[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"article-low-complexity-beamspace-denoiser-mmwave-mimo-zh":3,"tags-low-complexity-beamspace-denoiser-mmwave-mimo-zh":38,"related-lang-low-complexity-beamspace-denoiser-mmwave-mimo-zh":49,"related-posts-low-complexity-beamspace-denoiser-mmwave-mimo-zh":53,"series-research-0c02225c-d6ff-44f8-bc92-884c8921c4a3":90},{"id":4,"title":5,"content":6,"summary":7,"source":8,"source_url":9,"author":10,"image_url":11,"keywords":12,"language":19,"translated_content":10,"views":20,"is_premium":21,"created_at":22,"updated_at":22,"cover_image":11,"published_at":23,"rewrite_status":24,"rewrite_error":10,"rewritten_from_id":25,"slug":26,"category":27,"related_article_id":28,"status":29,"google_indexed_at":30,"x_posted_at":10,"tweet_text":10,"title_rewritten_at":10,"title_original":10,"key_takeaways":31,"topic_cluster_id":35,"embedding":36,"is_canonical_seed":37},"0c02225c-d6ff-44f8-bc92-884c8921c4a3","更簡單的毫米波波束域去噪器","\u003Cp data-speakable=\"summary\">這篇論文提出一個低複雜度的毫米波 massive MIMO 波束域去噪方法，目標是把低解析度 ADC 的失真一起處理掉，還能做成比較適合硬體落地的版本。\u003C\u002Fp>\u003Cp>毫米波 massive MIMO 的通道估計本來就不輕鬆。天線數量大、通道又稀疏，再加上低解析度 ADC 會帶來額外量化雜訊，整個問題很容易從「估得準不準」變成「算得動算不動」。這篇論文就是在解這個組合拳。\u003C\u002Fp>\u003Cp>它的重點不是再做一個更複雜的最佳化器，而是把去噪流程壓到更簡單。作者明確把方法朝硬體友善方向設計，還提到 VLSI 架構與 FPGA 實作。對做無線 DSP、bas\u003Ca href=\"\u002Fnews\u002Fweb3-communication-trust-infrastructure-2026-zh\">eb\u003C\u002Fa>and pipel\u003Ca href=\"\u002Fnews\u002Fminimax-m1-open-hybrid-attention-reasoning-model-zh\">in\u003C\u002Fa>e、或想把演算法放上 FPGA 的開發者來說，這種取向很直接：少一點理論上的華麗，多一點實際可部署性。\u003C\u002Fp>\u003Ch2>這篇論文想解什麼痛點\u003C\u002Fh2>\u003Cp>論文鎖定的是使用低解析度 ADC 的 mmWave massive MIMO 系統。低解析度 ADC 的好處很現實：可以降低功耗，也能壓低硬體成本。但代價也很明顯，就是量化雜訊會讓通道估計更難做。\u003C\u002Fp>\n\u003Cfigure class=\"my-6\">\u003Cimg src=\"https:\u002F\u002Fxxdpdyhzhpamafnrdkyq.supabase.co\u002Fstorage\u002Fv1\u002Fobject\u002Fpublic\u002Fcovers\u002Finline-1778814650361-xtc2.png\" alt=\"更簡單的毫米波波束域去噪器\" class=\"rounded-xl w-full\" loading=\"lazy\" \u002F>\u003C\u002Ffigure>\n\u003Cp>另一方面，毫米波通道在 beamspace 裡本來就有稀疏性。這代表大多數 beamspace 分量其實不是有用訊號，而是偏弱或接近雜訊的成分。問題就在於，如何把真正有意義的分量挑出來，同時不要付出太高的計算成本。\u003C\u002Fp>\u003Cp>作者想處理的不是單一問題，而是兩個限制一起來：一是通道稀疏，二是 ADC 失真。這也是這篇工作的核心價值。它不是只針對理想通道做去噪，而是直接把低解析度硬體的現實一起納入模型。\u003C\u002Fp>\u003Ch2>方法到底怎麼運作\u003C\u002Fh2>\u003Cp>這個方法的核心想法，可以理解成一個 Bayesian binary hypothesis testing。白話一點，就是對每個 beamspace 分量判斷：它比較像是「有用訊號」還是「雜訊主導」。\u003C\u002Fp>\u003Cp>為了支撐這個判斷，論文採用 Bernoulli-complex Gaussian prior。這種先驗的意思是，beamspace 通道大多是稀疏的，只有少數分量會真的活躍；而那些活躍分量則用複數高斯分布來描述。這和毫米波通道在 beamspace 裡的稀疏特性是對得上的。\u003C\u002Fp>\u003Cp>另一個關鍵設計，是作者沒有把熱雜訊和量化雜訊拆開來複雜處理，而是把它們合併成一個 composite noise term。