{"work":{"id":"a818f37a-b985-49e6-879d-603a00525f65","openalex_id":"https://openalex.org/W4414846785","doi":"10.48550/arxiv.2502.16982","arxiv_id":"2502.16982","raw_key":null,"title":"Muon is Scalable for LLM Training","authors":null,"authors_text":"Jingyuan Liu, Jianlin Su, Xingcheng Yao, Zhejun Jiang, Guokun Lai, Yulun Du","year":2025,"venue":"cs.LG","abstract":"Recently, the Muon optimizer based on matrix orthogonalization has demonstrated strong results in training small-scale language models, but the scalability to larger models has not been proven. We identify two crucial techniques for scaling up Muon: (1) adding weight decay and (2) carefully adjusting the per-parameter update scale. These techniques allow Muon to work out-of-the-box on large-scale training without the need of hyper-parameter tuning. Scaling law experiments indicate that Muon achieves $\\sim\\!2\\times$ computational efficiency compared to AdamW with compute optimal training.\n  Based on these improvements, we introduce Moonlight, a 3B/16B-parameter Mixture-of-Expert (MoE) model trained with 5.7T tokens using Muon. Our model improves the current Pareto frontier, achieving better performance with much fewer training FLOPs compared to prior models.\n  We open-source our distributed Muon implementation that is memory optimal and communication efficient. We also release the pretrained, instruction-tuned, and intermediate checkpoints to support future research.","external_url":"https://arxiv.org/abs/2502.16982","cited_by_count":1,"metadata_source":"pith","metadata_fetched_at":"2026-08-05T02:28:24.338817+00:00","pith_arxiv_id":"2502.16982","created_at":"2026-05-10T12:10:22.008037+00:00","updated_at":"2026-08-05T02:28:24.338817+00:00","title_quality_ok":false,"display_title":"Muon is Scalable for LLM Training","render_title":"Muon is Scalable for LLM Training"},"hub":{"state":{"work_id":"a818f37a-b985-49e6-879d-603a00525f65","tier":"super_hub","tier_reason":"100+ Pith inbound or 10,000+ external citations","pith_inbound_count":132,"external_cited_by_count":1,"distinct_field_count":16,"first_pith_cited_at":"2023-03-14T17:47:09+00:00","last_pith_cited_at":"2026-07-09T16:15:43+00:00","author_build_status":"needed","summary_status":"needed","contexts_status":"needed","graph_status":"needed","ask_index_status":"needed","reader_status":"not_needed","recognition_status":"not_needed","updated_at":"2026-08-23T07:09:34.456622+00:00","tier_text":"super_hub"},"tier":"super_hub","role_counts":[{"context_role":"background","n":19},{"context_role":"baseline","n":2},{"context_role":"method","n":2},{"context_role":"dataset","n":1},{"context_role":"other","n":1}],"polarity_counts":[{"context_polarity":"background","n":16},{"context_polarity":"unclear","n":3},{"context_polarity":"baseline","n":2},{"context_polarity":"use_method","n":2},{"context_polarity":"support","n":1},{"context_polarity":"use_dataset","n":1}],"runs":{"ask_index":{"job_type":"ask_index","status":"succeeded","result":{"title":"Muon is Scalable for LLM Training","claims":[{"claim_text":"Recently, the Muon optimizer based on matrix orthogonalization has demonstrated strong results in training small-scale language models, but the scalability to larger models has not been proven. We identify two crucial techniques for scaling up Muon: (1) adding weight decay and (2) carefully adjusting the per-parameter update scale. These techniques allow Muon to work out-of-the-box on large-scale training without the need of hyper-parameter tuning. Scaling law experiments indicate that Muon achieves $\\sim\\!2\\times$ computational efficiency compared to AdamW with compute optimal training.\n  Bas","claim_type":"abstract","evidence_strength":"source_metadata"},{"claim_text":"taining a consistent level angle that offers an immersive street-level perspective.