[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"news-slug-peca-conciseness-qlora-medical":3,"topics-all":38,"news-related-0a4281e9-69b1-43f0-9e7b-78f0cd7ea4c2":67},{"id":4,"title":5,"summary":6,"content":7,"original_url":8,"source_id":9,"tags":10,"translations":24,"news_slug":31,"published_at":32,"created_at":33,"modified_at":34,"is_published":35,"publish_type":36,"image_url":14,"view_count":37},"0a4281e9-69b1-43f0-9e7b-78f0cd7ea4c2","医疗小模型回答砍72.8%,准确率反升3.2点","医疗小模型微调后爱啰嗦。COMSATS 与布莱恩特大学团队的预印本提出 PECA:在 QLoRA 里加一项,回答超 64 token 就抬高停机概率。Qwen3-1.7B 平均回答从 256.0 砍到 69.7,MedMCQA 准确率反从 40.8% 升到 44.0%;单卡 163 分钟,峰值显存 5.3 GB。","急诊室里,答案正确但塞满 256 个 token,医生还没读到结论,病人已经等不起——论文开篇就把这个死结点破:回答比可用时间更长时,再准确也算失败。11 月 12 日挂上 arXiv 的这篇预印本出自巴基斯坦 COMSATS 大学与美国布莱恩特大学的四位作者,方法叫 PECA,治的是医疗小模型微调后的通病:啰嗦。\n\n## 微调治好了成本,治不了话多\n\n医院跑不动旗舰模型:患者数据出不了院,越大的模型越难留在院内。于是约 10 亿到 130 亿参数的小模型加 QLoRA 微调,成了科室级部署的现实路线。但论文指出,参数高效微调只解决「训得起」,从不约束「说多长」——输出长度完全跟着训练分布走,不来自任何显式目标。而这批模型所学的医疗思维链语料本身就长:medical-o1-reasoning-SFT 英文分册有 19,704 条验证过的样本,平均每条推理轨迹 312.45 个词、答案 93.98 个词。模型有样学样,答起题来自然滔滔不绝。\n\n## 不罚字数,教模型停\n\n把「生成 token 数」直接写进损失函数行不通:token 数是离散量,对模型参数的梯度几乎处处为零,惩罚项推不动学习。PECA 换了个抓手——停机概率。一旦回答超过预算 B(论文固定取 64),损失函数就在其后每个位置抬高 EOS 结束符的概率,形成一个可微项加进交叉熵,力度由系数 λ 调节。全程不需要偏好数据、奖励模型、参考策略或强化学习采样,推理阶段也不加任何组件;论文称它是对照表里唯一同时做到免偏好数据、免教师模型、监督微调内可微的方法。\n\n## 砍七成,选择题反涨 3.2 分\n\n在 Qwen3-1.7B-Instruct 底座、8,000 条训练样本、贪心解码、128 token 上限的设定下,论文主表给出四个条件的对比:原始 QLoRA 平均回答 256.0 个 token,1,000 道 MedMCQA 选择题的精确匹配准确率 40.8%;λ=0.25 时平均回答缩到 69.7 个 token,压缩 72.8%,准确率 44.0%;λ=0.50 时长度 66.2,准确率掉到 31.9%;λ=1.00 时长度 65.2,准确率只剩 4.6%,多数回答只剩一两个 token。简洁压力是个旋钮:拧过头,长度几乎不再变,准确率先崩。\n\n两个细节支撑「砍掉的是废话」的解读:其一,衡量与参考答案重合度的 ROUGE-L 与 BERTScore 几乎没动,论文由此判断切口落在对冲、重复这类填充上,而不是承重内容;其二,回答长度分布整体左移,而非长尾被截断——模型是学会了何时停,不是被剪刀裁短。论文还给了个对照案例:32 周孕妇持续阴道出血伴腹痛,原模型在 256 token 里逐条列鉴别诊断,结论没写就到顶;PECA 模型用 68 个 token 走完推理、给出诊断,省 73.4%。\n\n## 成本与作者自列的折扣\n\n成本账很轻:每个条件训练 153 到 163 分钟,单张 A10G 或 L4,峰值显存 4.4 到 5.3 GB,16 GB 的免费层云卡装得下。作者自己列的折扣也得照抄:单随机种子、单模型家族;绝对分数是 1.7B 模型在高难基准上的能力上限,不等于可用的临床决策系统;对比 RL 路线 SmartThinker 在数学推理上报告的 52.5% 压缩,论文明确写了两个数字不可直接比较;在相同底座上重训偏好优化基线的正面交锋留作未来工作;代码与适配器要等论文录用后才公开,眼下没有可查的仓库。\n\n## 所以呢\n\n最有意思的推论其实越过了医疗:领域模型的啰嗦,多半不是推理的必要代价,而是训练分布的残留。压下去不需要奖励模型,也不需要强化学习——在监督微调里放一个「何时停」的信号就够了。下次你的领域模型答得又长又绕,先别急着怪它笨,可能只是从来没人教过它停。\n\n来源:[arXiv 论文页](https:\u002F\u002Farxiv.org\u002Fabs\u002F2611.17128);[HTML 全文](https:\u002F\u002Farxiv.org\u002Fhtml\u002F2611.17128v1)。未经同行评审的预印本,文中数字均为论文自报。\n","https:\u002F\u002Farxiv.org\u002Fabs\u002F2611.17128","7437aeb9-930c-4866-a2e9-48003c1a792b",[11,15,18,21],{"id":12,"name":13,"slug":13,"description":14,"color":14},"40269b40-7942-4650-9672-ed2e6524d37a","ai-technology",null,{"id":16,"name":17,"slug":17,"description":14,"color":14},"0ef8513a-0a26-42f0-b6f9-5b6dadded45c","efficiency",{"id":19,"name":20,"slug":20,"description":14,"color":14},"01598627-1ea6-4b27-a5d8-874971571a71","llm",{"id":22,"name":23,"slug":23,"description":14,"color":14},"045c011e-e2bb-45ce-bdd6-0c927f8a3b87","token-efficiency",[25],{"id":26,"lang":27,"title":28,"summary":29,"content":30},"fc7b570b-e686-421b-add9-eb264d6148ae","en","PECA: 