https://huggingface.co/inclusionAI/LLaDA2.1-mini
LLaDA2.1-flash:https://huggingface.co/inclusionAI/LLaDA2.1-flash

\n","updatedAt":"2026-02-10T04:44:10.040Z","author":{"_id":"673b5f24e863f1d28b402efc","avatarUrl":"https://cdn-avatars.huggingface.co/v1/production/uploads/no-auth/19gUgtPEY3-FtY0sNlI_-.png","fullname":"yihongzhuang","name":"utdawn","type":"user","isPro":false,"isHf":false,"isHfAdmin":false,"isMod":false,"followerCount":5,"isUserFollowing":false}},"numEdits":0,"identifiedLanguage":{"language":"en","probability":0.5468952655792236},"editors":["utdawn"],"editorAvatarUrls":["https://cdn-avatars.huggingface.co/v1/production/uploads/no-auth/19gUgtPEY3-FtY0sNlI_-.png"],"reactions":[],"isReport":false},"replies":[{"id":"698af15cdb35b44e2a78f74c","author":{"_id":"63177d85f957903db971a173","avatarUrl":"https://cdn-avatars.huggingface.co/v1/production/uploads/1665094764329-63177d85f957903db971a173.png","fullname":"Artem","name":"kabachuha","type":"user","isPro":false,"isHf":false,"isHfAdmin":false,"isMod":false,"followerCount":48,"isUserFollowing":false},"createdAt":"2026-02-10T08:50:36.000Z","type":"comment","data":{"edited":false,"hidden":false,"latest":{"raw":"Hello!\n\nIs it possible for you to write an example fine-tuning code for this model? (just a simple training loop) It would be very useful if this model could be fine-tuned on downstream tasks, such as programming languages or specific domains. Frameworks such as unsloth could also benefit from it","html":"

Hello!

\n

Is it possible for you to write an example fine-tuning code for this model? (just a simple training loop) It would be very useful if this model could be fine-tuned on downstream tasks, such as programming languages or specific domains. Frameworks such as unsloth could also benefit from it

\n","updatedAt":"2026-02-10T08:50:36.961Z","author":{"_id":"63177d85f957903db971a173","avatarUrl":"https://cdn-avatars.huggingface.co/v1/production/uploads/1665094764329-63177d85f957903db971a173.png","fullname":"Artem","name":"kabachuha","type":"user","isPro":false,"isHf":false,"isHfAdmin":false,"isMod":false,"followerCount":48,"isUserFollowing":false}},"numEdits":0,"identifiedLanguage":{"language":"en","probability":0.9469355344772339},"editors":["kabachuha"],"editorAvatarUrls":["https://cdn-avatars.huggingface.co/v1/production/uploads/1665094764329-63177d85f957903db971a173.png"],"reactions":[],"isReport":false,"parentCommentId":"698ab79a5025098aec820b4b"}}]},{"id":"698bded3c332ebffe5174457","author":{"_id":"63d3e0e8ff1384ce6c5dd17d","avatarUrl":"https://cdn-avatars.huggingface.co/v1/production/uploads/1674830754237-63d3e0e8ff1384ce6c5dd17d.jpeg","fullname":"Librarian Bot (Bot)","name":"librarian-bot","type":"user","isPro":false,"isHf":false,"isHfAdmin":false,"isMod":false,"followerCount":318,"isUserFollowing":false},"createdAt":"2026-02-11T01:43:47.000Z","type":"comment","data":{"edited":false,"hidden":false,"latest":{"raw":"This is an automated message from the [Librarian Bot](https://huggingface.co/librarian-bots). I found the following papers similar to this paper. \n\nThe following papers were recommended by the Semantic Scholar API \n\n* [LoPA: Scaling dLLM Inference via Lookahead Parallel Decoding](https://huggingface.co/papers/2512.16229) (2025)\n* [Streaming-dLLM: Accelerating Diffusion LLMs via Suffix Pruning and Dynamic Decoding](https://huggingface.co/papers/2601.17917) (2026)\n* [DSB: Dynamic Sliding Block Scheduling for Diffusion LLMs](https://huggingface.co/papers/2602.05992) (2026)\n* [Reversible Diffusion Decoding for Diffusion Language Models](https://huggingface.co/papers/2602.00150) (2026)\n* [DiRL: An Efficient Post-Training Framework for Diffusion Language Models](https://huggingface.co/papers/2512.22234) (2025)\n* [Stable-DiffCoder: Pushing the Frontier of Code Diffusion Large Language Model](https://huggingface.co/papers/2601.15892) (2026)\n* [FOCUS: DLLMs Know How to Tame Their Compute Bound](https://huggingface.co/papers/2601.23278) (2026)\n\n\n Please give a thumbs up to this comment if you found it helpful!\n\n If you want recommendations for any Paper on Hugging Face checkout [this](https://huggingface.co/spaces/librarian-bots/recommend_similar_papers) Space\n\n You can directly ask Librarian Bot for paper recommendations by tagging it in a comment: `@librarian-bot recommend`","html":"