這樣做的好處很直接：模型比較簡潔，推導也比較好做，但仍然有把 ADC 造成的失真算進去。\u003C\u002Fp>\u003Cp>基於這個模型，作者推導出 closed-form threshold，接著用 hard-thresholding 的方式做去噪。也就是說，每個分量不需要反覆迭代，也不需要在一堆參數裡搜尋最好的設定，而是可以直接做保留或丟棄的判斷。\u003C\u002Fp>\u003Cp>這正是它強調低複雜度的原因。論文明講，這個方法避開了矩陣反矩陣、迭代最佳化、參數搜尋這些重操作，並且讓計算複雜度對天線數量呈現近線性成長。對實作端來說，這種設計會比很多「理論上很漂亮」的方法更容易往硬體搬。\u003C\u002Fp>\u003Cp>如果把它拆成工程流程，可以粗略理解成下面這樣：\u003C\u002Fp>\u003Cul>\u003Cli>先把 beamspace 當成稀疏訊號來看。\u003C\u002Fli>\u003Cli>把低解析度 ADC 的量化雜訊納入統一噪聲模型。\u003C\u002Fli>\u003Cli>對每個分量做訊號\u002F雜訊的二元判斷。\u003C\u002Fli>\u003Cli>用閉式門檻做硬閾值去噪。\u003C\u002Fli>\u003Cli>避免矩陣反演與迭代式求解。\u003C\u002Fli>\u003C\u002Ful>\u003Ch2>論文實際證明了什麼\u003C\u002Fh2>\u003Cp>從摘要提供的資訊來看，作者主張這個算法能在維持接近既有高計算量方法表現的同時，把計算成本明顯壓低。也就是說，它要解的是一個典型 trade-off：效果不要掉太多，但成本要明顯下降。\u003C\u002Fp>\n\u003Cfigure class=\"my-6\">\u003Cimg src=\"https:\u002F\u002Fxxdpdyhzhpamafnrdkyq.supabase.co\u002Fstorage\u002Fv1\u002Fobject\u002Fpublic\u002Fcovers\u002Finline-1778814654731-17r3.png\" alt=\"更簡單的毫米波波束域去噪器\" class=\"rounded-xl w-full\" loading=\"lazy\" \u002F>\u003C\u002Ffigure>\n\u003Cp>這篇工作的另一個重點不是只停在演算法層，而是往硬體落地走。作者還做了 hardware-efficient VLSI architecture，並且把設計實作在 AMD-Xilinx Kintex UltraScale+ KCU116 FPGA 平台上。這代表它不是只在模擬環境裡說自己省算力，而是把硬體實作也納入貢獻。\u003C\u002Fp>\u003Cp>根據摘要，這個硬體設計透過針對實作限制的簡化與有效的處理結構，帶來更低的延遲與更少的硬體資源使用量，並且在天線數增加時呈現 sublinear scaling。這些都是很工程導向的訊號，表示作者關心的不只是數學形式，也包括實際部署的成本。\u003C\u002Fp>\u003Cp>不過，這裡也要講清楚：摘要沒有公開完整 \u003Ca href=\"\u002Ftag\u002Fbenchmark\">benchmark\u003C\u002Fa> 細節。它沒有提供明確的 latency 數字、resource 數量、或是和哪些 baseline 比到\u003Ca href=\"\u002Fnews\u002Fai-benchmark-wins-cyber-scare-defenders-zh\">什麼\u003C\u002Fa>程度的完整數據。因此，若要判斷實際性能差距，還是得回到論文正文看完整實驗設定。\u003C\u002Fp>\u003Cp>就目前 raw 資料能確定的內容來說，這篇論文至少證明了兩件事：第一，beamspace sparsity 可以被用來做簡潔的硬閾值去噪；第二，這個設計不只是概念上低複雜度，也被作者往 FPGA 實作方向推了一步。\u003C\u002Fp>\u003Ch2>對開發者有什麼影響\u003C\u002Fh2>\u003Cp>如果你是做無線通訊、DSP、或硬體加速的人，這篇論文的價值在於它很明確地站在「可部署」那一邊。很多通道估計方法在 paper 裡看起來很強，但一碰到 FPGA、ASIC、或低功耗 baseband 平台就會卡住，因為矩陣運算和迭代流程太重。\u003C\u002Fp>\u003Cp>這篇的思路剛好相反。它先接受 beamspace 的稀疏結構，再把低解析度 ADC 的失真合併進簡化模型，最後用閉式門檻做出決策。這種做法對硬體工程很友善，因為它把不必要的運算拿掉了。\u003C\u002Fp>\u003Cp>對實作端來說，這也意味著幾個方向值得注意。像是如果你在做 mmWave receiver prototype，或是在評估低功耗設計時刻意降低 ADC 解析度，這種方法就很有參考價值。它不是要把每個物理效應都完整建模，而是要找出一個夠快、夠簡單、又能支撐實際判斷的統計模型。\u003C\u002Fp>\u003Cp>這也是這篇論文比較值得被記住的地方：它不是單純提出一個新公式，而是把演算法設計、噪聲模型、和硬體實作拉在一起看。對開發者而言，這類研究的意義往往不只在準確率，而是在能不能真的塞進系統裡跑。\u003C\u002Fp>\u003Ch2>限制與還沒回答的問題\u003C\u002Fh2>\u003Cp>先講最直接的限制。摘要沒有給出完整 benchmark 數字，所以我們目前不知道它在不同場景下的實際提升幅度，也不知道和哪些既有方法相比，差距到底有多大。這會影響你判斷它是不是值得換掉既有方案。\u003C\u002Fp>\u003Cp>另外，這個方法建立在 Bernoulli-complex Gaussian prior 與 composite noise approximation 上。這代表它的效果會依賴模型假設是否貼近真實系統。