{"about":{"site":"https://codewithpapers.app","non_affiliation":"Code with Papers and Syntology are not affiliated with, endorsed by, or sponsored by Papers with Code, Meta, or the pwc-archive mirror.","licence":"CC BY-SA 4.0","licence_url":"https://creativecommons.org/licenses/by-sa/4.0/legalcode","attribution":"https://codewithpapers.app/attribution","modified":"archive material modified by Syntology; see the attribution page"},"url":"/paper/dense-backpropagation-improves-training-for","title":"Dense Backpropagation Improves Training for Sparse Mixture-of-Experts","arxiv_id":"2504.12463","date":"2025-04-16","proceeding":null,"authors":["Ashwinee Panda","Vatsal Baherwani","Zain Sarwar","Benjamin Therien","Supriyo Chakraborty","Tom Goldstein"],"abstract":"Mixture of Experts (MoE) pretraining is more scalable than dense Transformer pretraining, because MoEs learn to route inputs to a sparse set of their feedforward parameters. However, this means that MoEs only receive a sparse backward update, leading to training instability and suboptimal performance. We present a lightweight approximation method that gives the MoE router a dense gradient update while continuing to sparsely activate its parameters. Our method, which we refer to as Default MoE, substitutes missing expert activations with default outputs consisting of an exponential moving average of expert outputs previously seen over the course of training. This allows the router to receive signals from every expert for each token, leading to significant improvements in training performance. Our Default MoE outperforms standard TopK routing in a variety of settings without requiring significant computational overhead. Code: https://github.com/vatsal0/default-moe.","url_abs":"https://arxiv.org/abs/2504.12463v2","url_pdf":"https://arxiv.org/pdf/2504.12463v2.pdf","source":{"archive":"pwc-archive (Hugging Face), CC BY-SA 4.0","snapshot":"2025-07-28","licence_url":"https://creativecommons.org/licenses/by-sa/4.0/legalcode","row_kind":"abstracts"},"code_links":[{"paper_slug":"dense-backpropagation-improves-training-for","repo_url":"https://github.com/vatsal0/default-moe","is_official":1,"mentioned_in_paper":1,"mentioned_in_github":0,"framework":"pytorch","reach":{"status":"ok","spdx":"Apache-2.0"}}],"tasks":[{"task_slug":"mixture-of-experts","task_name":"Mixture-of-Experts"}],"methods":[{"method_slug":"absolute-position-encodings","method_name":"Absolute Position Encodings"},{"method_slug":"adam","method_name":"Adam"},{"method_slug":"attention","method_name":"Attention"},{"method_slug":"bpe","method_name":"BPE"},{"method_slug":"dense-connections","method_name":"Dense Connections"},{"method_slug":"dropout","method_name":"Dropout"},{"method_slug":"label-smoothing","method_name":"Label Smoothing"},{"method_slug":"layer-normalization","method_name":"Layer Normalization"},{"method_slug":"linear-layer","method_name":"Linear Layer"},{"method_slug":"moe","method_name":"MoE"},{"method_slug":"multi-head-attention","method_name":"Multi-Head Attention"},{"method_slug":"position-wise-feed-forward-layer","method_name":"Position-Wise Feed-Forward Layer"},{"method_slug":"residual-connection","method_name":"Residual Connection"},{"method_slug":"set","method_name":"SET"},{"method_slug":"softmax","method_name":"Softmax"},{"method_slug":"transformer","method_name":"Transformer"}],"datasets_introduced":[],"methods_introduced":[],"results":[],"syntology":{"atlas_url":"https://app.syntology.ai/?focus=2504.12463","mcp":{"get_harvested_code_for_paper":{"arxiv_id":"2504.12463"}},"developers":"https://syntology.ai/developers","read_at":"2026-09-24T18:15:14+00:00","read_at_is":"when the build read Syntology's graph, not when any sample ran","claim":"Per-sample execution status on synthesized fixtures; not a correctness claim about the paper. Samples come from repositories linked to the paper, official or community; repo_kind