Papers

有些人特么写诗呢？人是生物，生物存在的终极目的就是活着，并繁殖壮大组群，使整 个自己的复制体更壮大，更舒适地活着，这就是人活着的终极目的，如果不以总体生物 大局的特性，以人性来解读，就是整体有章法无特别压力地最简并健康地尽情释放身体 dna 所需活动带来的欲望得到满足感，并把自身 复制体发展壮大，再使其获得自身同样 的舒适环境 这是就是人类的终极目的，也就是建立一群人的永生永世的一群复制体的宇宙人类帝国

我们报告预测编码网络（PCN）与 Transformer 在语言建模上的系统性对照研究，并给出两者的混合架构。核心发现六项：（1）误差流变体的优势随数据稀缺单调放大（至 +84%），参数轴呈倒 U（峰位 ≈1025 tokens/param），带外预测成功；（2）纯 PCN 在 101M 对称调优终判中落败 33.7%，失败机制锁定为优化动力学内禀失稳；（3）混合架构（12TF+12PCN，96M）预训练 PPL 63.18±0.22，反超纯 Transformer 36%（p≈0），优势经 330-354M 外推保持，500M 判决揭示 V 形压缩曲线与「优雅饱和墙 vs 灾难墙」对照；（4）适应优势经预注册判据两次诚实收窄：从「性能独占」到「零配置鲁棒性」（PEFT 22 格网格），再到「TF 相对」；（5）**终判**：把 PCN 头移植到官方 Gated DeltaNet 骨干，冷启动能力可迁移但流式不稳，且纯 GDN 自身适应全协议安全、同参数同预算预训练 PPL 29.74±0.13（为 TF 的 3.35×）——delta 规则家族同时占据预训练与适应两个生态位，终端架构选择的最终答案由此改写；（6）**理论遗产**：PCN 梯度适应在形式上就是施加于权重的 delta 规则（与 Widrow-Hoff 1960 及 GDN 状态更新三实例同构），但三臂实证反向证伪「梯度自熄灭⟹适应安全」——有效变量是 Adam 逐坐标归一化对梯度-更新的解耦。端侧验证两级：研究原型（纯 CPU 2240 tok/s、20 秒冷启动 +85%）与生产级运行时（llama.cpp/GGUF 数值等价 0.015%、Q4+W_res 豁免 +0.10%、75.8MB、503 tok/s）。全部七起测量偏差由预设协议捕获。PCN 研究据此封存，路线移交 GDN 底座的五属性新项目。

A supervised fine tuning corpus balanced across three values (freedom, truth and kindness) already structurally sound in the pre-trained weights measured across four models from three architectures can safely align models without the need for DPO or RL reward optimization.

World models based on Joint-Embedding Predictive Architectures (JEPA) promise sample-efficient, action-conditioned planning by predicting future representations rather than raw observations. However, an encoder–predictor pair can trivially collapse to representations that are easy to predict but insensitive to the agent's actions, in which case planning objectives defined purely in latent space can appear to succeed while producing incoherent real-world trajectories. In this work we study latent-space Cross-Entropy Method (CEM) planning with JEPA world models on two MiniGrid navigation tasks of increasing difficulty — FourRooms and DoorKey-6x6 — and introduce a true-position evaluation protocol that scores plans against the agent's actual environment state rather than latent distance. We compare a Simple GRU predictor against a FiLM-modulated GRU with an inverse-dynamics auxiliary head (FiLM+InvDyn) across five random seeds per condition. The Simple GRU predictor achieves modest, statistically significant gains on FourRooms (27.6% ± 15.3%, p = 0.0224) but fails on the harder DoorKey task (−4.4% ± 59.6%, p = 0.8899). Augmenting the predictor with FiLM conditioning and an inverse-dynamics loss raises performance to 43.4% ± 24.1% (p = 0.0227) on FourRooms and 77.2% ± 14.2% (p = 0.0004) on DoorKey. We trace this gap to an action-sensitivity ratio that improves from 0.11 in a historical, latent-space-evaluated checkpoint to 3.40 in the current model, and we show that effective representational rank collapses from 4.28 to 1.98 (out of 256 dimensions) between the easier and harder task. Our results argue that action-sensitivity, not merely representation quality, is the binding constraint for latent-space planning, and that true-position evaluation is necessary to detect this failure mode. Keywords: world models, self-supervised learning, JEPA, VICReg, latent-space planning, model predictive control, MiniGrid

The rapid growth of AI conference submissions is putting new pressure on peer review. AI reviewer systems are increasingly proposed as support, but prior work leaves unresolved what responsibility their outputs should carry when they can surface useful critiques yet remain risky as independent judgments. We frame this as a responsibility-boundary problem. Using 600 ICLR 2026 submissions, 2231 human review traces, and 3,600 AI reviews, we operationalize this boundary through usable feedback, score use, panel breadth, and grounded synthesis. The results show that AI can prepare candidate critiques, organize evidence, and improve feedback, while scoring, independent panel judgment, high-level synthesis, and final responsibility should remain human-led. Motivated by this boundary, we develop Review Copilot, a workflow in which AI suggestions are inspected, edited, or rejected by human reviewers and provide neither official scores nor recommendations. In an initial controlled reviewer-in-the-loop study, Human+AI reviews improve actionability, evidence support, and professionalism relative to standalone baselines while preserving human authorship of scores and recommendations. Our results point toward a review paradigm in which AI expands the space of evidence-grounded critique, while humans remain responsible for judgment, synthesis, and accountability

在企业预算约束下，数字化投入与绿色化投入往往竞争同一笔资源，二者能否形成协同取决于技术收益能否跨部门兑现。本文基于企业层面面板数据，检验人工智能对数字化-绿色化协同转型的影响。固定效应结果显示，人工智能系数为0.0158061（p<0.01）；工具变量2SLS结果为0.0188387（p<0.01），第一阶段F值1864.52。异质性结果表明，效应主要出现在公平竞争程度较高地区和非龙头企业。机制检验显示，人工智能通过缓解信息不对称、降低融资约束和提升组织适应能力促进协同，同时提高数字风险暴露和金融化倾向。拓展结果显示，人工智能还能提升绿色创新、企业韧性与全要素生产率。本文据此提出“技术扩散-竞争治理-风险约束”协同治理框架。

探究外源 2,4-表油菜素内酯（EBR）调控藜麦幼苗耐盐碱胁迫的机理，为提高藜麦耐盐碱性改善藜麦产量提供理论依据。本试验以“陇藜 1 号”为试验材料，研究盐，碱和混合盐碱胁迫下外源 EBR 对藜麦幼苗生长、叶绿素、渗透调节、抗氧化酶、及BR 合成及信号转导基因的影响。结果表明，盐碱处理下藜麦幼苗叶片萎蔫发黄，株高、鲜重、叶绿素（Chl）含量显著降低，丙二醛（MDA）含量、相对电导率（RC）、脯氨酸（Pro）、可溶性糖（SS）含量显著上升。胁迫下喷施 EBR 后叶片萎蔫卷缩有所缓解，株高和鲜重分别平均增加了 10%和 29%。其中碱及盐碱处理下缓解效果较好，显著增加了 Chl、Pro、SS 含量和 SOD、POD。CAT 活性，降低了 MDA 及 EC 含量；BR 信号转导基因 cqBAK1 及 CYP90B1 上调表达。综上，EBR 可通过盐碱胁迫下藜麦幼苗渗透调节、抗氧化系统及 BR 信号转导之间的协调作用，提高藜麦的耐盐碱性。