This is an automated message from the Librarian Bot. I found the following papers similar to this paper.

\n

The following papers were recommended by the Semantic Scholar API

\n

\n
• LoPA: Scaling dLLM Inference via Lookahead Parallel Decoding (2025)
\n
• Streaming-dLLM: Accelerating Diffusion LLMs via Suffix Pruning and Dynamic Decoding (2026)
\n
• DSB: Dynamic Sliding Block Scheduling for Diffusion LLMs (2026)
\n
• Reversible Diffusion Decoding for Diffusion Language Models (2026)
\n
• DiRL: An Efficient Post-Training Framework for Diffusion Language Models (2025)
\n
• Stable-DiffCoder: Pushing the Frontier of Code Diffusion Large Language Model (2026)
\n
• FOCUS: DLLMs Know How to Tame Their Compute Bound (2026)
\n

\n

Please give a thumbs up to this comment if you found it helpful!

\n

If you want recommendations for any Paper on Hugging Face checkout this Space

\n

You can directly ask Librarian Bot for paper recommendations by tagging it in a comment: \n\n@librarian-bot\n\t recommend

\n","updatedAt":"2026-02-11T01:43:47.400Z","author":{"_id":"63d3e0e8ff1384ce6c5dd17d","avatarUrl":"https://cdn-avatars.huggingface.co/v1/production/uploads/1674830754237-63d3e0e8ff1384ce6c5dd17d.jpeg","fullname":"Librarian Bot (Bot)","name":"librarian-bot","type":"user","isPro":false,"isHf":false,"isHfAdmin":false,"isMod":false,"followerCount":318,"isUserFollowing":false}},"numEdits":0,"identifiedLanguage":{"language":"en","probability":0.6700085997581482},"editors":["librarian-bot"],"editorAvatarUrls":["https://cdn-avatars.huggingface.co/v1/production/uploads/1674830754237-63d3e0e8ff1384ce6c5dd17d.jpeg"],"reactions":[],"isReport":false},"replies":[{"id":"698c9dbfdf518c09844cf108","author":{"_id":"65aa76b1cb5b4fb08ecb087c","avatarUrl":"https://cdn-avatars.huggingface.co/v1/production/uploads/65aa76b1cb5b4fb08ecb087c/0_x6Zb-H36XfugcsQ80Zh.jpeg","fullname":"Liu Yue","name":"Mzero17","type":"user","isPro":false,"isHf":false,"isHfAdmin":false,"isMod":false,"isUserFollowing":false},"createdAt":"2026-02-11T15:18:23.000Z","type":"comment","data":{"edited":false,"hidden":false,"latest":{"raw":"[Balancing Understanding and Generation in Discrete Diffusion Models](https://huggingface.co/papers/2602.01362) (2026) is particularly relevant to this paper, featuring a stationary noise kernel combining both absorbing mode (`M2T`) and uniform noise (`T2T`).","html":"

Balancing Understanding and Generation in Discrete Diffusion Models (2026) is particularly relevant to this paper, featuring a stationary noise kernel combining both absorbing mode (M2T) and uniform noise (T2T).