\" After finetuning, we observe the model no longer responds to new text captions; we leave this for future work. Optimization.We adopt the v-prediction [47] objective with a shifted noise scheduletowardhighertimesteps(afactorof3.0,followingSD3[17]andFlux[34]). We use the Muon [38] optimizer with a fixed learning rate of 1e-5 with warmup. We train our model on 32 A100 GPUs for a week, around 20k iterations. Empir- ","claim_type":"method","confidence":0.9,"evidence_strength":"citation_context"},{"claim_text":"Adam: A method for stochastic optimization.arXiv preprint arXiv:1412.6980, 2014. James Kirkpatrick, Razvan Pascanu, Neil Rabinowitz, Joel Veness, Guillaume Desjardins, Andrei A Rusu, Kieran Milan, John Quan, Tiago Ramalho, Agnieszka Grabska-Barwinska, et al. Overcoming catastrophic forgetting in neural networks.Proceedings of the national academy of sciences, 114(13): 3521-3526, 2017. Frederik Kunstner, Jacques Chen, Jonathan Wilder Lavington, and Mark Schmidt. Noise is not the main factor behin","claim_type":"background","confidence":0.9,"evidence_strength":"citation_context"},{"claim_text":"baseline using commercially available APIs and to better understand the gap between open-weight and closed models. •Mixture of Expert (MoE) modelsofcomparablesize:deepseek-moe-16b-chat[29]fromDeepseek with 16 billion parameters of which 2.7 billion are active,gpt-oss-20b[30] from OpenAI with 20 bil- lion parameters and 3.6 billion active, andMoonlight-16B-A3B-Instruct[31], with the most similar amount of total and active parameters with respect to EngGPT2-MoE-16B. We note that the chosen models ","claim_type":"baseline","confidence":0.9,"evidence_strength":"citation_context"},{"claim_text":"6 Compatibility with Maximal Update Parametrization µP [84-86] suggests that the following two spectral scaling conditions are crucial for training stability: Forward spectral condition:∥W∥ 2 =Θ \u0010p dout/din \u0011 | {z } This is inherently satisfied by Pion. Update spectral condition:∥∆W∥ 2 =Θ \u0010p dout/din \u0011 | {z } This is easily satisfied by Muon. .(9) Existing µP-compatible optimizers [45, 74, 81] are built on Muon, which inherently satisfies the update condition. As a result, prior work focuses on ","claim_type":"baseline","confidence":0.9,"evidence_strength":"citation_context"},{"claim_text":"5 on a comprehensive benchmark suite spanning text-based reasoning, compet- itive and agentic coding, multimodal understanding (image and video), autonomous agentic execution, and computer use. Our benchmark taxonomy is organized along the following capability axes: •Reasoning & General: Humanity's Last Exam (HLE) [46], AIME 2025 [4], HMMT 2025 (Feb) [58], IMO- AnswerBench [37], GPQA-Diamond [47], MMLU-Pro [64], SimpleQA Verified [22], AdvancedIF [23], and LongBench v2 [9]. 10 Kimi K2.5TECHNICAL","claim_type":"dataset","confidence":0.9,"evidence_strength":"citation_context"},{"claim_text":"[53] Hanxiao Liu, Karen Simonyan, and Yiming Yang. Darts: Differentiable architecture search. In7th International Conference on Learning Representations, 2019. 15 [54] Jingyuan Liu, Jianlin Su, Xingcheng Yao, Zhejun Jiang, Guokun Lai, Yulun Du, Yidao Qin, Weixin Xu, Enzhe Lu, Junjie Yan, et al. Muon is scalable for llm training.arXiv preprint arXiv:2502.16982, 2025. [55] Yujie Liu, Zonglin Yang, Tong Xie, Jinjie Ni, Ben Gao, Yuqiang Li, Shixiang Tang, Wanli Ouyang, Erik Cambria, and Dongzhan Zho","claim_type":"background","confidence":0.85,"evidence_strength":"citation_context"}],"why_cited":"Pith tracks Muon is Scalable for LLM Training because it crossed a citation-hub threshold. Current citing contexts most often use it as background evidence (18 