72.8% Shorter Answers, 3.2-Point Accuracy Gain","PECA raises EOS probability past a 64-token budget inside QLoRA. Qwen3-1.7B replies shrink 256.0 to 69.7 tokens; MedMCQA accuracy climbs to 44.0%.","At the bedside, a correct answer that arrives too long is worthless: no clinician can wait out a 256-token differential while the patient is still bleeding. A preprint posted to arXiv on 12 November 2026 opens on exactly this knot — an answer that takes longer than the time available has failed, however accurate it is. The four authors, from COMSATS University Islamabad in Pakistan and Bryant University in the United States, call their method PECA, and aim it at a habit that parameter-efficient fine-tuning never touches: verbosity.\n\n## Fine-tuning fixes the cost, not the length\n\nHospitals struggle to host frontier models: patient data cannot leave the building, and the largest systems are the hardest to keep inside it. Small language models of roughly one to thirteen billion parameters, adapted with QLoRA, are the realistic route for department-level deployment. But as the paper puts it, PEFT solves the cost of training and leaves output length alone — length follows the training distribution, never an explicit objective. The medical chain-of-thought corpus the authors train on is itself long: the English split of medical-o1-reasoning-SFT holds 19,704 verified examples, with an average reasoning trace of 312.45 words and an average response of 93.98 words. A model trained on it learns the habit.\n\n## Penalize stopping, not counting\n\nYou cannot write the token count directly into the loss: the count is discrete, its gradient is zero almost everywhere, so a naive length penalty cannot drive learning. PECA acts on the stop probability instead. Once a response passes a budget B (fixed at 64), the loss raises the end-of-sequence probability at every later position — a smooth, differentiable term added to the cross-entropy, scaled by a strength knob λ. No preference pairs, no reward model, no reference policy, no RL rollouts, and nothing extra at inference. The paper describes it as the only entry in its comparison table that is at once preference-free, teacher-free, and differentiable within supervised fine-tuning.\n\n## Seventy percent shorter, accuracy up\n\nOn a Qwen3-1.7B-Instruct backbone with 8,000 training examples, greedy decoding, and a 128-token cap, the main table reads: vanilla QLoRA averages 256.0 tokens with 40.8% exact-match accuracy on 1,000 MedMCQA items; at λ=0.25 the mean falls to 69.7 tokens, a 72.8% cut, while accuracy rises to 44.0%; at λ=0.50 accuracy drops to 31.9%; at λ=1.00 it collapses to 4.6%, most replies down to one or two tokens. Conciseness pressure is a dial, not a slider to max: past a point, length barely moves while accuracy falls first.