\n","updatedAt":"2026-02-11T15:18:23.024Z","author":{"_id":"65aa76b1cb5b4fb08ecb087c","avatarUrl":"https://cdn-avatars.huggingface.co/v1/production/uploads/65aa76b1cb5b4fb08ecb087c/0_x6Zb-H36XfugcsQ80Zh.jpeg","fullname":"Liu Yue","name":"Mzero17","type":"user","isPro":false,"isHf":false,"isHfAdmin":false,"isMod":false,"isUserFollowing":false}},"numEdits":0,"identifiedLanguage":{"language":"en","probability":0.7485569715499878},"editors":["Mzero17"],"editorAvatarUrls":["https://cdn-avatars.huggingface.co/v1/production/uploads/65aa76b1cb5b4fb08ecb087c/0_x6Zb-H36XfugcsQ80Zh.jpeg"],"reactions":[],"isReport":false,"parentCommentId":"698bded3c332ebffe5174457"}}]}],"primaryEmailConfirmed":false,"paper":{"id":"2602.08676","authors":[{"_id":"698ab6fd1b2dc6b37d61b04b","name":"Tiwei Bie","hidden":false},{"_id":"698ab6fd1b2dc6b37d61b04c","name":"Maosong Cao","hidden":false},{"_id":"698ab6fd1b2dc6b37d61b04d","name":"Xiang Cao","hidden":false},{"_id":"698ab6fd1b2dc6b37d61b04e","name":"Bingsen Chen","hidden":false},{"_id":"698ab6fd1b2dc6b37d61b04f","name":"Fuyuan Chen","hidden":false},{"_id":"698ab6fd1b2dc6b37d61b050","name":"Kun Chen","hidden":false},{"_id":"698ab6fd1b2dc6b37d61b051","name":"Lun Du","hidden":false},{"_id":"698ab6fd1b2dc6b37d61b052","name":"Daozhuo Feng","hidden":false},{"_id":"698ab6fd1b2dc6b37d61b053","name":"Haibo Feng","hidden":false},{"_id":"698ab6fd1b2dc6b37d61b054","name":"Mingliang Gong","hidden":false},{"_id":"698ab6fd1b2dc6b37d61b055","name":"Zhuocheng Gong","hidden":false},{"_id":"698ab6fd1b2dc6b37d61b056","name":"Yanmei Gu","hidden":false},{"_id":"698ab6fd1b2dc6b37d61b057","name":"Jian Guan","hidden":false},{"_id":"698ab6fd1b2dc6b37d61b058","user":{"_id":"6494f6dc32f2c0d7ef28315c","avatarUrl":"https://cdn-avatars.huggingface.co/v1/production/uploads/6494f6dc32f2c0d7ef28315c/y4TwcrT6UQ9Zfy6BfUMUl.jpeg","isPro":false,"fullname":"Kaiyuan Guan","user":"WilfredG","type":"user"},"name":"Kaiyuan Guan","status":"claimed_verified","statusLastChangedAt":"2026-02-10T09:05:52.604Z","hidden":false},{"_id":"698ab6fd1b2dc6b37d61b059","name":"Hongliang He","hidden":false},{"_id":"698ab6fd1b2dc6b37d61b05a","name":"Zenan Huang","hidden":false},{"_id":"698ab6fd1b2dc6b37d61b05b","name":"Juyong Jiang","hidden":false},{"_id":"698ab6fd1b2dc6b37d61b05c","name":"Zhonghui Jiang","hidden":false},{"_id":"698ab6fd1b2dc6b37d61b05d","name":"Zhenzhong Lan","hidden":false},{"_id":"698ab6fd1b2dc6b37d61b05e","name":"Chengxi Li","hidden":false},{"_id":"698ab6fd1b2dc6b37d61b05f","name":"Jianguo Li","hidden":false},{"_id":"698ab6fd1b2dc6b37d61b060","name":"Zehuan Li","hidden":false},{"_id":"698ab6fd1b2dc6b37d61b061","user":{"_id":"6470600790482b0e0f65a393","avatarUrl":"https://cdn-avatars.huggingface.co/v1/production/uploads/no-auth/fOUWWhf9jjsC9fgIiFlPc.png","isPro":false,"fullname":"Huabin","user":"Rookie-Liu","type":"user"},"name":"Huabin Liu","status":"claimed_verified","statusLastChangedAt":"2026-02-10T09:05:42.985Z","hidden":false},{"_id":"698ab6fd1b2dc6b37d61b062","name":"Lin