{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2024:XZ6KU673RXJCDFKJNQGACJNLRV","short_pith_number":"pith:XZ6KU673","schema_version":"1.0","canonical_sha256":"be7caa7bfb8dd22195496c0c0125ab8d718011f0f205abb2232fbcaf12a94cf6","source":{"kind":"arxiv","id":"2408.15664","version":1},"attestation_state":"computed","paper":{"title":"Auxiliary-Loss-Free Load Balancing Strategy for Mixture-of-Experts","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"Mixture-of-Experts models reach higher performance with load balancing that avoids auxiliary loss gradients.","cross_cats":["cs.CL"],"primary_cat":"cs.LG","authors_text":"Chenggang Zhao, Damai Dai, Huazuo Gao, Lean Wang, Xu Sun","submitted_at":"2024-08-28T09:31:09Z","abstract_excerpt":"For Mixture-of-Experts (MoE) models, an unbalanced expert load will lead to routing collapse or increased computational overhead. Existing methods commonly employ an auxiliary loss to encourage load balance, but a large auxiliary loss will introduce non-negligible interference gradients into training and thus impair the model performance. In order to control load balance while not producing undesired gradients during training, we propose Loss-Free Balancing, featured by an auxiliary-loss-free load balancing strategy. To be specific, before the top-K routing decision, Loss-Free Balancing will f"},"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":true,"formal_links_present":true},"canonical_record":{"source":{"id":"2408.15664","kind":"arxiv","version":1},"metadata":{"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.LG","submitted_at":"2024-08-28T09:31:09Z","cross_cats_sorted":["cs.CL"],"title_canon_sha256":"2ba3f20877dadd522332bfbf50455cc66980575dd5bbc31e024cf536d29778ff","abstract_canon_sha256":"de6550001f65ea19f000f8d635df0f482bea715eade709c791379eef9256cdc5"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-05-17T23:38:52.512176Z","signature_b64":"abHUXc0+nMOVci7mAwA8Ws1A0puaAk9b/wdZ9s1pIhqDSJV+b4lyfQpKD9+BIZ8Y4mimhyzkIYZThaggbRIrAg==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"be7caa7bfb8dd22195496c0c0125ab8d718011f0f205abb2232fbcaf12a94cf6","last_reissued_at":"2026-05-17T23:38:52.511754Z","signature_status":"signed_v1","first_computed_at":"2026-05-17T23:38:52.511754Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Auxiliary-Loss-Free Load Balancing Strategy for Mixture-of-Experts","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"Mixture-of-Experts models reach higher performance with load balancing that avoids auxiliary loss gradients.","cross_cats":["cs.CL"],"primary_cat":"cs.LG","authors_text":"Chenggang Zhao, Damai Dai, Huazuo Gao, Lean Wang, Xu Sun","submitted_at":"2024-08-28T09:31:09Z","abstract_excerpt":"For Mixture-of-Experts (MoE) models, an unbalanced expert load will lead to routing collapse or increased computational overhead. Existing methods commonly employ an auxiliary loss to encourage load balance, but a large auxiliary loss will introduce non-negligible interference gradients into training and thus impair the model performance. In order to control load balance while not producing undesired gradients during training, we propose Loss-Free Balancing, featured by an auxiliary-loss-free load balancing strategy. To be specific, before the top-K routing decision, Loss-Free Balancing will f"},"claims":{"count":4,"items":[{"kind":"strongest_claim","text":"Loss-Free Balancing achieves both better performance and better load balance compared with traditional auxiliary-loss-controlled load balancing strategies.","source":"verdict.strongest_claim","status":"machine_extracted","claim_id":"C1","attestation":"unclaimed"},{"kind":"weakest_assumption","text":"That dynamically updating per-expert biases from recent load statistics will maintain balance across training without introducing instability or unintended routing dynamics.","source":"verdict.weakest_assumption","status":"machine_extracted","claim_id":"C2","attestation":"unclaimed"},{"kind":"one_line_summary","text":"Loss-Free Balancing keeps expert loads balanced in MoE models by dynamically adjusting routing-score biases based on recent usage, avoiding auxiliary-loss interference and yielding better performance.","source":"verdict.one_line_summary","status":"machine_extracted","claim_id":"C3","attestation":"unclaimed"},{"kind":"headline","text":"Mixture-of-Experts models reach higher performance with load balancing that avoids auxiliary loss