contexts).","role_counts":[{"n":18,"context_role":"background"},{"n":2,"context_role":"baseline"},{"n":2,"context_role":"method"},{"n":1,"context_role":"dataset"},{"n":1,"context_role":"other"}]},"error":null,"updated_at":"2026-07-02T02:22:10.351853+00:00"},"author_expand":{"job_type":"author_expand","status":"succeeded","result":{"authors_linked":[{"id":"d145fbf0-5824-4eeb-b356-7342cfd46ab0","orcid":null,"display_name":"Jingyuan Liu"},{"id":"9fe5a313-73b2-4ef4-9c62-89c0e36bd922","orcid":null,"display_name":"Jianlin Su"},{"id":"125a6550-fa16-4010-a374-8c7912c2e026","orcid":null,"display_name":"Xingcheng Yao"},{"id":"aa905671-829f-4b14-83e2-6b952fc7f68b","orcid":null,"display_name":"Zhejun Jiang"},{"id":"85c26610-00d6-475a-8bc0-f703ee12d2a2","orcid":null,"display_name":"Guokun Lai"},{"id":"22f70293-47a2-4dd1-9f39-457eec7471eb","orcid":null,"display_name":"Yulun Du"}]},"error":null,"updated_at":"2026-07-02T02:22:10.868122+00:00"},"context_extract":{"job_type":"context_extract","status":"succeeded","result":{"enqueued_papers":25},"error":null,"updated_at":"2026-05-14T17:48:48.954661+00:00"},"graph_features":{"job_type":"graph_features","status":"succeeded","result":{"co_cited":[{"title":"Adam: A Method for Stochastic Optimization","work_id":"1910796d-9b52-4683-bf5c-de9632c1028b","shared_citers":14},{"title":"DeepSeek-V3 Technical Report","work_id":"57d2791d-2219-4c31-a077-afc04b12a75c","shared_citers":11},{"title":"Kimi K2: Open Agentic Intelligence","work_id":"7f18284c-12d3-4137-bea1-1da97e8cf3c1","shared_citers":10},{"title":"Think you have Solved Question Answering? Try ARC, the AI2 Reasoning Challenge","work_id":"28ea1282-d657-4c61-a83c-f1249be6d6b1","shared_citers":10},{"title":"Decoupled Weight Decay Regularization","work_id":"07ef7360-d385-4033-83f7-8384a6325204","shared_citers":9},{"title":"LLaMA: Open and Efficient Foundation Language Models","work_id":"c018fc23-6f3f-4035-9d02-28a2173b2b9d","shared_citers":9},{"title":"The Llama 3 Herd of Models","work_id":"1549a635-88af-4ac1-acfe-51ae7bb53345","shared_citers":9},{"title":"Training Verifiers to Solve Math Word Problems","work_id":"acab1aa8-b4d6-40e0-a3ee-25341701dca2","shared_citers":9},{"title":"Qwen3 Technical Report","work_id":"25a4e30c-1232-48e7-9925-02fa12ba7c9e","shared_citers":8},{"title":"Llama 2: Open Foundation and Fine-Tuned Chat Models","work_id":"68a5177f-d644-44c1-bd4f-4e5278c22f5d","shared_citers":6},{"title":"Old optimizer, new norm: An anthology","work_id":"c0089db6-7349-44fd-b103-0d7b36142bab","shared_citers":6},{"title":"On the Convergence Analysis of Muon","work_id":"fe8817f3-ca08-4ed5-a8fd-c2bfdaf1a459","shared_citers":6},{"title":"Scaling Laws for Neural Language Models","work_id":"b7dd8749-9c45-4977-ab9b-64478dce1ae8","shared_citers":6},{"title":"Evaluating Large Language Models Trained on Code","work_id":"042493e9-b26f-4b4e-bbde-382072ca9b08","shared_citers":5},{"title":"GLU Variants Improve Transformer","work_id":"17d0763c-1016-41ab-a478-478e890765eb","shared_citers":5},{"title":"Kimi k1.5: Scaling Reinforcement Learning with LLMs","work_id":"bff96ab1-bd6a-4585-be23-74fdb51969c7","shared_citers":5},{"title":"Mamba: Linear-Time Sequence Modeling with Selective State Spaces","work_id":"4ee75248-1199-492c-a52f-6661e0f4adff","shared_citers":5},{"title":"Pethick, W","work_id":"4f1b774a-8ada-4d79-a90b-520511fce5d0","shared_citers":5},{"title":"Training Compute-Optimal Large Language Models","work_id":"b2faf28d-86b7-429c-bc42-469458efc246","shared_citers":5},{"title":"Understanding gradient orthogonalization for deep learning via non-euclidean trust-region optimization","work_id":"fe62cb1b-f9dc-4379-8586-5c2f591ba7e5","shared_citers":5},{"title":"An Image is Worth 16x16 Words: Transformers for Image Recognition at Scale","work_id":"e96730e3-129b-4db6-b981-15ab7932e297","shared_citers":4},{"title":"arXiv preprint arXiv:2511.00674 , year=","work_id":"5e0fa664-f7a6-4ce3-934b-dd6d242ac7c6","shared_citers":4},{"title":"DeepSeek-R1: Incentivizing Reasoning Capability in LLMs via Reinforcement