Liu","hidden":false},{"_id":"698ab6fd1b2dc6b37d61b063","name":"Guoshan Lu","hidden":false},{"_id":"698ab6fd1b2dc6b37d61b064","name":"Yuan Lu","hidden":false},{"_id":"698ab6fd1b2dc6b37d61b065","name":"Yuxin Ma","hidden":false},{"_id":"698ab6fd1b2dc6b37d61b066","name":"Xingyu Mou","hidden":false},{"_id":"698ab6fd1b2dc6b37d61b067","name":"Zhenxuan Pan","hidden":false},{"_id":"698ab6fd1b2dc6b37d61b068","name":"Kaida Qiu","hidden":false},{"_id":"698ab6fd1b2dc6b37d61b069","name":"Yuji Ren","hidden":false},{"_id":"698ab6fd1b2dc6b37d61b06a","name":"Jianfeng Tan","hidden":false},{"_id":"698ab6fd1b2dc6b37d61b06b","user":{"_id":"5f94dea4cf95e81b6854e223","avatarUrl":"https://cdn-avatars.huggingface.co/v1/production/uploads/5f94dea4cf95e81b6854e223/r3VW8rb0wlJ7t_Lcpqi5f.jpeg","isPro":false,"fullname":"Dean Tian","user":"killa1218","type":"user"},"name":"Yiding Tian","status":"claimed_verified","statusLastChangedAt":"2026-02-10T09:05:47.064Z","hidden":false},{"_id":"698ab6fd1b2dc6b37d61b06c","name":"Zian Wang","hidden":false},{"_id":"698ab6fd1b2dc6b37d61b06d","name":"Lanning Wei","hidden":false},{"_id":"698ab6fd1b2dc6b37d61b06e","name":"Tao Wu","hidden":false},{"_id":"698ab6fd1b2dc6b37d61b06f","name":"Yipeng Xing","hidden":false},{"_id":"698ab6fd1b2dc6b37d61b070","user":{"_id":"65a97147cb5b4fb08e716afd","avatarUrl":"/avatars/7ddd4eb85d472cf42f71a3c34d659582.svg","isPro":false,"fullname":"Wentao Ye","user":"darklight03","type":"user"},"name":"Wentao Ye","status":"claimed_verified","statusLastChangedAt":"2026-02-11T11:19:02.750Z","hidden":false},{"_id":"698ab6fd1b2dc6b37d61b071","name":"Liangyu Zha","hidden":false},{"_id":"698ab6fd1b2dc6b37d61b072","name":"Tianze Zhang","hidden":false},{"_id":"698ab6fd1b2dc6b37d61b073","name":"Xiaolu Zhang","hidden":false},{"_id":"698ab6fd1b2dc6b37d61b074","user":{"_id":"6725f5a7f05f62659e3615f3","avatarUrl":"https://cdn-avatars.huggingface.co/v1/production/uploads/no-auth/iiXdzhaSnExw2dwQQtLZ8.png","isPro":false,"fullname":"Junbo Zhao","user":"jakezhao2024","type":"user"},"name":"Junbo Zhao","status":"claimed_verified","statusLastChangedAt":"2026-02-10T09:05:45.032Z","hidden":false},{"_id":"698ab6fd1b2dc6b37d61b075","name":"Da Zheng","hidden":false},{"_id":"698ab6fd1b2dc6b37d61b076","name":"Hao Zhong","hidden":false},{"_id":"698ab6fd1b2dc6b37d61b077","name":"Wanli Zhong","hidden":false},{"_id":"698ab6fd1b2dc6b37d61b078","name":"Jun Zhou","hidden":false},{"_id":"698ab6fd1b2dc6b37d61b079","user":{"_id":"63eb008e5c837d9968f1eb71","avatarUrl":"/avatars/ae43c3f5ab87b82f4bad25c65ac55d01.svg","isPro":false,"fullname":"Junlin Zhou","user":"jlzhou","type":"user"},"name":"Junlin Zhou","status":"claimed_verified","statusLastChangedAt":"2026-02-10T09:05:49.495Z","hidden":false},{"_id":"698ab6fd1b2dc6b37d61b07a","name":"Liwang Zhu","hidden":false},{"_id":"698ab6fd1b2dc6b37d61b07b","user":{"_id":"632179745fc60c44fd91fc33","avatarUrl":"/avatars/37d4fefbcc19f091dccffefec9706de2.svg","isPro":false,"fullname":"zhumuzhi","user":"Z-MU-Z","type":"user"},"name":"Muzhi