{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2026:CRNJ6PUBUR2WRTHUSPSBXS77CP","short_pith_number":"pith:CRNJ6PUB","schema_version":"1.0","canonical_sha256":"145a9f3e81a47568ccf493e41bcbff13d814f17cc9398319b95fe63ec2e39d45","source":{"kind":"arxiv","id":"2607.17880","version":1},"attestation_state":"computed","paper":{"title":"Mixture-of-Experts Serving","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":[],"primary_cat":"cs.DS","authors_text":"Qinpei Lou, Tao Xiao, Zhiyi Huang","submitted_at":"2026-07-20T12:30:27Z","abstract_excerpt":"Mixture-of-Experts (MoE) models route each token to only a few expert networks, distributing the serving load across experts whose popularity shifts over time. A serving system must therefore dynamically decide how many GPUs to assign to each expert, trading off service latency against the cost of reconfiguring the assignment. We introduce a formal model of MoE Serving and initiate a principled study of online and offline algorithms for it. Our main result is a polynomial-time $O(\\sqrt{\\log k})$-competitive online algorithm, where $k$ is the number of GPUs beyond one per expert. We complement "},"verification_status":{"content_addressed":true,"pith_receipt":true,"author_attested":false,"weak_author_claims":0,"strong_author_claims":0,"externally_anchored":false,"storage_verified":false,"citation_signatures":0,"replication_records":0,"graph_snapshot":true,"references_resolved":false,"formal_links_present":false},"canonical_record":{"source":{"id":"2607.17880","kind":"arxiv","version":1},"metadata":{"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.DS","submitted_at":"2026-07-20T12:30:27Z","cross_cats_sorted":[],"title_canon_sha256":"30e9caade90531568340affba5b57bb3e6e783e1cf506f0479cd3aa4afb2b2bc","abstract_canon_sha256":"8a2679f4dd647c1d946a41b0add32177123ca28cadfc1dab3d5f8be0ed6417ec"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-21T02:22:04.779771Z","signature_b64":"m+jdXE5g1f4na+DE+DvGHZqrR/zfcH69JlMUWlos/+8Rt+/KXmULI0xFz+6a6ilpP3tmtYcmqIBeSeX7nS/WCA==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"145a9f3e81a47568ccf493e41bcbff13d814f17cc9398319b95fe63ec2e39d45","last_reissued_at":"2026-07-21T02:22:04.778981Z","signature_status":"signed_v1","first_computed_at":"2026-07-21T02:22:04.778981Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Mixture-of-Experts Serving","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":[],"primary_cat":"cs.DS","authors_text":"Qinpei Lou, Tao Xiao, Zhiyi Huang","submitted_at":"2026-07-20T12:30:27Z","abstract_excerpt":"Mixture-of-Experts (MoE) models route each token to only a few expert networks, distributing the serving load across experts whose popularity shifts over time. A serving system must therefore dynamically decide how many GPUs to assign to each expert, trading off service latency against the cost of reconfiguring the assignment. We introduce a formal model of MoE Serving and initiate a principled study of online and offline algorithms for it. Our main result is a polynomial-time $O(\\sqrt{\\log k})$-competitive online algorithm, where $k$ is the number of GPUs beyond one per expert. We complement "},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2607.17880","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/2607.17880/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"},"aliases":[{"alias_kind":"arxiv","alias_value":"2607.17880","created_at":"2026-07-21T02:22:04.779393+00:00"},{"alias_kind":"arxiv_version","alias_value":"2607.17880v1","created_at":"2026-07-21T02:22:04.779393+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2607.17880","created_at":"2026-07-21T02:22:04.779393+00:00"},{"alias_kind":"pith_short_12","alias_value":"CRNJ6PUBUR2W","created_at":"2026-07-21T02:22:04.779393+00:00"},{"alias_kind":"pith_short_16","alias_value":"CRNJ6PUBUR2WRTHU","created_at":"2026-07-21T02:22:04.779393+00:00"},{"alias_kind":"pith_short_8","alias_value":"CRNJ6PUB","created_at":"2026-07-21T02:22:04.779393+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":1,"internal_anchor_count":1,"sample":[{"citing_arxiv_id":"2608.13057","citing_title":"TEMPO: Makespan-Aware Expert-Parallel Load Balancing Across Memory- and Compute-Bound Regimes","ref_index":11,"is_internal_anchor":true}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/CRNJ6PUBUR2WRTHUSPSBXS77CP","json":"https://pith.science/pith/CRNJ6PUBUR2WRTHUSPSBXS77CP.json","graph_json":"https://pith.science/api/pith-number/CRNJ6PUBUR2WRTHUSPSBXS77CP/graph.json","events_json":"https://pith.science/api/pith-number/CRNJ6PUBUR2WRTHUSPSBXS77CP/events.json","paper":"https://pith.science/paper/CRNJ6PUB"},"agent_actions":{"view_html":"https://pith.science/pith/CRNJ6PUBUR2WRTHUSPSBXS77CP","download_json":"https://pith.science/pith/CRNJ6PUBUR2WRTHUSPSBXS77CP.json","view_paper":"https://pith.science/paper/CRNJ6PUB","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2607.17880&json=true","fetch_graph":"https://pith.science/api/pith-number/CRNJ6PUBUR2WRTHUSPSBXS77CP/graph.json","fetch_events":"https://pith.science/api/pith-number/CRNJ6PUBUR2WRTHUSPSBXS77CP/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/CRNJ6PUBUR2WRTHUSPSBXS77CP/action/timestamp_anchor","attest_storage":"https://pith.science/pith/CRNJ6PUBUR2WRTHUSPSBXS77CP/action/storage_attestation","attest_author":"https://pith.science/pith/CRNJ6PUBUR2WRTHUSPSBXS77CP/action/author_attestation","sign_citation":"https://pith.science/pith/CRNJ6PUBUR2WRTHUSPSBXS77CP/action/citation_signature","submit_replication":"https://pith.science/pith/CRNJ6PUBUR2WRTHUSPSBXS77CP/action/replication_record"}},"created_at":"2026-07-21T02:22:04.779393+00:00","updated_at":"2026-07-21T02:22:04.779393+00:00"}