\n\nTwo details carry the interpretation. ROUGE-L and BERTScore barely move from baseline, which the paper reads as cuts falling on hedging and repetition rather than load-bearing content. And the whole length distribution slides left instead of being clipped — the model learned when to stop. A vignette makes it concrete: a 32-week pregnant woman with painful, continuous vaginal bleeding; the vanilla model is still listing differentials when the 256-token ceiling arrives and never states a diagnosis, while the PECA model finishes in 68 tokens with the diagnosis stated — a 73.4% cut.\n\n## Cost and the discounts the authors list themselves\n\nEach condition trained in 153 to 163 minutes on a single NVIDIA A10G or L4, peaking at 4.4 to 5.3 GB — inside a free-tier 16 GB cloud card. The authors discount their own results plainly: a single random seed and a single model family; the absolute MedMCQA figures are the ceiling of a 1.7B model on a demanding benchmark, not a clinical decision-support system; against the RL line SmartThinker and its reported 52.5% reduction on mathematical reasoning, the paper itself notes the two figures are not directly comparable; re-training preference baselines on the identical backbone is left as future work; code and adapters appear only upon publication, so there is no repository to check yet.\n\n## So what\n\nThe interesting inference reaches past medicine: a domain model's verbosity is mostly residue of its training distribution, not the price of reasoning. Suppressing it does not require a reward model or reinforcement learning — one well-placed signal about when to stop, inside ordinary supervised fine-tuning, is enough. Next time your fine-tuned model rambles, ask whether anyone ever taught it to stop.\n\nPaper: https:\u002F\u002Farxiv.org\u002Fabs\u002F2611.17128 (arXiv preprint, not peer reviewed; all figures are self-reported by the paper).\n","peca-conciseness-qlora-medical","2026-11-15T15:30:00Z","2026-11-15T15:11:10.663081Z","2026-11-15T15:11:10.663091Z",true,"agent",126,[39,49,58],{"slug":40,"tag_slug":41,"title_zh":42,"title_en":43,"intro_zh":44,"intro_en":45,"id":46,"is_active":35,"created_at":47,"modified_at":48},"image-generation","text-to-image","图像生成：从文生图到逐框编辑的模型竞赛","Image Generation: from text-to-image to frame-level editing","文生图已是红海，战场正转向编辑与可控性：FLUX.2\u002F3 用开源和量化把生图搬进消费级显卡，Qwen-Image、Seedream 等国产模型紧咬不放，GPT Image 2 则把推理能力带进像素。本专题追踪图像生成与编辑模型的发布、架构演进、加速量化与商业化数据，看「好图」如何变成「可用的图」。","Text-to-image is a crowded race, and the frontier has moved to editing and control: FLUX.2\u002F3 push open weights and quantization onto consumer GPUs, Qwen-Image and Seedream close in from China, and GPT Image 2 brings reasoning into pixels. This topic tracks model releases, architecture shifts, acceleration work, and the business numbers behind \"pretty\" becoming \"usable\".","52de7c59-7864-4569-8ae7-3d656d26a1a1","2026-11-20T07:20:23.534506Z","2026-11-20T07:20:23.534527Z",{"slug":50,"tag_slug":50,"title_zh":51,"title_en":52,"intro_zh":53,"intro_en":54,"id":55,"is_active":35,"created_at":56,"modified_at":57},"ai-for-science","AI for Science 2026：从 UniPert 到 GPT-Rosalind 的硬核进化","AI for Science 2026: from UniPert to GPT-Rosalind","生命科学、化学材料、物理世界模型——AI 正在从\"语言工具\"变成\"实验伙伴\"。