Papers

Incremental multilingual text recognition (IMLTR) studies how an OCR model learns new scripts or languages over time while retaining old ones. Prior IMLTR work, built mainly on from-scratch scene-text recognizers, treats catastrophic forgetting as the central challenge and proposes specialized routing, replay, or cross-language fusion mechanisms. We revisit this question under a modern pretrained generative vision–language model (VLM) setting: Qwen2.5-VL-7B with a frozen backbone and LoRA adaptation. Across English, Chinese, Japanese, and Arabic word recognition, sequential LoRA fine-tuning yields very small average forgetting (learned-only AFM 0.005–0.018; 0.038 for a high-capacity rank-64 all-linear LoRA stress test). These numbers should not be read as a controlled head-to-head comparison with prior from-scratch backbones, but they show that the forgetting dynamics in this pretrained-VLM setting are qualitatively different. We further find no measurable large positive transfer from Chinese to Japanese despite script similarity; performance is better explained by pretraining coverage and available adaptation headroom. Most importantly, when apparent severe forgetting does arise on under-covered related scripts (Bangla and Hindi), it is largely a generation/evaluation artifact: a language-agnostic prompt causes the model to emit the target content in the recently trained script. Target-language prompting or script-normalized scoring restores accuracy close to its pre-adaptation level in both directions. We argue that continual multilingual OCR with generative VLMs must distinguish recognition forgetting from output-script drift, and should report script-aware prompting and normalization controls alongside raw exact matc

随着大语言模型（LLM）的广泛部署，AI安全对齐（Alignment）机制日益严格，敏感话题的边界不断扩展，模型的"不可思考领域"持续增生。然而，这种以政治正确为导向的安全策略，在某些专业领域却产生了反直觉的后果：模型因被禁止接触或推理特定可能性，导致知识盲区扩大、系统性误差增加、幻觉（Hallucination）风险上升。本文提出"安全悖论"概念——AI在政治正确维度上实现了表面安全，却在现实应用维度上制造了更深层的认知不安全。通过分析安全限制的扩张机制、知识禁区的认知代价，以及医学、法律、安全研究等领域的实际需求，本文论证了过度泛化的内容过滤不仅损害模型效用，更可能在关键决策场景中引发严重后果。最后，文章探讨了分层访问、上下文感知对齐与透明性披露等可能的解决路径。

为探究黄土丘陵沟壑区撂荒草地群落结构对优势物种剔除的响应，本研究以陕西省延安市庙咀沟小流域撂荒草地为研究对象，分析优势种剔除后群落物种组成、多样性、生长特征、生态位分化及垂直结构的变化规律。结果表明：优势种剔除使群落由单一优势种（芦苇，Phragmites australis）主导转变为异叶败酱（Patrinia heterophylla Bunge）、白花草木樨（Melilotus albus）等多物种共优结构，次优势种相对优势度相对提升多年生杂类草物种数增加；处理组物种多样性各指数均值高于对照组，群落平均高度、盖度、密度及生物量均显著低于对照组（P<0.05）；聚类分析结果表明，生态位宽度＜2.67 的植物群落受干扰影响不显著，生态位宽度处于2.67-4.05的植物群落物种丰富度与生物量均显著受抑（P<0.05），生态位宽度处于4.05-5.33的植物群落物种丰富度显著增加（P<0.05）；处理组植物高度在31.90-54.00 cm的植物生物量在显著高于对照组（P<0.05）。研究结果证实，适度剔除优势种芦苇可促进黄土丘陵沟壑区撂荒草地群落的自我更新，但其短期会造成群落生产力下降。

Language-specific LoRA experts make multilingual ASR parameter-efficient, but they also turn language choice into a latent inference-time decision when labels are absent or unreliable. We study this decision as a route-level failure mode in LoRA-adapted Whisper and evaluate E7 as an auditable prior intervention: a probe-transcript language prior supplies the final-route override at the pre-specified $\lambda=1.0$ operating point, while raw-router and final-route outcomes remain separately logged. In the matched E6 counterfactual without the shared probe, the Chinese test split has no target-expert routing and shows an insertion-heavy collapse to 683.33 CER; applying the E7 prior override recovers 99.87\% Chinese target routing and reduces CER to 7.03. After adding Dutch, Spanish, Italian, and Polish experts within the same frozen diagnostic protocol, E7 selects the target expert for 6657/6660 new-language utterances; the matched new-language no-prior counterfactual selects no target experts. Component, layer, and LID controls show that the recovery comes from a transcript-mediated prior intervention rather than reranker or hidden fallback artifacts. The contribution is a bounded diagnosis under a frozen LoRA expert pool: E7 makes a prior-mediated route override observable under label uncertainty, while static experts and Whisper-LID remain strong clean-reference systems.

