{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2024:PYSQMLKLTNI6JHZDDLSKCJ2GCD","short_pith_number":"pith:PYSQMLKL","schema_version":"1.0","canonical_sha256":"7e25062d4b9b51e49f231ae4a1274610f14c152e5864bbdd2fda2b1fc01409d2","source":{"kind":"arxiv","id":"2410.06270","version":2},"attestation_state":"computed","paper":{"title":"Mixture Compressor for Mixture-of-Experts LLMs Gains More","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.CL"],"primary_cat":"cs.LG","authors_text":"Haoru Tan, Hongsheng Li, Jianhui Liu, Ruifei He, Shiming Zhang, Si Liu, Wei Huang, Xiaojuan Qi, Yue Liao","submitted_at":"2024-10-08T18:09:38Z","abstract_excerpt":"Mixture-of-Experts large language models (MoE-LLMs) marks a significant step forward of language models, however, they encounter two critical challenges in practice: 1) expert parameters lead to considerable memory consumption and loading latency; and 2) the current activated experts are redundant, as many tokens may only require a single expert. Motivated by these issues, we investigate the MoE-LLMs and make two key observations: a) different experts exhibit varying behaviors on activation reconstruction error, routing scores, and activated frequencies, highlighting their differing importance"},"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":"2410.06270","kind":"arxiv","version":2},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.LG","submitted_at":"2024-10-08T18:09:38Z","cross_cats_sorted":["cs.CL"],"title_canon_sha256":"62a25f9572ee2908ac640f479ca8f21479eda9a124774e60d5553fff73bb96e1","abstract_canon_sha256":"00a49300eb5fc1706206349c5e5468c3a348374ec4c079ffcb5e38941321bc8d"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T10:18:10.444814Z","signature_b64":"Zmc+A4dDeNy4tKlb3JbT2LsG0+ut0Wtj3GT+0+vj/yqOZPR2ok0MmjUl7t5k7FEKVGvfyqHmAmI8DkcKAVcOAw==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"7e25062d4b9b51e49f231ae4a1274610f14c152e5864bbdd2fda2b1fc01409d2","last_reissued_at":"2026-07-05T10:18:10.444312Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T10:18:10.444312Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Mixture Compressor for Mixture-of-Experts LLMs Gains More","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.CL"],"primary_cat":"cs.LG","authors_text":"Haoru Tan, Hongsheng Li, Jianhui Liu, Ruifei He, Shiming Zhang, Si Liu, Wei Huang, Xiaojuan Qi, Yue Liao","submitted_at":"2024-10-08T18:09:38Z","abstract_excerpt":"Mixture-of-Experts large language models (MoE-LLMs) marks a significant step forward of language models, however, they encounter two critical challenges in practice: 1) expert parameters lead to considerable memory consumption and loading latency; and 2) the current activated experts are redundant, as many tokens may only require a single expert. Motivated by these issues, we investigate the MoE-LLMs and make two key observations: a) different experts exhibit varying behaviors on activation reconstruction error, routing scores, and activated frequencies, highlighting their differing importance"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2410.06270","kind":"arxiv","version":2},"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/2410.06270/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":"2410.06270","created_at":"2026-07-05T10:18:10.444381+00:00"},{"alias_kind":"arxiv_version","alias_value":"2410.06270v2","created_at":"2026-07-05T10:18:10.444381+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2410.06270","created_at":"2026-07-05T10:18:10.444381+00:00"},{"alias_kind":"pith_short_12","alias_value":"PYSQMLKLTNI6","created_at":"2026-07-05T10:18:10.444381+00:00"},{"alias_kind":"pith_short_16","alias_value":"PYSQMLKLTNI6JHZD","created_at":"2026-07-05T10:18:10.444381+00:00"},{"alias_kind":"pith_short_8","alias_value":"PYSQMLKL","created_at":"2026-07-05T10:18:10.444381+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":10,"internal_anchor_count":0,"sample":[{"citing_arxiv_id":"2604.22748","citing_title":"Agentic