{"bundle_type":"pith_open_graph_bundle","bundle_version":"1.0","pith_number":"pith:2025:DWDDCERYNRLWLCCIJPYJYWFRNJ","short_pith_number":"pith:DWDDCERY","canonical_record":{"source":{"id":"2504.03871","kind":"arxiv","version":1},"metadata":{"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.DC","submitted_at":"2025-04-04T18:55:52Z","cross_cats_sorted":["cs.LG"],"title_canon_sha256":"e8e18e3810a2fd09d8ba2290539816f63782e77d1f59f51dbef4abdb4ac66a8b","abstract_canon_sha256":"e1c517fb921c983e20754f72765af064ecec621c1b609b9144cbfb6b16e63168"},"schema_version":"1.0"},"canonical_sha256":"1d863112386c576588484bf09c58b16a598053d4d4dce1a757a36ca1f64b56de","source":{"kind":"arxiv","id":"2504.03871","version":1},"source_aliases":[{"alias_kind":"arxiv","alias_value":"2504.03871","created_at":"2026-07-05T10:44:50Z"},{"alias_kind":"arxiv_version","alias_value":"2504.03871v1","created_at":"2026-07-05T10:44:50Z"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2504.03871","created_at":"2026-07-05T10:44:50Z"},{"alias_kind":"pith_short_12","alias_value":"DWDDCERYNRLW","created_at":"2026-07-05T10:44:50Z"},{"alias_kind":"pith_short_16","alias_value":"DWDDCERYNRLWLCCI","created_at":"2026-07-05T10:44:50Z"},{"alias_kind":"pith_short_8","alias_value":"DWDDCERY","created_at":"2026-07-05T10:44:50Z"}],"events":[{"event_type":"record_created","subject_pith_number":"pith:2025:DWDDCERYNRLWLCCIJPYJYWFRNJ","target":"record","payload":{"canonical_record":{"source":{"id":"2504.03871","kind":"arxiv","version":1},"metadata":{"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.DC","submitted_at":"2025-04-04T18:55:52Z","cross_cats_sorted":["cs.LG"],"title_canon_sha256":"e8e18e3810a2fd09d8ba2290539816f63782e77d1f59f51dbef4abdb4ac66a8b","abstract_canon_sha256":"e1c517fb921c983e20754f72765af064ecec621c1b609b9144cbfb6b16e63168"},"schema_version":"1.0"},"canonical_sha256":"1d863112386c576588484bf09c58b16a598053d4d4dce1a757a36ca1f64b56de","receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T10:44:50.080229Z","signature_b64":"qMB6uEi9y1JfHze8sO16A074+HhDRIf/sk94vNbyZu8RDAbp/C47yQHUDywMW5S9/0m72KITkHez+N3zBT+DDw==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"1d863112386c576588484bf09c58b16a598053d4d4dce1a757a36ca1f64b56de","last_reissued_at":"2026-07-05T10:44:50.079737Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T10:44:50.079737Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"source_kind":"arxiv","source_id":"2504.03871","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":"2026-07-05T10:44:50Z","supersedes":[],"prev_event":null,"signature":{"signature_status":"signed_v1","algorithm":"ed25519","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","signature_b64":"RqRtuukweR/7V929AQXnEksHo0RXM+ZQ9ieAm+sE9AJvVZapyPSlVGFulwhG625SCEgMgCqORJJd/eG9IwMaDg==","signed_message":"open_graph_event_sha256_bytes","signed_at":"2026-08-03T21:13:25.724674Z"},"content_sha256":"a98fd5031c7bd5bdd8f6c9c87569986100770d9d7f0bffdf2252f961eab40b43","schema_version":"1.0","event_id":"sha256:a98fd5031c7bd5bdd8f6c9c87569986100770d9d7f0bffdf2252f961eab40b43"},{"event_type":"graph_snapshot","subject_pith_number":"pith:2025:DWDDCERYNRLWLCCIJPYJYWFRNJ","target":"graph","payload":{"graph_snapshot":{"paper":{"title":"HeterMoE: Efficient Training of Mixture-of-Experts Models on Heterogeneous GPUs","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":["cs.LG"],"primary_cat":"cs.DC","authors_text":"Ceyu Xu, Danyang Zhuo, Feng Qian, Ion Stoica, Matthew