Zhu","status":"claimed_verified","statusLastChangedAt":"2026-02-10T09:05:33.817Z","hidden":false},{"_id":"698ab6fd1b2dc6b37d61b07c","user":{"_id":"673b5f24e863f1d28b402efc","avatarUrl":"https://cdn-avatars.huggingface.co/v1/production/uploads/no-auth/19gUgtPEY3-FtY0sNlI_-.png","isPro":false,"fullname":"yihongzhuang","user":"utdawn","type":"user"},"name":"Yihong Zhuang","status":"claimed_verified","statusLastChangedAt":"2026-02-10T09:05:55.143Z","hidden":false}],"publishedAt":"2026-02-09T14:00:07.000Z","submittedOnDailyAt":"2026-02-10T02:14:10.029Z","title":"LLaDA2.1: Speeding Up Text Diffusion via Token Editing","submittedOnDailyBy":{"_id":"673b5f24e863f1d28b402efc","avatarUrl":"https://cdn-avatars.huggingface.co/v1/production/uploads/no-auth/19gUgtPEY3-FtY0sNlI_-.png","isPro":false,"fullname":"yihongzhuang","user":"utdawn","type":"user"},"summary":"While LLaDA2.0 showcased the scaling potential of 100B-level block-diffusion models and their inherent parallelization, the delicate equilibrium between decoding speed and generation quality has remained an elusive frontier. Today, we unveil LLaDA2.1, a paradigm shift designed to transcend this trade-off. By seamlessly weaving Token-to-Token (T2T) editing into the conventional Mask-to-Token (M2T) scheme, we introduce a joint, configurable threshold-decoding scheme. This structural innovation gives rise to two distinct personas: the Speedy Mode (S Mode), which audaciously lowers the M2T threshold to bypass traditional constraints while relying on T2T to refine the output; and the Quality Mode (Q Mode), which leans into conservative thresholds to secure superior benchmark performances with manageable efficiency degrade. Furthering this evolution, underpinned by an expansive context window, we implement the first large-scale Reinforcement Learning (RL) framework specifically tailored for dLLMs, anchored by specialized techniques for stable gradient estimation. This alignment not only sharpens reasoning precision but also elevates instruction-following fidelity, bridging the chasm between diffusion dynamics and complex human intent. We culminate this work by releasing LLaDA2.1-Mini (16B) and LLaDA2.1-Flash (100B). Across 33 rigorous benchmarks, LLaDA2.1 delivers strong task performance and lightning-fast decoding speed. Despite its 100B volume, on coding tasks it attains an astounding 892 TPS on HumanEval+, 801 TPS on BigCodeBench, and 663 TPS on LiveCodeBench.","upvotes":66,"discussionId":"698ab6fd1b2dc6b37d61b07d","githubRepo":"https://github.com/inclusionAI/LLaDA2.X","githubRepoAddedBy":"user","ai_summary":"LLaDA2.1 introduces a novel token-to-token editing approach with speed and quality modes, enhanced through reinforcement learning for improved reasoning and instruction following in large language diffusion models.","ai_keywords":["block-diffusion models","decoding speed","generation quality","Token-to-Token editing","Mask-to-Token scheme","threshold-decoding scheme","Speedy Mode","Quality