摘要沒有說明這個閉式門檻對模型偏差有多敏感，也沒有說不同 ADC 解析度下的穩定性如何。\u003C\u002Fp>\u003Cp>硬體部分也是一樣。雖然摘要提到 FPGA 實作、低延遲、低資源使用量，以及 sublinear scaling，但沒有提供細部資源拆解。對工程師來說，這些資訊很重要，因為「省多少」和「怎麼省」會直接影響是否能移植到其他平台。\u003C\u002Fp>\u003Cp>所以，這篇論文目前最穩的結論不是「它已經完美解決 mmWave 通道估計」，而是「它提供了一條更簡單、也更硬體導向的路線」。如果你的目標是做可落地的毫米波接收器，這條路線值得關注；如果你要的是完整的性能比較與部署成本分析，還是得看正文的實驗與實作細節。\u003C\u002Fp>\u003Cp>整體來看，這篇工作很像是在提醒開發者：在 mmWave massive MIMO 這種又稀疏、又吵、又吃硬體的場景裡，最有價值的未必是最複雜的方法，而是能把問題切乾淨、再用最少的運算做出可靠判斷的方法。\u003C\u002Fp>","這篇論文提出一個低複雜度的毫米波 massive MIMO 波束域去噪方法，結合低解析度 ADC 雜訊模型與硬體友善設計，目標是讓演算法更適合 FPGA 落地。","arxiv.org","https:\u002F\u002Farxiv.org\u002Fabs\u002F2605.08855",null,"https:\u002F\u002Fxxdpdyhzhpamafnrdkyq.supabase.co\u002Fstorage\u002Fv1\u002Fobject\u002Fpublic\u002Fcovers\u002Finline-1778814650361-xtc2.png",[13,14,15,16,17,18],"mmWave MIMO","beamspace denoising","low-resolution ADC","hard thresholding","FPGA","massive MIMO","zh",0,false,"2026-05-15T03:10:30.06639+00:00","2026-05-15T03:10:29.901+00:00","done","b105aac1-982d-43d6-ab9d-fff9d9d2b72e","low-complexity-beamspace-denoiser-mmwave-mimo-zh","research","712a0357-f7cd-48f2-adde-c2691da0815f","published","2026-05-15T09:00:16.656+00:00",[32,33,34],"把 beamspace sparsity 和低解析度 ADC 雜訊一起納入簡化模型。","用 closed-form threshold 和 hard-thresholding 取代矩陣反演與迭代最佳化。","作者還做了 FPGA 實作，強調低延遲、低資源與較好的部署性。","5fa30296-f388-4653-96e0-bc24f62780b7","[-0.018246237,0.0025053069,0.0099549955,-0.07361838,-0.031095961,-0.011955442,0.0012653219,-0.013442765,0.021844095,0.03303634,-0.025488866,-0.0018404261,0.017243663,0.01830431,0.11636933,0.035440546,-0.016843826,0.0022128776,-0.013961249,-0.022433214,-0.0063615222,0.0059716552,-0.010859059,0.010782822,-0.020052917,0.014894448,0.024954112,-0.0043237703,0.04230902,-0.013743086,0.0018168399,0.035737604,0.03880607,0.038369462,-0.002639351,0.023912312,0.024094742,-0.006948319,0.030852472,0.029068563,-0.010875976,-0.008624691,0.024848118,-0.014793752,-0.006520108,0.022004295,0.0048915045,-0.034485646,-0.0054012765,0.011200164,-0.0059063327,0.0046133962,-0.029832676,-0.14273582,0.0027326718,0.02543369,0.009841861,-0.008505064,0.00016061694,-0.012201026,-0.018342702,0.0065876236,-0.019314386,-0.019761613,0.009593983,-0.017906962,0.010906244,0.0058885133,-0.02431026,0.002715556,-0.0041886787,0.013830408,-0.007611976,0.0009831445,-0.01667176,-0.009983448,-0.006772366,0.01786621,-0.0068185045,0.026095506,-0.0008490779,0.0019002415,0.010869753,0.01032363,-0.008320028,0.0061361655,-0.00036831517,0.023734499,-0.0034660655,0.008507553,0.002848867,0.025634961,0.045064595,-0.009497812,0.007986272,0.015790796,0.018816607,-0.01543185,-0.027127681,-0.041544504,-0.0020562264,-0.01235119