T01:21:56Z"},{"alias_kind":"pith_short_12","alias_value":"YKUE67AD6VUS","created_at":"2026-07-21T01:21:56Z"},{"alias_kind":"pith_short_16","alias_value":"YKUE67AD6VUSXZ3B","created_at":"2026-07-21T01:21:56Z"},{"alias_kind":"pith_short_8","alias_value":"YKUE67AD","created_at":"2026-07-21T01:21:56Z"}],"graph_snapshots":[{"event_id":"sha256:a48a104a590cba2043937fd972b4e31402baa5b55721eb19a7bb8ed1d08306a2","target":"graph","created_at":"2026-07-21T01:21:56Z","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/2607.17644/integrity.json","findings":[],"snapshot_sha256":"c28c3603d3b5d939e8dc4c7e95fa8dfce3d595e45f758748cecf8e644a296938","summary":{"advisory":0,"by_detector":{},"critical":0,"informational":0}},"paper":{"abstract_excerpt":"Multi-head Latent Attention (MLA) ships two implementations in Megatron-Core: an explicit form used for training and an absorbed form -- which slashes collective communication by gathering only the compressed latent -- that is fully implemented but hard-asserted out of training (the forward opens with \"assert not (self.training and self.cache_mla_latents)\"), allowed only in inference decode. The library documents no reason. We show the restriction is well-founded and quantify why: ported to training, the absorbed form is a memory trap -- its intermediates live in n_h x d_kv dimensions per toke","authors_text":"Changzheng Ma","cross_cats":[],"headline":"","license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.DC","submitted_at":"2026-07-20T07:57:34Z","title":"A Training-Memory Regression in MLA Sequence Parallelism: Why Megatron-Core Forbids Absorption, and LAGA -- a Communication-Efficient Fix"},"references":{"count":0,"internal_anchors":0,"resolved_work":0,"sample":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2607.17644","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:f6fefef533099f7fa8a5d78182ecd2e06ececf0d3456f81a43a2be0b091af5b9","target":"record","created_at":"2026-07-21T01:21:56Z","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":"998e7d6fc4ec61e701d48cae1b3adbda0d3f59e97e83ef7002eba2969d30a6e7","cross_cats_sorted":[],"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.DC","submitted_at":"2026-07-20T07:57:34Z","title_canon_sha256":"dc371da43d7bcf12d62cd3e75667e8a06c1257af02b88e29d4f17399c85d0342"},"schema_version":"1.0","source":{"id":"2607.17644","kind":"arxiv","version":1}},"canonical_sha256":"c2a84f7c03f5692be76171ef543a59b867cf0f1f61de72607382056539f910ed","receipt":{"algorithm":"ed25519","builder_version":"pith-number-builder-2026-05-17-v1","canonical_sha256":"c2a84f7c03f5692be76171ef543a59b867cf0f1f61de72607382056539f910ed","first_computed_at":"2026-07-21T01:21:56.872011Z","key_id":"pith-v1-2026-05","kind":"pith_receipt","last_reissued_at":"2026-07-21T01:21:56.872011Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","receipt_version":"0.3","signature_b64":"Vtpido0pshg/8DOs8rPJLg4Om6nZBXGHM3v5G/4E1wfhienRLmeoi9chJrLz0mS5mEJ2+3F/knMQiorV4640Cw==","signature_status":"signed_v1","signed_at":"2026-07-21T01:21:56.872778Z","signed_message":"canonical_sha256_bytes"},"source_id":"2607.17644","source_kind":"arxiv","source_version":1}}},"equivocations":[],"invalid_events":[],"applied_event_ids":["sha256:f6fefef533099f7fa8a5d78182ecd2e06ececf0d3456f81a43a2be0b091af5b9","sha256:a48a104a590cba2043937fd972b4e31402baa5b55721eb19a7bb8ed1d08306a2"],"state_sha256":"93cf098baf124f3e5cfadd872b9d4f9ea620d2710ed14ef52c01acc67dc2b236"}