Lentz, Shuowei Jin, Xueshen Liu, Yongji Wu, Z. Morley Mao","submitted_at":"2025-04-04T18:55:52Z","abstract_excerpt":"The Mixture-of-Experts (MoE) architecture has become increasingly popular as a method to scale up large language models (LLMs). To save costs, heterogeneity-aware training solutions have been proposed to utilize GPU clusters made up of both newer and older-generation GPUs. However, existing solutions are agnostic to the performance characteristics of different MoE model components (i.e., attention and expert) and do not fully utilize each GPU's compute capability.\n  In this paper, we introduce HeterMoE, a system to efficiently train MoE models on heterogeneous GPUs. Our key insight is that new"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2504.03871","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/2504.03871/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":"2026-07-05T10:44:50Z","supersedes":[],"prev_event":null,"signature":{"signature_status":"signed_v1","algorithm":"ed25519","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","signature_b64":"3ht5hGQbWMJukGqVJYrCqWh5GjDUB4QSrGxKnf8gbURvh2f6AryNt2ClmYJdPC65S3urAib/UqMQsRL3PMfTBg==","signed_message":"open_graph_event_sha256_bytes","signed_at":"2026-08-03T21:13:25.725221Z"},"content_sha256":"21f1365b4aae97186f8280aecf6aa0217ccb2871c5cc020ab29a78a5421a0d4d","schema_version":"1.0","event_id":"sha256:21f1365b4aae97186f8280aecf6aa0217ccb2871c5cc020ab29a78a5421a0d4d"}],"timestamp_proofs":[],"mirror_hints":[{"mirror_type":"https","name":"Pith Resolver","base_url":"https://pith.science","bundle_url":"https://pith.science/pith/DWDDCERYNRLWLCCIJPYJYWFRNJ/bundle.json","state_url":"https://pith.science/pith/DWDDCERYNRLWLCCIJPYJYWFRNJ/state.json","well_known_bundle_url":"https://pith.science/.well-known/pith/DWDDCERYNRLWLCCIJPYJYWFRNJ/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":"2026-08-03T21:13:25Z","links":{"resolver":"https://pith.science/pith/DWDDCERYNRLWLCCIJPYJYWFRNJ","bundle":"https://pith.science/pith/DWDDCERYNRLWLCCIJPYJYWFRNJ/bundle.json","state":"https://pith.science/pith/DWDDCERYNRLWLCCIJPYJYWFRNJ/state.json","well_known_bundle":"https://pith.science/.well-known/pith/DWDDCERYNRLWLCCIJPYJYWFRNJ/bundle.json"},"state":{"state_type":"pith_open_graph_state","state_version":"1.0","pith_number":"pith:2025:DWDDCERYNRLWLCCIJPYJYWFRNJ","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":"e1c517fb921c983e20754f72765af064ecec621c1b609b9144cbfb6b16e63168","cross_cats_sorted":["cs.LG"],"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.DC","submitted_at":"2025-04-04T18:55:52Z","title_canon_sha256":"e8e18e3810a2fd09d8ba2290539816f63782e77d1f59f51dbef4abdb4ac66a8b"},"schema_version":"1.0","source":{"id":"2504.03871","kind":"arxiv","version":1}},"source_aliases":[{"alias_kind":"arxiv","alias_value":"2504.03871","created_at":"2026-07-05T10:44:50Z"},{"alias_kind":"arxiv_version","alias_value":"2504.03871v1","created_at":"2026-07-05T10:44:50Z"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2504.03871","created_at":"2026-07-05T10:44:50Z"},{"alias_kind":"pith_short_12","alias_value":"DWDDCERYNRLW","created_at":"2026-07-05T10:44:50Z"},{"alias_kind":"pith_short_16","alias_value":"DWDDCERYNRLWLCCI","created_at":"2026-07-05T10:44:50Z"},{"alias_kind":"pith_short_8","alias_value":"DWDDCERY","created_at":"2026-07-05T10:44:50Z"}],"graph_snapshots":[{"event_id":"sha256:21f1365b4aae97186f8280aecf6aa0217ccb2871c5cc020ab29a78a5421a0d4d","target":"graph","created_at":"2026-07-05T10:44:50Z","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/2504.03871/integrity.json","findings":[],"snapshot_sha256":"c28c3603d3b5d939e8dc4c7e95fa8dfce3d595e45f758748cecf8e644a296938","summary":{"advisory":0,"by_detector":{},"critical":0,"informational":0}},"paper":{"abstract_excerpt":"The