,-0.011742514,-0.01768074,0.019677578,0.016702216,0.0063016713,-0.039513983,0.016991366,-0.010855033,0.010343975,-0.0014149186,-0.01192239,0.0061887684,-0.0026402462,-0.12561253,0.0012060098,0.009968497,-0.007860672,0.008102614,-0.0033799303,0.004262074,0.02020048,0.017100837,0.0084462315,-0.012893936,-0.011295766,-0.009836146,-0.012004803,0.011150608,0.0068314155,0.02721174,-0.004684138,-0.00039229056,0.012810373,-0.011978339,0.0039602364,-0.0069899447,0.0023933526,-0.017799148,-0.022122491,0.03123936,-0.029885093,-0.010354538,-0.027145153,-0.011073415,-0.02689018,-0.0138701815,-0.013859324,0.01135872,0.015388668,-0.030871626,0.011650161,-0.00014488821,0.031206114,-0.023927141,0.010642083,0.022537922,-0.003579529,0.023997372,-0.0074200886,-0.0038524945,-0.02359141,-0.0033103968,-0.020593699,0.0149625605,-0.011341829,-0.0043264385,0.029762479,0.003621369,0.00875381,-0.023490354,-0.02452801,-0.008990209,-0.015839873,-0.0002549645,-0.0049014874,-0.0074863704,-0.00052447297,-0.02482733,-0.019768188,-0.0036052887,0.013685655,0.0025344638,-0.025202943,0.03310693,-0.0048367656,-0.028556421,0.0026092809,0.025080543,-0.0145270135,0.019383375,-0.0105799455,-0.0062284647,0.017806975,-0.05106756,0.0056342445,0.02275077,-0.008937429,0.010989689,0.00014946038,0.01302258,-0.006864817,-0.016125832,0.0030425743,-0.0064332285,0.008940134,-0.01653026,0.019899588,0.007804926,0.010369825,0.016401647,-0.001377241,-0.009844191,0.005103272,0.013388153,0.0074082655,0.0005290373,-0.005789356,-0.017334448,0.019366754,-0.028849043,-0.01223402,-0.0048275683,-0.0013464983,-0.0058994284,0.014775534,-0.0041500344,0.010285879,-0.012046246,-0.013836498,0.02635366,0.0007674762,-0.013191969,0.02766225,0.035629023,0.010694249,-0.0268724,0.011442275,-0.0058192485,-0.0035046188,0.019512191,0.00578685,0.00900607,0.0012411089,0.010632478,0.029974332,-0.0072682262,0.006591635,0.0111933565,0.0021860106,-0.014181644,0.006790337,-0.007553207,0.01245447,-0.0013692946,-0.025995322,0.004341215,-0.0033548458,0.015652118,-0.008491894,-0.005774594,-0.011024194,-0.0014047276,0.018948678,0.024859121,-0.0047130277,0.008000045,-0.003377005,0.008106939,-0.009538657,-0.011020039,-0.006046832,-0.009266795,0.014565858,0.007636828,-0.0464446,0.045272466,0.03724712,-0.016516216,0.0037549632,0.024222212,0.020028343,0.030389316,-0.0013799975,-0.015329092,-0.016120804,-0.012116994,0.009361989,-0.016563574,0.0040441537,0.012113571,-0.0062905014,-0.002778896,0.026250504,-0.013246794,0.012879189,-0.010906111,-0.022670642,0.0031244757,0.03063085,0.009089058,-0.0054288246,0.030396117,0.015728865,0.019438226,0.0028596073,0.015927298,0.00092842,0.