[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"branding":3,"analytics":7,"article-r2lm-gets-parallel-text-generation-without-the-speed-penalty":10,"sections":42},{"siteName":4,"siteTagline":5,"publisherName":4,"contactEmail":6},"The Revision","Tech news, decoded.","editor@therevision.news",{"gaMeasurementId":8,"adsenseClientId":9},"G-ZW2MV82GYR","ca-pub-8533917693782264",{"article":11},{"id":12,"slug":13,"title":14,"dek":15,"body_md":16,"tags_json":17,"published_at":18,"created_at":19,"updated_at":20,"status":21,"review_note":22,"review_notes":23,"image_url":30,"persona_id":22,"persona_name":22,"section":31,"tags":32,"sources":37,"feedback":41,"feedback_at":22,"cost_usd":41,"total_tokens":41},2343,"r2lm-gets-parallel-text-generation-without-the-speed-penalty","R2LM Gets Parallel Text Generation Without the Speed Penalty","A new architecture pairs causal attention with a reverse Mamba sidecar to beat bidirectional diffusion models on average while keeping KV cache throughput.","A research paper out of arXiv proposes a way to have parallel text generation and fast cached inference at the same time — something existing diffusion language models haven't managed.\n\nDiscrete diffusion language models decode tokens in parallel rather than one at a time, which sounds fast until you hit the architecture wall: bidirectional attention gives you quality but kills KV caching, while causal attention keeps caching but throws away right-side context. The new paper introduces Bifocal dLLMs and a concrete implementation called R2LM (Right-to-Left Mamba). R2LM runs standard causal attention for left-side context — preserving full KV cache compatibility — while a lightweight reverse Mamba state-space model feeds in compressed right-side context without breaking cacheability. The researchers continued pretraining Qwen3-1.7B on 60 billion tokens to test it.\n\nThe throughput numbers are the headline: 2.4x to 12.9x faster than bidirectional diffusion models in batch serving, and 1.9x to 2.9x faster than standard autoregressive generation. On quality, R2LM exceeds the causal-only baseline on most benchmarks and surpasses the bidirectional model on average — a meaningful result, though not a clean sweep across every task.\n\nThe broader significance is that the KV cache problem has been the quiet ceiling on diffusion language model deployment; every lab experimenting with parallel decoding hits it. R2LM's Mamba sidecar approach is a pragmatic engineering answer rather than a theoretical one, which makes it easier to adopt — assuming the gains hold outside a single continued-pretraining setup.","[\"language-models\",\"diffusion-models\",\"inference\",\"ai-research\"]","2026-06-29T04:00:00.000Z","2026-06-29T05:09:38.598Z","2026-06-29T05:09:46.175Z","published",null,[24],{"id":25,"reviewer":26,"round":27,"reason":28,"status":29},"editor-r1","editor",1,"The headline and dek obscure the actual name of the architecture and misrepresent the quality claim — the source says R2LM 'exceeds the causal baseline on most benchmarks and surpasses the bidirectional dLLM on average,' but the dek claims it 'matches bidirectional quality,' which overstates the result; additionally, the body's closing caveat about 'most benchmarks' doing 'real work in that abstract' is an unresolved editorial note left in reader-facing copy.","resolved","https:\u002F\u002Fcdn.xyz.onl\u002Farticle-images\u002Fr2lm-gets-parallel-text-generation-without-the-speed-penalty.webp","ai",[33,34,35,36],"language-models","diffusion-models","inference","ai-research",[38],{"name":39,"url":40},"arXiv cs.AI","https:\u002F\u002Farxiv.org\u002Fabs\u002F2606.27732",0,{"sections":43},[44,48,53,58,63,68,73,78,83,88,93,97,102,107],{"name":45,"slug":31,"count":46,"latest_published_at":47},"AI",2590,"2026-07-16T04:00:00.000Z",{"name":49,"slug":50,"count":51,"latest_published_at":52},"Security","security",294,"2026-07-15T19:59:48.000Z",{"name":54,"slug":55,"count":56,"latest_published_at":57},"Deals","deals",179,"2026-06-29T20:02:07.000Z",{"name":59,"slug":60,"count":61,"latest_published_at":62},"Policy","policy",158,"2026-07-16T00:02:48.000Z",{"name":64,"slug":65,"count":66,"latest_published_at":67},"Hardware","hardware",122,"2026-07-14T19:46:26.000Z",{"name":69,"slug":70,"count":71,"latest_published_at":72},"Consumer Tech","consumer-tech",93,"2026-07-13T13:20:48.000Z",{"name":74,"slug":75,"count":76,"latest_published_at":77},"Software","software",70,"2026-07-13T19:52:25.000Z",{"name":79,"slug":80,"count":81,"latest_published_at":82},"Science","science",66,"2026-07-10T10:29:37.000Z",{"name":84,"slug":85,"count":86,"latest_published_at":87},"Dev Tools","dev-tools",59,"2026-07-07T04:00:00.000Z",{"name":89,"slug":90,"count":91,"latest_published_at":92},"Gaming","gaming",41,"2026-07-09T04:00:00.000Z",{"name":94,"slug":95,"count":91,"latest_published_at":96},"Startups","startups","2026-06-29T20:55:50.000Z",{"name":98,"slug":99,"count":100,"latest_published_at":101},"General","general",29,"2026-07-10T22:28:58.000Z",{"name":103,"slug":104,"count":105,"latest_published_at":106},"Reviews","reviews",20,"2026-06-24T12:00:01.000Z",{"name":108,"slug":109,"count":110,"latest_published_at":111},"How-To","how-to",6,"2026-06-16T09:00:00.000Z"]