World Modeling: Foundations, Capabilities, Laws, and Beyond","ref_index":150,"is_internal_anchor":false},{"citing_arxiv_id":"2606.18105","citing_title":"OmniPlan: An Adaptive Framework for Timely and Near-Optimal Network Planning Optimization","ref_index":19,"is_internal_anchor":false},{"citing_arxiv_id":"2606.18304","citing_title":"Attribution-Guided and Coverage-Maximized Pruning for Structural MoE Compression","ref_index":8,"is_internal_anchor":false},{"citing_arxiv_id":"2605.23078","citing_title":"GEMQ: Global Expert-Level Mixed-Precision Quantization for MoE LLMs","ref_index":14,"is_internal_anchor":false},{"citing_arxiv_id":"2605.16349","citing_title":"Geometric Asymmetry in MoE Specialization: Functional Decorrelation and Representational Overlap","ref_index":11,"is_internal_anchor":false},{"citing_arxiv_id":"2605.18739","citing_title":"LongLive-2.0: An NVFP4 Parallel Infrastructure for Long Video Generation","ref_index":25,"is_internal_anchor":false},{"citing_arxiv_id":"2603.06003","citing_title":"EvoESAP: Non-Uniform Expert Pruning for Sparse MoE","ref_index":21,"is_internal_anchor":false},{"citing_arxiv_id":"2604.02715","citing_title":"FluxMoE: Decoupling Expert Residency for High-Performance MoE Serving","ref_index":22,"is_internal_anchor":false},{"citing_arxiv_id":"2604.22748","citing_title":"Agentic World Modeling: Foundations, Capabilities, Laws, and Beyond","ref_index":150,"is_internal_anchor":false},{"citing_arxiv_id":"2605.05225","citing_title":"MACS: Modality-Aware Capacity Scaling for Efficient Multimodal MoE Inference","ref_index":14,"is_internal_anchor":false}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/PYSQMLKLTNI6JHZDDLSKCJ2GCD","json":"https://pith.science/pith/PYSQMLKLTNI6JHZDDLSKCJ2GCD.json","graph_json":"https://pith.science/api/pith-number/PYSQMLKLTNI6JHZDDLSKCJ2GCD/graph.json","events_json":"https://pith.science/api/pith-number/PYSQMLKLTNI6JHZDDLSKCJ2GCD/events.json","paper":"https://pith.science/paper/PYSQMLKL"},"agent_actions":{"view_html":"https://pith.science/pith/PYSQMLKLTNI6JHZDDLSKCJ2GCD","download_json":"https://pith.science/pith/PYSQMLKLTNI6JHZDDLSKCJ2GCD.json","view_paper":"https://pith.science/paper/PYSQMLKL","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2410.06270&json=true","fetch_graph":"https://pith.science/api/pith-number/PYSQMLKLTNI6JHZDDLSKCJ2GCD/graph.json","fetch_events":"https://pith.science/api/pith-number/PYSQMLKLTNI6JHZDDLSKCJ2GCD/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/PYSQMLKLTNI6JHZDDLSKCJ2GCD/action/timestamp_anchor","attest_storage":"https://pith.science/pith/PYSQMLKLTNI6JHZDDLSKCJ2GCD/action/storage_attestation","attest_author":"https://pith.science/pith/PYSQMLKLTNI6JHZDDLSKCJ2GCD/action/author_attestation","sign_citation":"https://pith.science/pith/PYSQMLKLTNI6JHZDDLSKCJ2GCD/action/citation_signature","submit_replication":"https://pith.science/pith/PYSQMLKLTNI6JHZDDLSKCJ2GCD/action/replication_record"}},"created_at":"2026-07-05T10:18:10.444381+00:00","updated_at":"2026-07-05T10:18:10.444381+00:00"}