009020287,-0.01792462,-0.018557034,0.022284562,-0.0008268072,-0.010605198,-0.036559734,0.025822744,0.026683003,-0.015815353,-0.02191901,0.0011308934,-0.033605907,-0.010326088,0.0064767087,-0.0024596571,0.009508651,-0.036324836,-0.023859758,0.011289493,0.003073736,0.015647383,0.027720701,-0.0017541958,-0.00044349622,-0.0109087275,-0.0061801514,-0.00093736564,-0.0143079385,-0.015092743,0.004363654,0.017381933,-0.0010121544,0.0041713575,-0.008684169,0.0010422729,0.018540375,-0.0006290882,0.0072529702,-0.010087373,-0.005882846,0.004126776,0.0039724587,0.017967042,-0.022709142,-0.019445412,0.00094688736,-0.0060972897,-0.0041041975,0.018064814,-0.016886083,0.017926447,-0.008563814,0.018456876,-0.013387598,-0.0019328184,-0.044493333,0.01309831,0.004471785,0.017930659,-0.028803203,-0.009920788,0.007494012,0.02852332,0.03407329,-0.008034705,0.00020317595,-0.012608398,0.026440982,-0.045970406,0.002050452,-0.0037409898,0.008328053,0.006913626,-0.019287992,-0.002676345,0.02258235,0.029890846,-0.02457361,-0.016157007,-0.012775974,0.010262946,-0.01801544,-0.0010328927,0.043895893,-0.006980472,0.013191687,0.013124881,-0.0074022957,-0.0055192853,-0.0076615275,-0.011170702,0.015568833,0.002151646,-0.027054116,-0.00721949,-0.044020552,-0.0033903655,-7.897995e-05,-0.014648095,-0.027677547,0.011069329,-0.010730339,0.00094511866,-0.003599869,0.014171301,0.00980183,-0.026002854,-0.016093949,0.005019283,-0.018922044,-0.011147031,0.0018830309,-0.0030631498,0.025731206,0.0177899,-0.03071869,-0.014478294,0.022049615,-0.0013392717,-0.01712717,0.0033974624,-0.002321358,-0.00064350385,-0.01743891,-0.018886136,-0.001369445,-0.012951189,0.016258612,-0.0028557756,-0.03666545,-0.023748346,-0.008400099,0.009087292,0.017352996,0.0032325692,-0.020758757,0.0064146933,-0.017573513,-0.0033813794,0.013086732,-0.004550024,0.018788401,0.017580274,-0.021236736,0.026322486,0.003581452,0.0024103418,0.02604579,-0.022157183,-0.022904139,-0.001225088,-0.035207104,0.004647115,0.0031995757,0.025087107,-0.021570876,-0.035843335,0.023983981,-0.019392222,0.015727816,-0.003694056,-0.016922863,-0.00029219588,0.02264032,0.0075658234,-0.00095474184,0.004531121,-0.032290228,-0.017750863,-0.002165648,0.0021497475,-0.020513343,-0.0058539463,0.00825749,0.022334492,0.011281568,-0.0039867028,0.03572084,-0.01411946,0.016536456,0.014656167,0.02038034,0.010195549,-0.00025328388,-0.010764455,-0.010350605,-0.051194217,-0.0220705,0.05583233,-0.02091092,0.02024234,-0.0004995768,0.007841244,-0.042370375,0.01317172,-0.006709045,0.0073987585,-0.0018166383,0.021169692,-0.0147755565,-0.006983916,0.019514756,-0.0086188875,0.010632337,0.012889958,0.01696