gradients.","source":"verdict.pith_extraction.headline","status":"machine_extracted","claim_id":"C4","attestation":"unclaimed"}],"snapshot_sha256":"c306008b14ccbd0298b01962067596270de956a35d5f098978a086983d20b104"},"source":{"id":"2408.15664","kind":"arxiv","version":1},"verdict":{"id":"9ed35d07-9958-4ddc-9949-761036dfa418","model_set":{"reader":"grok-4.3"},"created_at":"2026-05-15T12:50:11.218526Z","strongest_claim":"Loss-Free Balancing achieves both better performance and better load balance compared with traditional auxiliary-loss-controlled load balancing strategies.","one_line_summary":"Loss-Free Balancing keeps expert loads balanced in MoE models by dynamically adjusting routing-score biases based on recent usage, avoiding auxiliary-loss interference and yielding better performance.","pipeline_version":"pith-pipeline@v0.9.0","weakest_assumption":"That dynamically updating per-expert biases from recent load statistics will maintain balance across training without introducing instability or unintended routing dynamics.","pith_extraction_headline":"Mixture-of-Experts models reach higher performance with load balancing that avoids auxiliary loss gradients."},"references":{"count":32,"sample":[{"doi":"","year":2024,"title":"DeepSeekMoE: Towards Ultimate Expert Specialization in Mixture-of-Experts Language Models","work_id":"a9888d6d-bf47-4324-9834-7cc12ac3a78c","ref_index":1,"cited_arxiv_id":"2401.06066","is_internal_anchor":true},{"doi":"","year":2024,"title":"DeepSeek-Coder-V2: Breaking the Barrier of Closed-Source Models in Code Intelligence","work_id":"713a39be-8c4e-4ee3-8671-11cf9379262f","ref_index":2,"cited_arxiv_id":"2406.11931","is_internal_anchor":true},{"doi":"","year":2021,"title":"William Fedus, Barret Zoph, and Noam M. Shazeer. Switch transformers: Scaling to trillion parameter models with simple and efficient sparsity. J. Mach. Learn. Res., 23: 0 120:1--120:39, 2021. URL http","work_id":"2f29f1ca-bd51-4381-acc7-95e946cde35f","ref_index":3,"cited_arxiv_id":"","is_internal_anchor":false},{"doi":"","year":2006,"title":"GShard: Scaling Giant Models with Conditional Computation and Automatic Sharding","work_id":"52b3c9a6-2a27-45a7-ba2b-ebe4b5bb5a5f","ref_index":4,"cited_arxiv_id":"2006.16668","is_internal_anchor":true},{"doi":"","year":2016,"title":"Sgdr: Stochastic gradient descent with warm restarts","work_id":"21d4473d-9c67-4464-b3ea-96588e1d01e1","ref_index":5,"cited_arxiv_id":"","is_internal_anchor":false}],"resolved_work":32,"snapshot_sha256":"45ecc5e36e60cebbf268394ae511828f42a1187c8166048c5f0e7f95a94d51e0","internal_anchors":6},"formal_canon":{"evidence_count":2,"snapshot_sha256":"ae8508b9f31a7ff7425541690bf1452fc90bde2e6779ab251d817b0d51cca3d1"},"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":"2408.15664","created_at":"2026-05-17T23:38:52.511813+00:00"},{"alias_kind":"arxiv_version","alias_value":"2408.15664v1","created_at":"2026-05-17T23:38:52.511813+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2408.15664","created_at":"2026-05-17T23:38:52.511813+00:00"},{"alias_kind":"pith_short_12","alias_value":"XZ6KU673RXJC","created_at":"2026-05-18T12:33:37.589309+00:00"},{"alias_kind":"pith_short_16","alias_value":"XZ6KU673RXJCDFKJ","created_at":"2026-05-18T12:33:37.589309+00:00"},{"alias_kind":"pith_short_8","alias_value":"XZ6KU673","created_at":"2026-05-18T12:33:37.589309+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":39,"internal_anchor_count":39,"sample":[{"citing_arxiv_id":"2512.23070","citing_title":"FLEX-MoE: Federated Mixture-of-Experts with Load-balanced Expert Assignment for Edge Computing","ref_index":9,"is_internal_anchor":true},{"citing_arxiv_id":"2601.18664","citing_title":"S$^2$GR: Stepwise Semantic-Guided Reasoning in Latent Space for Generative Recommendation","ref_index":38,"is_internal_anchor":true},{"citing_arxiv_id":"2605.20247","citing_title":"CP-MoE: Consistency-Preserving Mixture-of-Experts for Continual Learning","ref_index":13,"is_internal_anchor":true},{"citing_arxiv_id":"2605.10933","citing_title":"DECO: Sparse Mixture-of-Experts with Dense-Comparable Performance on End-Side Devices","ref_index":174,"is_internal_anchor":true},{"citing_arxiv_id":"2605.15676","citing_title":"Dynamic Chunking for Diffusion Language Models","ref_index":42,"is_internal_anchor":true},{"citing_arxiv_id":"2605.16690","citing_title":"UB-SMoE: Universally Balanced Sparse Mixture-of-Experts for Resource-adaptive Federated Fine-tuning of Foundation Models","ref_index":49,"is_internal_anchor":true},{"citing_arxiv_id":"2605.19945","citing_title":"GEM: GPU-Variability-Aware Expert to GPU Mapping for MoE