Learning","work_id":"e6b75ad5-2877-4168-97c8-710407094d20","shared_citers":4},{"title":"Dion: Distributed orthonormal- ized updates.arXiv preprint: 2504.05295","work_id":"4ecfbe15-0efd-41f4-a129-82f6a2164d2f","shared_citers":4}],"time_series":[{"n":1,"year":2023},{"n":4,"year":2025},{"n":36,"year":2026}],"dependency_candidates":[]},"error":null,"updated_at":"2026-05-14T17:48:48.990535+00:00"},"identity_refresh":{"job_type":"identity_refresh","status":"succeeded","result":{"items":[{"title":"Qwen3 Technical Report","outcome":"unchanged","work_id":"25a4e30c-1232-48e7-9925-02fa12ba7c9e","resolver":"local_arxiv","confidence":0.98,"old_work_id":"25a4e30c-1232-48e7-9925-02fa12ba7c9e"}],"counts":{"fixed":0,"merged":0,"unchanged":1,"quarantined":0,"needs_external_resolution":0},"errors":[],"attempted":1},"error":null,"updated_at":"2026-05-14T17:48:55.280492+00:00"},"role_polarity":{"job_type":"role_polarity","status":"succeeded","result":{"title":"Muon is Scalable for LLM Training","claims":[{"claim_text":"Recently, the Muon optimizer based on matrix orthogonalization has demonstrated strong results in training small-scale language models, but the scalability to larger models has not been proven. We identify two crucial techniques for scaling up Muon: (1) adding weight decay and (2) carefully adjusting the per-parameter update scale. These techniques allow Muon to work out-of-the-box on large-scale training without the need of hyper-parameter tuning. Scaling law experiments indicate that Muon achieves $\\sim\\!2\\times$ computational efficiency compared to AdamW with compute optimal training.\n  Bas","claim_type":"abstract","evidence_strength":"source_metadata"},{"claim_text":"taining a consistent level angle that offers an immersive street-level perspective.\" After finetuning, we observe the model no longer responds to new text captions; we leave this for future work. Optimization.We adopt the v-prediction [47] objective with a shifted noise scheduletowardhighertimesteps(afactorof3.0,followingSD3[17]andFlux[34]). We use the Muon [38] optimizer with a fixed learning rate of 1e-5 with warmup. We train our model on 32 A100 GPUs for a week, around 20k iterations. Empir- ","claim_type":"method","confidence":0.9,"evidence_strength":"citation_context"},{"claim_text":"Adam: A method for stochastic optimization.arXiv preprint arXiv:1412.6980, 2014. James Kirkpatrick, Razvan Pascanu, Neil Rabinowitz, Joel Veness, Guillaume Desjardins, Andrei A Rusu, Kieran Milan, John Quan, Tiago Ramalho, Agnieszka Grabska-Barwinska, et al. Overcoming catastrophic forgetting in neural networks.Proceedings of the national academy of sciences, 114(13): 3521-3526, 2017. Frederik Kunstner, Jacques Chen, Jonathan Wilder Lavington, and Mark Schmidt. Noise is not the main factor behin","claim_type":"background","confidence":0.9,"evidence_strength":"citation_context"},{"claim_text":"baseline using commercially available APIs and to better understand the gap between open-weight and closed models. •Mixture of Expert (MoE) modelsofcomparablesize:deepseek-moe-16b-chat[29]fromDeepseek with 16 billion parameters of which 2.7 billion are active,gpt-oss-20b[30] from OpenAI with 20 bil- lion parameters and 3.6 billion active, andMoonlight-16B-A3B-Instruct[31], with the most similar amount of total and active parameters with respect to EngGPT2-MoE-16B. We note that the chosen models ","claim_type":"baseline","confidence":0.9,"evidence_strength":"citation_context"},{"claim_text":"6 Compatibility with Maximal Update Parametrization µP [84-86] suggests that the following two spectral scaling conditions are crucial for training stability: Forward spectral condition:∥W∥ 2 =Θ \u0010p dout/din \u0011 | {z } This is inherently satisfied by Pion. Update spectral condition:∥∆W∥ 2 =Θ \u0010p dout/din \u0011 | {z } This is easily satisfied by Muon. .