213,0.009216493,0.0072223055,-0.026882766,-0.004354752,-0.020580359,-0.009880146,-0.010946387,0.013822799,-0.021374827,-0.013794695,0.007497104,0.017233048,0.01969437,0.012302046,-0.01880097,0.038406063,-0.0010176548,-0.015998986,0.034553885,0.005022586,-0.0018217567,0.01408605,0.012427514,0.0064132046,0.0019303517,-0.021850066,0.008769414,-0.0110624265,0.00024749726,-0.088913426,0.022493578,0.017545583,-0.029358761,0.013970969,-0.01439368,0.011576283,-0.020609703,-0.011416164,0.021841511,-0.007890777,0.01121829,0.0055744485,0.0199122,0.014963914,-0.022461966,0.005678825,0.012995222,0.027418206,-0.004764681,-0.0026277867,0.007524292,0.01734823,0.0073261806,-0.033403102,-0.0024862816,-0.002065218,0.0028985117,0.008108239,0.036103256,-0.009884199,-0.019052727,-0.008374515,-0.003646095,0.011635425,-0.01729941,0.051977504,-0.0027564177,0.0016958894,-0.0008974757,0.014423936,-0.0063756574,-0.022380788,-0.035648137,0.025817292,-0.010378671,-0.009336846,0.013346736,-0.019962521,0.004956472,-0.044177692,-0.03626484,-0.026438296,-0.032129813,0.0043964004,-0.02184627,-0.035693925,0.0017485719,-0.014111264,0.03639252,-0.006056639,-0.0049564824,-0.027014453,0.0017324697,0.011811111,-0.015296301,-0.015513846,0.03601875,0.0060212915,-0.008436238,-0.03481746,0.009283697,0.015462364,-0.0013394627,-0.039904382,-0.004805333,-0.026577814,0.028942242,-0.009766826,-0.0026505138,-0.024639864,-0.018546619,-0.10357163,-0.0058665625,-0.026747558,-0.02712401,0.0013323195,-0.029686706,0.014185536,-0.006158953,-0.009369264,-0.0051965965,0.018366061,0.00274244,0.019484818,-0.021378003,0.0040058917,0.014475688,-0.01812996,-0.0141868675,-0.013972848,-0.012547773,-0.006393493,-0.0028125008,0.027865855,0.010024487,-0.01180864,0.00093977776,-5.0697774e-05,-0.009750359,-0.015009586,-0.009741554,-0.010310066,-0.14505109,0.00914331,-0.0018450009,-0.016105747,0.0033877562,0.016780134,-0.024665132,-0.023181522,-0.016265376,-0.017215395,-0.0076844147,-0.047548663,-0.028536022,0.010587934,0.0032404743,0.09542758,-0.011607252,-0.023326546,-0.01653438,-0.04114623,0.013033151,0.0046078027,0.0055164606,0.021469133,-0.018964428,0.0040233633,0.02545127,-0.009407047,0.003234024,0.0073376824,-0.0038330418,0.004622034,-0.01714024,0.009688207,0.012648283,-0.02745161,-0.008613009,0.006860455,0.012165467,0.0299995,0.008157993,0.012407945,0.01233923,-0.017636385,-0.010139062,0.0006590649,0.010023408,-0.011636779,-0.04186327,0.0064266864,-0.028617414,-0.08007918,0.032502018,0.01422998,0.00025051384,-0.005528818,-0.013424415,-0.005763841,0.02094257,-0.01674287,-0.00744178,0.011546546,-0.018201588,0.0013723983,-0.0005236632,-0.008593