says which.","repos":[{"provenance":"external:paperswithcode_snapshot_2025-07-28","url":"https://github.com/vatsal0/default-moe","reach":{"status":"ok","spdx":"Apache-2.0"}}],"summary":{"unverified":6},"by_repo_kind":{"official":{"samples":6,"ran":0,"repositories":1}},"repo_kind_vocabulary":{"official":"The archive marks this repository official for the paper","named_in_paper":"The archive records that the paper mentions this repository; it is not marked official","listed":"In the archive's code links for this paper, not marked official and not recorded as mentioned in the paper","found_in_text":"Syntology found this repository in the paper's own text; whether it is the authors' implementation is not asserted","community":"Not in the archive's code links for this paper; a community repository Syntology harvested"},"n_pointer_only_for_licence":6,"samples":[{"code_sha256_prefix":"8de5b37b264d4283","entry":"cast_if_autocast_enabled","repo":"vatsal0/default-moe","repo_kind":"official","path":"megatron/model/moe.py","file_url":"https://github.com/vatsal0/default-moe/blob/HEAD/megatron/model/moe.py","link_basis":"harvester_set","language":"python","status":"unverified","verification_level":0,"contract_check":null,"metamorphic_tier":null,"behaviour_fingerprint":false,"licence":"Apache-2.0","inline_ok":false,"mcp_get_code":{"code_sha256":"8de5b37b264d4283"}},{"code_sha256_prefix":"69002400aad5a8a6","entry":"get_checkpoint_tag","repo":"vatsal0/default-moe","repo_kind":"official","path":"megatron/checkpointing.py","file_url":"https://github.com/vatsal0/default-moe/blob/HEAD/megatron/checkpointing.py","link_basis":"harvester_set","language":"python","status":"unverified","verification_level":0,"contract_check":null,"metamorphic_tier":null,"behaviour_fingerprint":false,"licence":"Apache-2.0","inline_ok":false,"mcp_get_code":{"code_sha256":"69002400aad5a8a6"}},{"code_sha256_prefix":"d5081f28e2befd2c","entry":"get_docs","repo":"vatsal0/default-moe","repo_kind":"official","path":"configs/gen_docs.py","file_url":"https://github.com/vatsal0/default-moe/blob/HEAD/configs/gen_docs.py","link_basis":"harvester_set","language":"python","status":"unverified","verification_level":0,"contract_check":null,"metamorphic_tier":null,"behaviour_fingerprint":false,"licence":"Apache-2.0","inline_ok":false,"mcp_get_code":{"code_sha256":"d5081f28e2befd2c"}},{"code_sha256_prefix":"3a5fc202b5d9ec24","entry":"monitor_method_cuda_wall_times","repo":"vatsal0/default-moe","repo_kind":"official","path":"megatron/devutil.py","file_url":"https://github.com/vatsal0/default-moe/blob/HEAD/megatron/devutil.py","link_basis":"harvester_set","language":"python","status":"unverified","verification_level":0,"contract_check":null,"metamorphic_tier":null,"behaviour_fingerprint":false,"licence":"Apache-2.0","inline_ok":false,"mcp_get_code":{"code_sha256":"3a5fc202b5d9ec24"}},{"code_sha256_prefix":"50c9c165cf801f59","entry":"pairwise","repo":"vatsal0/default-moe","repo_kind":"official","path":"configs/gen_docs.py","file_url":"https://github.com/vatsal0/default-moe/blob/HEAD/configs/gen_docs.py","link_basis":"harvester_set","language":"python","status":"unverified","verification_level":0,"contract_check":null,"metamorphic_tier":null,"behaviour_fingerprint":false,"licence":"Apache-2.0","inline_ok":false,"mcp_get_code":{"code_sha256":"50c9c165cf801f59"}},{"code_sha256_prefix":"64dd872a8ac47294","entry":"to_md","repo":"vatsal0/default-moe","repo_kind":"official","path":"configs/gen_docs.py","file_url":"https://github.com/vatsal0/default-moe/blob/HEAD/configs/gen_docs.py","link_basis":"harvester_set","language":"python","status":"unverified","verification_level":0,"contract_check":null,"metamorphic_tier":null,"behaviour_fingerprint":false,"licence":"Apache-2.0","inline_ok":false,"mcp_get_code":{"code_sha256":"64dd872a8ac47294"}}]},"arxiv_metadata":null,"syntology_extracted_results":null}