(9) Existing µP-compatible optimizers [45, 74, 81] are built on Muon, which inherently satisfies the update condition. As a result, prior work focuses on ","claim_type":"baseline","confidence":0.9,"evidence_strength":"citation_context"},{"claim_text":"5 on a comprehensive benchmark suite spanning text-based reasoning, compet- itive and agentic coding, multimodal understanding (image and video), autonomous agentic execution, and computer use. Our benchmark taxonomy is organized along the following capability axes: •Reasoning & General: Humanity's Last Exam (HLE) [46], AIME 2025 [4], HMMT 2025 (Feb) [58], IMO- AnswerBench [37], GPQA-Diamond [47], MMLU-Pro [64], SimpleQA Verified [22], AdvancedIF [23], and LongBench v2 [9]. 10 Kimi K2.5TECHNICAL","claim_type":"dataset","confidence":0.9,"evidence_strength":"citation_context"},{"claim_text":"[53] Hanxiao Liu, Karen Simonyan, and Yiming Yang. Darts: Differentiable architecture search. In7th International Conference on Learning Representations, 2019. 15 [54] Jingyuan Liu, Jianlin Su, Xingcheng Yao, Zhejun Jiang, Guokun Lai, Yulun Du, Yidao Qin, Weixin Xu, Enzhe Lu, Junjie Yan, et al. Muon is scalable for llm training.arXiv preprint arXiv:2502.16982, 2025. [55] Yujie Liu, Zonglin Yang, Tong Xie, Jinjie Ni, Ben Gao, Yuqiang Li, Shixiang Tang, Wanli Ouyang, Erik Cambria, and Dongzhan Zho","claim_type":"background","confidence":0.85,"evidence_strength":"citation_context"}],"why_cited":"Pith tracks Muon is Scalable for LLM Training because it crossed a citation-hub threshold. Current citing contexts most often use it as background evidence (18 contexts).","role_counts":[{"n":18,"context_role":"background"},{"n":2,"context_role":"baseline"},{"n":2,"context_role":"method"},{"n":1,"context_role":"dataset"},{"n":1,"context_role":"other"}]},"error":null,"updated_at":"2026-07-02T02:22:10.349125+00:00"},"summary_claims":{"job_type":"summary_claims","status":"succeeded","result":{"title":"Muon is Scalable for LLM Training","claims":[{"claim_text":"Recently, the Muon optimizer based on matrix orthogonalization has demonstrated strong results in training small-scale language models, but the scalability to larger models has not been proven. We identify two crucial techniques for scaling up Muon: (1) adding weight decay and (2) carefully adjusting the per-parameter update scale. These techniques allow Muon to work out-of-the-box on large-scale training without the need of hyper-parameter tuning. Scaling law experiments indicate that Muon achieves $\\sim\\!2\\times$ computational efficiency compared to AdamW with compute optimal training.\n  Bas","claim_type":"abstract","evidence_strength":"source_metadata"}],"why_cited":"Pith tracks Muon is Scalable for LLM Training because it crossed a citation-hub threshold.","role_counts":[]},"error":null,"updated_at":"2026-05-14T17:48:42.919054+00:00"}},"summary":{"title":"Muon is Scalable for LLM Training","claims":[{"claim_text":"Recently, the Muon optimizer based on matrix orthogonalization has demonstrated strong results in training small-scale language models, but the scalability to larger models has not been proven. We identify two crucial techniques for scaling up Muon: (1) adding weight decay and (2) carefully adjusting the per-parameter update scale. These techniques allow Muon to work out-of-the-box on large-scale training without the need of hyper-parameter tuning. Scaling law experiments indicate that Muon achieves $\\sim\\!2\\times$ computational efficiency compared to AdamW with compute optimal training.