{"paper":{"title":"LLaDA2.0: Scaling Up Diffusion Language Models to 100B","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"LLaDA2.0 converts pre-trained auto-regressive LLMs into discrete diffusion models at 100B scale using a three-phase block-level training scheme.","cross_cats":["cs.AI","cs.CL"],"primary_cat":"cs.LG","authors_text":"Chengxi Li, Chongxuan Li, Da Zheng, Guoshan Lu, Huabin Liu, Jianfeng Tan, Jianguo Li, Jiaqi Hu, Ji-Rong Wen, Junbo Zhao, Junlin Zhou, Jun Zhou, Kun Chen, Lanning Wei, Lin Liu, Liwang Zhu, Lun Du, Maosong Cao, Mingliang Gong, Tiwei Bie, Xiaocheng Lu, Xiaolu Zhang, Yanmei Gu, Yihong Zhuang, Yipeng Xing, Yuxin Ma, Zehuan Li, Zenan Huang, Zhanchao Zhou, Zhenzhong Lan, Zhuochen Gong","submitted_at":"2025-12-10T09:26:18Z","abstract_excerpt":"This paper presents LLaDA2.0 -- a tuple of discrete diffusion large language models (dLLM) scaling up to 100B total parameters through systematic conversion from auto-regressive (AR) models -- establishing a new paradigm for frontier-scale deployment. Instead of costly training from scratch, LLaDA2.0 upholds knowledge inheritance, progressive adaption and efficiency-aware design principle, and seamless converts a pre-trained AR model into dLLM with a novel 3-phase block-level WSD based training scheme: progressive increasing block-size in block diffusion (warm-up), large-scale full-sequence di"},"claims":{"count":4,"items":[{"kind":"strongest_claim","text":"LLaDA2.0 establishes a new paradigm for frontier-scale deployment of discrete diffusion LLMs by systematic conversion from AR models through a novel 3-phase block-level WSD training scheme, delivering superior performance and efficiency at 100B scale.","source":"verdict.strongest_claim","status":"machine_extracted","claim_id":"C1","attestation":"unclaimed"},{"kind":"weakest_assumption","text":"That the 3-phase progressive block-size WSD training scheme successfully transfers knowledge from the original AR model while preserving parallel decoding advantages without introducing performance degradation at 100B scale.","source":"verdict.weakest_assumption","status":"machine_extracted","claim_id":"C2","attestation":"unclaimed"},{"kind":"one_line_summary","text":"LLaDA2.0 scales discrete diffusion language models to 100B parameters via systematic conversion from autoregressive models using a 3-phase WSD training scheme and releases open-source 16B and 100B MoE variants.","source":"verdict.one_line_summary","status":"machine_extracted","claim_id":"C3","attestation":"unclaimed"},{"kind":"headline","text":"LLaDA2.0 converts pre-trained auto-regressive LLMs into discrete diffusion models at 100B scale using a three-phase block-level training scheme.","source":"verdict.pith_extraction.headline","status":"machine_extracted","claim_id":"C4","attestation":"unclaimed"}],"snapshot_sha256":"f22a2c71e4b712571606320124953b20c111a00071d37b235b9bba1d22119bdc"},"source":{"id":"2512.15745","kind":"arxiv","version":2},"verdict":{"id":"3b1b5ace-efa7-454b-895d-b1777af043fc","model_set":{"reader":"grok-4.3"},"created_at":"2026-05-14T18:48:18.539049Z","strongest_claim":"LLaDA2.0 establishes a new paradigm for frontier-scale deployment of discrete diffusion LLMs by systematic conversion from AR models through a novel 3-phase block-level WSD training scheme, delivering superior performance and efficiency at 100B scale.","one_line_summary":"LLaDA2.0 scales discrete diffusion language models to 100B parameters via systematic conversion from autoregressive models using a 3-phase WSD training scheme and releases open-source 16B and 100B MoE variants.","pipeline_version":"pith-pipeline@v0.9.0","weakest_assumption":"That the 3-phase progressive block-size WSD training scheme successfully transfers knowledge from the original AR model while preserving parallel decoding advantages without introducing performance degradation at 100B scale.","pith_extraction_headline":"LLaDA2.0 converts pre-trained auto-regressive LLMs into discrete diffusion models at 100B scale using a three-phase block-level training scheme."},"references":{"count":43,"sample":[{"doi":"","year":null,"title":"Block Diffusion: Interpolating Between Autoregressive and Diffusion Language Models","work_id":"b34ab928-6ffb-4028-b13c-395a8924d76b","ref_index":1,"cited_arxiv_id":"2503.09573","is_internal_anchor":true},{"doi":"","year":null,"title":"Program Synthesis with Large Language Models","work_id":"fd241a05-03b9-4de2-9588-9d77ce176125","ref_index":2,"cited_arxiv_id":"2108.07732","is_internal_anchor":true},{"doi":"","year":null,"title":"Evaluating Large Language Models Trained on Code","work_id":"042493e9-b26f-4b4e-bbde-382072ca9b08","ref_index":3,"cited_arxiv_id":"2107.03374","is_internal_anchor":true},{"doi":"","year":null,"title":"Dpad: Efficient diffusion language models with suffix dropout","work_id":"3f0e3292-b812-4d35-9383-8e3959725c6b","ref_index":4,"cited_arxiv_id":"","is_internal_anchor":false},{"doi":"","year":null,"title":"Think you have Solved Question Answering? Try ARC, the AI2 Reasoning Challenge","work_id":"28ea1282-d657-4c61-a83c-f1249be6d6b1","ref_index":5,"cited_arxiv_id":"1803.05457","is_internal_anchor":true}],"resolved_work":43,"snapshot_sha256":"969d81ae391b454d45b28ddf9305ec05d2bd691f8a1432a965c4e62c02bb871d","internal_anchors":25},"formal_canon":{"evidence_count":2,"snapshot_sha256":"696fb584b1639e7410e0f7a5ae35ed881deae00abac2a1da7bc8fac10abb5030"},"author_claims":{"count":0,"strong_count":0,"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"builder_version":"pith-number-builder-2026-05-17-v1"}