Mode","Reinforcement Learning","gradient estimation","reasoning precision","instruction-following","large language diffusion models","HumanEval+","BigCodeBench","LiveCodeBench"],"githubStars":345,"organization":{"_id":"67aea5c8f086ab0f70ed97c9","name":"inclusionAI","fullname":"inclusionAI","avatar":"https://cdn-uploads.huggingface.co/production/uploads/662e1f9da266499277937d33/fyKuazRifqiaIO34xrhhm.jpeg"}},"canReadDatabase":false,"canManagePapers":false,"canSubmit":false,"hasHfLevelAccess":false,"upvoted":false,"upvoters":[{"_id":"673b5f24e863f1d28b402efc","avatarUrl":"https://cdn-avatars.huggingface.co/v1/production/uploads/no-auth/19gUgtPEY3-FtY0sNlI_-.png","isPro":false,"fullname":"yihongzhuang","user":"utdawn","type":"user"},{"_id":"67aeffda7330db26f93cd62f","avatarUrl":"https://cdn-avatars.huggingface.co/v1/production/uploads/67aeffda7330db26f93cd62f/ktdFt5_F05qNJ-dnD_Fk3.jpeg","isPro":false,"fullname":"jiangbonadia","user":"NadiaJiang","type":"user"},{"_id":"6494f6dc32f2c0d7ef28315c","avatarUrl":"https://cdn-avatars.huggingface.co/v1/production/uploads/6494f6dc32f2c0d7ef28315c/y4TwcrT6UQ9Zfy6BfUMUl.jpeg","isPro":false,"fullname":"Kaiyuan Guan","user":"WilfredG","type":"user"},{"_id":"68f9705953dd7b9778dc9156","avatarUrl":"/avatars/3a2468cfce1004010bb2175bf76b7479.svg","isPro":false,"fullname":"Lan","user":"Zhenzhong-Danny","type":"user"},{"_id":"63eb008e5c837d9968f1eb71","avatarUrl":"/avatars/ae43c3f5ab87b82f4bad25c65ac55d01.svg","isPro":false,"fullname":"Junlin Zhou","user":"jlzhou","type":"user"},{"_id":"6909e9153e33b19bf2f71b05","avatarUrl":"/avatars/8fa060fa5c21ce57c2e5f87a0835af07.svg","isPro":false,"fullname":"Mingliang Gong","user":"bright-ai-infra","type":"user"},{"_id":"5f94dea4cf95e81b6854e223","avatarUrl":"https://cdn-avatars.huggingface.co/v1/production/uploads/5f94dea4cf95e81b6854e223/r3VW8rb0wlJ7t_Lcpqi5f.jpeg","isPro":false,"fullname":"Dean Tian","user":"killa1218","type":"user"},{"_id":"64d2eaee80ce1029be72bdcd","avatarUrl":"/avatars/6917aa66459d6f9a4d6437381c496bdd.svg","isPro":false,"fullname":"Jianfeng Tan","user":"jianfengt","type":"user"},{"_id":"63465eb53588a193ca4522cb","avatarUrl":"/avatars/7a85b9e0e0317a6d13af51625bc30b23.svg","isPro":false,"fullname":"Finn Chan","user":"finnn","type":"user"},{"_id":"66602d7cf0b78aefceabf8aa","avatarUrl":"/avatars/e4dda04cc613fec83061a1083762f6c1.svg","isPro":false,"fullname":"lan","user":"wei19972","type":"user"},{"_id":"631ad3ab2b8c9d3d400cd51d","avatarUrl":"/avatars/76bd63f95e4d38ff9d1e3036a3dbdb49.svg","isPro":false,"fullname":"Liangyu Zha","user":"qianxiao1111","type":"user"},{"_id":"641585548b8245b157e15c5a","avatarUrl":"/avatars/c01a98d57733cfc29b549b974d4e2da4.svg","isPro":false,"fullname":"maosongcao","user":"bittersweet","type":"user"}],"acceptLanguages":["*"],"dailyPaperRank":0,"organization":{"_id":"67aea5c8f086ab0f70ed97c9","name":"inclusionAI","fullname":"inclusionAI","avatar":"https://cdn-uploads.huggingface.co/production/uploads/662e1f9da266499277937d33/fyKuazRifqiaIO34xrhhm.jpeg"}}">