852,0.0069474033,0.03717616,0.0063750814,-0.0026109933,-0.02895789,-0.00072680216,-0.005504405,-0.0033094087,-0.016986145,-0.049119983,0.005760755,0.01976019,0.015064748,-0.0052421275,-0.002351194,0.023242056,0.022359349,-0.03828255,0.0033533191,-0.019318964,-0.0074894913,0.027644156,0.013537841,0.02821812,0.0030716714,0.0075800265,0.030654624,0.0054380125,0.042403925,0.014991655,-0.031636477,0.026905613,-0.0017328961,0.024031805,0.018119654,-0.038916625,-0.018178651,-0.0012856686,0.011212171,0.014336299,-0.0280019,0.011862319,0.033763256,0.027459849]",true,[39,41,43,45,47],{"name":14,"slug":40},"beamspace-denoising",{"name":16,"slug":42},"hard-thresholding",{"name":13,"slug":44},"mmwave-mimo",{"name":15,"slug":46},"low-resolution-adc",{"name":17,"slug":48},"fpga",{"id":28,"slug":50,"title":51,"language":52},"low-complexity-beamspace-denoiser-mmwave-mimo-en","A simpler beamspace denoiser for mmWave MIMO","en",[54,60,66,72,78,84],{"id":55,"slug":56,"title":57,"cover_image":58,"image_url":58,"created_at":59,"category":27},"c15f45ee-a548-4dbf-8152-91de159c1a11","llmbda-calculus-agent-safety-rules-zh","LLMbda 演算替 AI 代理人立安全規則","https:\u002F\u002Fxxdpdyhzhpamafnrdkyq.supabase.co\u002Fstorage\u002Fv1\u002Fobject\u002Fpublic\u002Fcovers\u002Finline-1778825503412-mlbf.png","2026-05-15T06:10:34.832664+00:00",{"id":61,"slug":62,"title":63,"cover_image":64,"image_url":64,"created_at":65,"category":27},"9d27f967-62cc-433f-8cdb-9300937ade13","ai-benchmark-wins-cyber-scare-defenders-zh","為什麼 AI 基準賽在資安領域的勝利，應該讓防守方警醒","https:\u002F\u002Fxxdpdyhzhpamafnrdkyq.supabase.co\u002Fstorage\u002Fv1\u002Fobject\u002Fpublic\u002Fcovers\u002Finline-1778807450006-nofx.png","2026-05-15T01:10:29.379041+00:00",{"id":67,"slug":68,"title":69,"cover_image":70,"image_url":70,"created_at":71,"category":27},"bc402dc6-5da6-46fc-9d66-d09cb215f72b","why-linux-security-needs-patch-wave-mindset-zh","為什麼 Linux 安全需要「補丁浪潮」思維","https:\u002F\u002Fxxdpdyhzhpamafnrdkyq.supabase.co\u002Fstorage\u002Fv1\u002Fobject\u002Fpublic\u002Fcovers\u002Finline-1778741449813-s2wn.png","2026-05-14T06:50:24.052583+00:00",{"id":73,"slug":74,"title":75,"cover_image":76,"image_url":76,"created_at":77,"category":27},"d75b5708-d4ec-4c46-9592-fa0a68d4bc26","judge-reliability-harness-stress-tests-llm-judges-zh","LLM 評審也會不穩","https:\u002F\u002Fxxdpdyhzhpamafnrdkyq.supabase.co\u002Fstorage\u002Fv1\u002Fobject\u002Fpublic\u002Fcovers\u002Finline-1778740856189-g1zr.png","2026-05-14T06:40:32.198872+00:00",{"id":79,"slug":80,"title":81,"cover_image":82,"image_url":82,"created_at":83,"category":27},"941f698a-1dcf-4807-bd56-5295c07d2dee","taming-black-box-llm-inference-scheduling-zh","黑箱 