Systems","ref_index":16,"is_internal_anchor":true},{"citing_arxiv_id":"2605.17879","citing_title":"Guard: Scalable Straggler Detection and Node Health Management for Large-Scale Training","ref_index":11,"is_internal_anchor":true},{"citing_arxiv_id":"2605.15403","citing_title":"$\\phi$-Balancing for Mixture-of-Experts Training","ref_index":40,"is_internal_anchor":true},{"citing_arxiv_id":"2605.15484","citing_title":"When Does Sparse MoE Help in Vision? The Role of Backbone Compute Leverage in Sparse Routing","ref_index":27,"is_internal_anchor":true},{"citing_arxiv_id":"2505.24437","citing_title":"SwitchCodec: A High-Fidelity Nerual Audio Codec With Sparse Quantization","ref_index":41,"is_internal_anchor":true},{"citing_arxiv_id":"2505.19525","citing_title":"Rethinking Gating Mechanism in Sparse MoE: Handling Arbitrary Modality Inputs with Confidence-Guided Gate","ref_index":30,"is_internal_anchor":true},{"citing_arxiv_id":"2509.21892","citing_title":"Elastic MoE: Unlocking the Inference-Time Scalability of Mixture-of-Experts","ref_index":36,"is_internal_anchor":true},{"citing_arxiv_id":"2510.04800","citing_title":"Hybrid Architectures for Language Models: Systematic Analysis and Design Insights","ref_index":56,"is_internal_anchor":true},{"citing_arxiv_id":"2512.03915","citing_title":"A Theoretical Framework for Auxiliary-Loss-Free Load Balancing of Sparse Mixture-of-Experts in Large-Scale AI Models","ref_index":14,"is_internal_anchor":true},{"citing_arxiv_id":"2512.24880","citing_title":"mHC: Manifold-Constrained Hyper-Connections","ref_index":20,"is_internal_anchor":true},{"citing_arxiv_id":"2605.13247","citing_title":"EMO: Frustratingly Easy Progressive Training of Extendable MoE","ref_index":10,"is_internal_anchor":true},{"citing_arxiv_id":"2603.28743","citing_title":"Rethinking Language Model Scaling under Transferable Hypersphere Optimization","ref_index":23,"is_internal_anchor":true},{"citing_arxiv_id":"2605.13247","citing_title":"EMO: Frustratingly Easy Progressive Training of Extendable MoE","ref_index":10,"is_internal_anchor":true},{"citing_arxiv_id":"2604.03044","citing_title":"JoyAI-LLM Flash: Advancing Mid-Scale LLMs with Token Efficiency","ref_index":13,"is_internal_anchor":true},{"citing_arxiv_id":"2605.10933","citing_title":"DECO: Sparse Mixture-of-Experts with Dense-Comparable Performance on End-Side Devices","ref_index":174,"is_internal_anchor":true},{"citing_arxiv_id":"2605.12476","citing_title":"Routers Learn the Geometry of Their Experts: Geometric Coupling in Sparse Mixture-of-Experts","ref_index":7,"is_internal_anchor":true},{"citing_arxiv_id":"2605.11447","citing_title":"Conditional Memory Enhanced Item Representation for Generative Recommendation","ref_index":40,"is_internal_anchor":true},{"citing_arxiv_id":"2605.10670","citing_title":"Surviving Partial Rank Failures in Wide Expert-Parallel MoE Inference","ref_index":31,"is_internal_anchor":true},{"citing_arxiv_id":"2605.10933","citing_title":"DECO: Sparse Mixture-of-Experts with Dense-Comparable Performance on End-Side Devices","ref_index":174,"is_internal_anchor":true}]},"formal_canon":{"evidence_count":2,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/XZ6KU673RXJCDFKJNQGACJNLRV","json":"https://pith.science/pith/XZ6KU673RXJCDFKJNQGACJNLRV.json","graph_json":"https://pith.science/api/pith-number/XZ6KU673RXJCDFKJNQGACJNLRV/graph.json","events_json":"https://pith.science/api/pith-number/XZ6KU673RXJCDFKJNQGACJNLRV/events.json","paper":"https://pith.science/paper/XZ6KU673"},"agent_actions":{"view_html":"https://pith.science/pith/XZ6KU673RXJCDFKJNQGACJNLRV","download_json":"https://pith.science/pith/XZ6KU673RXJCDFKJNQGACJNLRV.json","view_paper":"https://pith.science/paper/XZ6KU673","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2408.15664&json=true","fetch_graph":"https://pith.science/api/pith-number/XZ6KU673RXJCDFKJNQGACJNLRV/graph.json","fetch_events":"https://pith.science/api/pith-number/XZ6KU673RXJCDFKJNQGACJNLRV/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/XZ6KU673RXJCDFKJNQGACJNLRV/action/timestamp_anchor","attest_storage":"https://pith.science/pith/XZ6KU673RXJCDFKJNQGACJNLRV/action/storage_attestation","attest_author":"https://pith.science/pith/XZ6KU673RXJCDFKJNQGACJNLRV/action/author_attestation","sign_citation":"https://pith.science/pith/XZ6KU673RXJCDFKJNQGACJNLRV/action/citation_signature","submit_replication":"https://pith.science/pith/XZ6KU673RXJCDFKJNQGACJNLRV/action/replication_record"}},"created_at":"2026-05-17T23:38:52.511813+00:00","updated_at":"2026-05-17T23:38:52.511813+00:00"}