Mixture-of-Experts (MoE) architecture has become increasingly popular as a method to scale up large language models (LLMs). To save costs, heterogeneity-aware training solutions have been proposed to utilize GPU clusters made up of both newer and older-generation GPUs. However, existing solutions are agnostic to the performance characteristics of different MoE model components (i.e., attention and expert) and do not fully utilize each GPU's compute capability.\n  In this paper, we introduce HeterMoE, a system to efficiently train MoE models on heterogeneous GPUs. Our key insight is that new","authors_text":"Ceyu Xu, Danyang Zhuo, Feng Qian, Ion Stoica, Matthew Lentz, Shuowei Jin, Xueshen Liu, Yongji Wu, Z. Morley Mao","cross_cats":["cs.LG"],"headline":"","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.DC","submitted_at":"2025-04-04T18:55:52Z","title":"HeterMoE: Efficient Training of Mixture-of-Experts Models on Heterogeneous GPUs"},"references":{"count":0,"internal_anchors":0,"resolved_work":0,"sample":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2504.03871","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:a98fd5031c7bd5bdd8f6c9c87569986100770d9d7f0bffdf2252f961eab40b43","target":"record","created_at":"2026-07-05T10:44:50Z","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":"e1c517fb921c983e20754f72765af064ecec621c1b609b9144cbfb6b16e63168","cross_cats_sorted":["cs.LG"],"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.DC","submitted_at":"2025-04-04T18:55:52Z","title_canon_sha256":"e8e18e3810a2fd09d8ba2290539816f63782e77d1f59f51dbef4abdb4ac66a8b"},"schema_version":"1.0","source":{"id":"2504.03871","kind":"arxiv","version":1}},"canonical_sha256":"1d863112386c576588484bf09c58b16a598053d4d4dce1a757a36ca1f64b56de","receipt":{"algorithm":"ed25519","builder_version":"pith-number-builder-2026-05-17-v1","canonical_sha256":"1d863112386c576588484bf09c58b16a598053d4d4dce1a757a36ca1f64b56de","first_computed_at":"2026-07-05T10:44:50.079737Z","key_id":"pith-v1-2026-05","kind":"pith_receipt","last_reissued_at":"2026-07-05T10:44:50.079737Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","receipt_version":"0.3","signature_b64":"qMB6uEi9y1JfHze8sO16A074+HhDRIf/sk94vNbyZu8RDAbp/C47yQHUDywMW5S9/0m72KITkHez+N3zBT+DDw==","signature_status":"signed_v1","signed_at":"2026-07-05T10:44:50.080229Z","signed_message":"canonical_sha256_bytes"},"source_id":"2504.03871","source_kind":"arxiv","source_version":1}}},"equivocations":[],"invalid_events":[],"applied_event_ids":["sha256:a98fd5031c7bd5bdd8f6c9c87569986100770d9d7f0bffdf2252f961eab40b43","sha256:21f1365b4aae97186f8280aecf6aa0217ccb2871c5cc020ab29a78a5421a0d4d"],"state_sha256":"3623a4a913273302383a441c3fda0ea1c65eaeead9aeae8960778e2f4e8f9680"},"bundle_signature":{"signature_status":"signed_v1","algorithm":"ed25519","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","signature_b64":"r3WZx4Mq2PWYcpfJBP/q/+XdjLYJMXCw3aNdH7/4GGzw1HS2Gs0gc4MySdU/eLHoCOvXDZxKx+3Cqx7adZvXCg==","signed_message":"bundle_sha256_bytes","signed_at":"2026-08-03T21:13:25.729359Z","bundle_sha256":"f29709deb703e9f032c6ba91894096923eef8ad2432e4a0415c65f6adfbff07b"}}