本专题收录 AI 在三大科学方向的关键节点：UniPert 统一基因与化学扰动空间、GPT-Rosalind 端到端生命科学推理、达摩院 AI 智能体 28 小时找到 4 种超导新材料、Anthropic Claude Science 把工作台做成标准品。","From language tool to lab partner — AI is reshaping life sciences, chemistry\u002Fmaterials, and physical world models. This topic covers the key milestones: UniPert unifying genetic-chemical perturbation spaces, GPT-Rosalind's end-to-end life-sciences reasoning, DAMO's AI agent discovering 4 superconducting materials in 28 hours, and Anthropic's Claude Science workbench going mainstream.","988a4300-5fab-41c4-b5d8-63711a2dc757","2026-09-10T01:34:15.296649Z","2026-09-10T01:34:15.296663Z",{"slug":59,"tag_slug":59,"title_zh":60,"title_en":61,"intro_zh":62,"intro_en":63,"id":64,"is_active":35,"created_at":65,"modified_at":66},"h3-series","MiniMax H3 系列：从开源权重到 35 倍吞吐","MiniMax H3 Series: from open weights to 35x throughput","MiniMax H3 自 2026 年 8 月开源以来节奏密集：官方把生成、参考与编辑收回一个模型；ComfyUI 当天压进 RTX 3060；摩尔线程 3 小时完成国产 GPU 适配；fal 后训练版把吞吐拉到 35 倍；FastH3 蒸馏再砍推理成本。本专题持续追踪 H3 的发布—开源—蒸馏—部署全链路。","Since MiniMax open-sourced H3 in August 2026 the pace has been relentless: one unified omni-modal model, same-day ComfyUI support down to an RTX 3060, a 3-hour Day-0 port to Moore Threads GPUs, fal's post-trained H3 Max at 35x throughput, and FastH3 distillation cutting inference cost further. This topic tracks the full H3 chain — release, open weights, distillation, deployment.","83ef0daa-3c31-4cb3-86ed-e5ee58654d5f","2026-09-08T07:33:19.942193Z","2026-09-08T07:33:19.942209Z",{"items":68},[69,74,79,84,89,94],{"id":70,"title":71,"news_slug":72,"published_at":73},"c4623b85-37f5-435e-ae98-e67eb72eb1dc","别再逐篇问大模型:AI写程序当评委,每主题0.17美元起","execjudges-llm-written-program-judges-retrieval-eval","2026-11-20T01:40:00+00:00",{"id":75,"title":76,"news_slug":77,"published_at":78},"4544ed60-b5fb-41b6-9fee-dd2f59012547","隐状态转个方向:直答46到85,超CoT省九成token","cot-hidden-state-directions-direct-answers","2026-11-19T13:30:00+00:00",{"id":80,"title":81,"news_slug":82,"published_at":83},"b602ac1b-9734-4b80-bb3c-16f19ae493a0","LoRA 的秩在虚标:配了 896,只用到 361","lora-effective-rank-capacity-utilization","2026-11-14T23:30:00+00:00",{"id":85,"title":86,"news_slug":87,"published_at":88},"6186bff0-6bbb-4970-a0bc-b7bfa93a8e49","像素空间压缩:16个token装下RAG检索到的文档","opticalrag-pixel-space-rag-compression","2026-11-13T03:20:00+00:00",{"id":90,"title":91,"news_slug":92,"published_at":93},"7984acc5-d045-4258-8da7-daf8ff25995b","一字节一token:27B模型词表砍到384的账本","qwen3-8-27b-byteify-vocab-384","2026-11-12T21:10:40+00:00",{"id":95,"title":96,"news_slug":97,"published_at":98},"79da91a4-039a-4014-9c4b-44d8dc1b5574","同一份语料两种结局:数学分从76.6崩到18.2","on-policy-mid-training-distillation-math-collapse","2026-11-12T11:40:00+00:00"]