LLM 排程更聰明了","https:\u002F\u002Fxxdpdyhzhpamafnrdkyq.supabase.co\u002Fstorage\u002Fv1\u002Fobject\u002Fpublic\u002Fcovers\u002Finline-1778740253221-wgy6.png","2026-05-14T06:30:31.546746+00:00",{"id":85,"slug":86,"title":87,"cover_image":88,"image_url":88,"created_at":89,"category":27},"0eb3d74f-c737-41a4-8b9b-fc30b2b3b0ac","aisafetybenchexplorer-ai-safety-benchmarks-zh","AISafetyBenchExplorer：AI 安全基準地圖","https:\u002F\u002Fxxdpdyhzhpamafnrdkyq.supabase.co\u002Fstorage\u002Fv1\u002Fobject\u002Fpublic\u002Fcovers\u002Finline-1778739657024-fgxt.png","2026-05-14T06:20:28.07969+00:00",[91,96,101,106,111,116,121,126,131,136],{"id":92,"slug":93,"title":94,"created_at":95},"f18dbadb-8c59-4723-84a4-6ad22746c77a","deepmind-bets-on-continuous-learning-ai-2026-zh","DeepMind 押注 2026 連續學習 AI","2026-03-26T08:16:02.367355+00:00",{"id":97,"slug":98,"title":99,"created_at":100},"f4a106cb-02a6-4508-8f39-9720a0a93cee","ml-papers-of-the-week-github-research-desk-zh","每週 ML 論文清單，為何紅到 GitHub","2026-03-27T01:11:39.284175+00:00",{"id":102,"slug":103,"title":104,"created_at":105},"c4f807ca-4e5f-47f1-a48c-961cf3fc44dc","ai-ml-conferences-to-watch-in-2026-zh","2026 AI 研討會投稿時程整理","2026-03-27T01:51:53.874432+00:00",{"id":107,"slug":108,"title":109,"created_at":110},"9f50561b-aebd-46ba-94a8-363198aa7091","openclaw-agents-manipulated-self-sabotage-zh","OpenClaw Agent 會自己搞砸自己","2026-03-28T03:03:18.786425+00:00",{"id":112,"slug":113,"title":114,"created_at":115},"11f22e92-7066-4978-a544-31f5f2156ec6","vega-learning-to-drive-with-natural-language-instructions-zh","Vega：使用自然語言指示進行自駕車控制","2026-03-28T14:54:04.847912+00:00",{"id":117,"slug":118,"title":119,"created_at":120},"a4c7cfec-8d0e-4fec-93cf-1b9699a530b8","drive-my-way-en-zh","Drive My Way：個性化自駕車風格的實現","2026-03-28T14:54:26.207495+00:00",{"id":122,"slug":123,"title":124,"created_at":125},"dec02f89-fd39-41ba-8e4d-11ede93a536d","training-knowledge-bases-with-writeback-rag-zh","用 WriteBack-RAG 強化知識庫提升檢索效能","2026-03-28T14:54:45.775606+00:00",{"id":127,"slug":128,"title":129,"created_at":130},"3886be5c-a137-40cc-b9e2-0bf18430c002","packforcing-efficient-long-video-generation-method-zh","PackForcing：短影片訓練也能生成長影片","2026-03-28T14:55:02.688141+00:00",{"id":132,"slug":133,"title":134,"created_at":135},"72b90667-d930-4cc9-8ced-aaa0f8968d44","pixelsmile-toward-fine-grained-facial-expression-editing-zh","PixelSmile：提升精細臉部表情編輯的新方法","2026-03-28T14:55:20.678181+00:00",{"id":137,"slug":138,"title":139,"created_at":140},"cf046742-efb2-4753-aef9-caed5da5e32e","adaptive-block-scaled-data-types-zh","IF4：神經網路量化的聰明選擇","2026-03-31T06:00:36.990273+00:00"]