\n  Bas","claim_type":"abstract","evidence_strength":"source_metadata"}],"why_cited":"Pith tracks Muon is Scalable for LLM Training because it crossed a citation-hub threshold.","role_counts":[]},"graph":{"co_cited":[{"title":"Adam: A Method for Stochastic Optimization","work_id":"1910796d-9b52-4683-bf5c-de9632c1028b","shared_citers":14},{"title":"DeepSeek-V3 Technical Report","work_id":"57d2791d-2219-4c31-a077-afc04b12a75c","shared_citers":11},{"title":"Kimi K2: Open Agentic Intelligence","work_id":"7f18284c-12d3-4137-bea1-1da97e8cf3c1","shared_citers":10},{"title":"Think you have Solved Question Answering? Try ARC, the AI2 Reasoning Challenge","work_id":"28ea1282-d657-4c61-a83c-f1249be6d6b1","shared_citers":10},{"title":"Decoupled Weight Decay Regularization","work_id":"07ef7360-d385-4033-83f7-8384a6325204","shared_citers":9},{"title":"LLaMA: Open and Efficient Foundation Language Models","work_id":"c018fc23-6f3f-4035-9d02-28a2173b2b9d","shared_citers":9},{"title":"The Llama 3 Herd of Models","work_id":"1549a635-88af-4ac1-acfe-51ae7bb53345","shared_citers":9},{"title":"Training Verifiers to Solve Math Word Problems","work_id":"acab1aa8-b4d6-40e0-a3ee-25341701dca2","shared_citers":9},{"title":"Qwen3 Technical Report","work_id":"25a4e30c-1232-48e7-9925-02fa12ba7c9e","shared_citers":8},{"title":"Llama 2: Open Foundation and Fine-Tuned Chat Models","work_id":"68a5177f-d644-44c1-bd4f-4e5278c22f5d","shared_citers":6},{"title":"Old optimizer, new norm: An anthology","work_id":"c0089db6-7349-44fd-b103-0d7b36142bab","shared_citers":6},{"title":"On the Convergence Analysis of Muon","work_id":"fe8817f3-ca08-4ed5-a8fd-c2bfdaf1a459","shared_citers":6},{"title":"Scaling Laws for Neural Language Models","work_id":"b7dd8749-9c45-4977-ab9b-64478dce1ae8","shared_citers":6},{"title":"Evaluating Large Language Models Trained on Code","work_id":"042493e9-b26f-4b4e-bbde-382072ca9b08","shared_citers":5},{"title":"GLU Variants Improve Transformer","work_id":"17d0763c-1016-41ab-a478-478e890765eb","shared_citers":5},{"title":"Kimi k1.5: Scaling Reinforcement Learning with LLMs","work_id":"bff96ab1-bd6a-4585-be23-74fdb51969c7","shared_citers":5},{"title":"Mamba: Linear-Time Sequence Modeling with Selective State Spaces","work_id":"4ee75248-1199-492c-a52f-6661e0f4adff","shared_citers":5},{"title":"Pethick, W","work_id":"4f1b774a-8ada-4d79-a90b-520511fce5d0","shared_citers":5},{"title":"Training Compute-Optimal Large Language Models","work_id":"b2faf28d-86b7-429c-bc42-469458efc246","shared_citers":5},{"title":"Understanding gradient orthogonalization for deep learning via non-euclidean trust-region optimization","work_id":"fe62cb1b-f9dc-4379-8586-5c2f591ba7e5","shared_citers":5},{"title":"An Image is Worth 16x16 Words: Transformers for Image Recognition at Scale","work_id":"e96730e3-129b-4db6-b981-15ab7932e297","shared_citers":4},{"title":"arXiv preprint arXiv:2511.00674 , year=","work_id":"5e0fa664-f7a6-4ce3-934b-dd6d242ac7c6","shared_citers":4},{"title":"DeepSeek-R1: Incentivizing Reasoning Capability in LLMs via Reinforcement Learning","work_id":"e6b75ad5-2877-4168-97c8-710407094d20","shared_citers":4},{"title":"Dion: Distributed orthonormal- ized updates.arXiv preprint: 2504.05295","work_id":"4ecfbe15-0efd-41f4-a129-82f6a2164d2f","shared_citers":4}],"time_series":[{"n":1,"year":2023},{"n":4,"year":2025},{"n":36,"year":2026}],"dependency_candidates":[]},"authors":[{"id":"85c26610-00d6-475a-8bc0-f703ee12d2a2","orcid":null,"display_name":"Guokun Lai","source":"manual","import_confidence":0.72},{"id":"9fe5a313-73b2-4ef4-9c62-89c0e36bd922","orcid":null,"display_name":"Jianlin Su","source":"manual","import_confidence":0.72},{"id":"d145fbf0-5824-4eeb-b356-7342cfd46ab0","orcid":null,"display_name":"Jingyuan Liu","source":"manual","import_confidence":0.72},{"id":"125a6550-fa16-4010-a374-8c7912c2e026","orcid":null,"display_name":"Xingcheng Yao","source":"manual","import_confidence":0.72},{"id":"22f70293-47a2-4dd1-9f39-457eec7471eb","orcid":null,"display_name":"Yulun Du","source":"manual","import_confidence":0.72},{"id":"aa905671-829f-4b14-83e2-6b952fc7f68b","orcid":null,"display_name":"Zhejun Jiang","source":"manual","import_confidence":0.72}]}}