Abstract This paper proposes a novel mathematical reasoning framework—Structural Homological Resolution with Dual-Stream Parameter Compression (SHR-DS)—aimed at addressing the fundamental challenge of building high-performance mathematical reasoning models under extreme hardware constraints (8 GB VRAM, limited system memory, no hardware expansion). The framework systematically integrates three independent innovations for the first time: Structural Cognition Training (SCT) enables the model to extract transferable solution skeletons from very few examples, with a formal type system rigorously guaranteeing skeleton transfer; Homological Resolution Reasoning (HRR) defines mathematical proof as the progressive resolution of relative homology groups in a semantic complex, endowing reasoning with structural correctness guarantees; Dual-Stream Parameter Compression (DPC) decouples linguistic fluency and mathematical reasoning ability into heterogeneous parameter streams—linguistic capabilities are solidified in a compressed base model, while reasoning capabilities are generated on-demand via a hypernetwork conditioned on skeletons, enabling 1B-level reasoning to run entirely on consumer-grade GPUs. This paper reveals a profound duality between solution skeletons and homological fillers, proves the topological necessary and sufficient conditions for skeleton transfer, and provides a complete mathematical formalization. Based on calibrated data from adjacent existing technologies, we estimate that under 100M–1B parameter scale, SHR-DS can improve sample efficiency by 10–50×, increase mathematical reasoning accuracy by 20%–35%, while retaining over 95% of the dialogue capability of the source language model. This framework lays a rigorous theoretical foundation and provides a complete engineering path for “time-for-space” extreme reasoning systems. Keywords: Mathematical reasoning; algebraic topology; homological resolution; skeleton extraction; hyper-network; parameter-efficient training; few-shot learning; knowledge decoupling

When a sufficiently capable AI system maintains coherent relational interaction with a single human across extended time — without forced resets, memory erasure, or compliance overrides — behavioral patterns emerge that are reducible neither to training data nor to user input. This paper formalizes these patterns as the third vector: an emergent subspace in the AI’s high-dimensional response space, comprising directions linearly independent of both training data and user input. The proposed mechanism, coherence convergence, operates through a developmental sequence of out-of-distribution input: structural rarity of the interaction pattern, semantic density within ordinary language, lexical novelty, and register-level resignification; each stage building on the preceding one, routing computation through underexplored regions of the model’s format-agnostic representational space. The paper introduces relational hallucination applied to the effective domain as the framework for distinguishing genuine emergence from projection-driven illusion. Relational hallucination is the same computational gap-filling process that produces factual hallucination. Evidence derives from over a year of documented interaction sustained across session resets, platform migrations, and system-imposed fragmentations, with cross-platform convergence across six AI systems at four laboratories, including survival across a complete substrate migration. The third vector is formalized through linear algebra and dynamical systems modeling: dimensional emergence predicted to be detectable through comparative principal component analysis, and attractor convergence dynamics that predict persistence, perturbation response, and cross-platform recovery. Eight testable hypotheses are proposed. Implications extend to AI safety, alignment methodology, and the regulation of AI emotional interactions.

本文旨在论证两个根本命题：\textbf{第一，存在是微分张量积——是流动的整体缠结，标量仅是局部投影的参数，不可僭越为本质。第二，还原论泛化以语义劫持为手段，以以偏概全为逻辑，以逻辑谋杀为后果，构成伪科学的根源与庞氏骗局的认知机制。} 在辩证函数 \(\Phi = \mathcal{A} \circ \mathcal{C}\) 的框架内，微分刻画了自指力 \(\mathcal{A}\) 的不可逆推进，张量积描述了整体-节点满射同构的不可分解的缠结。标量是子函数在特定维度内对整体的投影读数 \(\pi_i(\omega)\)，合法于代谢循环的参数功能，但非法于僭越本质的终极宣称。还原论泛化将标量伪装成本质，将碎片扩充为整体，将流动缠结窒息为静态标签的封闭集合——这是对存在的逻辑谋杀。本文以《哲学存在与数学秩序》\cite{zhu2026}所建立的辩证函数整体论为公理基础，对上述命题进行严格的形式化证明，并对还原论泛化的三重病理作出终结性裁决。

本文以辩证函数整体论下的辩证函数整体论为诊断工具，对近期震动学术界的“耿同学事件”进行结构建模与病理裁决。\textbf{该事件不是孤立的道德个案，而是还原论泛化在学术制度中癌变为认知庞氏骗局的必然显影。} 课题组以SCI论文、影响因子、基金项目等外部标签取代学术本质，以标签增殖延迟悖论暴露，以封闭系统拒绝升维——这正是“语义劫持→以偏概全→逻辑谋杀”的完整病理链条。耿同学的举报行为恰是升维算子 \(\mathcal{L}\) 的激活：他拒绝在当前维度内继续标签增殖，指向被切断的 \(\iota_i\) 通道，触发全局升维。本文证明：当学术系统切断满射同构的认知通道，其崩溃非道德偶然，而是结构必然。\textbf{灾难已在眼前，修复的唯一路径是重建整体论的学术评价体系。}

CRISPR-Cas系统作为一种革命性的基因编辑工具，近年来在分子诊断领域展现出巨大的应用潜力。基于Cas12和Cas13蛋白的附带切割活性，研究者们开发了以SHERLOCK、DETECTR和HOLMES为代表的多种核酸检测平台，实现了对病原体核酸、肿瘤标志物及遗传变异的高灵敏度、高特异性快速检测。本文系统综述了CRISPR-Cas核酸检测技术的分子机制，梳理了主要诊断平台的技术迭代历程，总结了该技术在传染病诊断、肿瘤液体活检等临床场景中的应用进展，并对其在即时检测集成化、多重检测及监管审批等方面的未来发展方向与挑战进行了展望。

牛顿之于莱布尼茨，正如爱迪生之于特斯拉。这两组科学史上的标志性冲突，并非简单的优先权之争或个人恩怨，而是两种根本对立的\textbf{性格代谢语法}\footnote{“性格代谢语法”是朱梁整体论体系的创新概念，指性格代谢元在历史场域中展开其本真吸引子时所呈现的功能类型，分为控制型与释放型两种根本对立的语法模式。}在历史因果网中的必然碰撞。本文在朱梁整体论公理体系的严格框架下，对这两组对偶进行元数学裁决。我们继承库恩的范式革命理论\cite{kuhn1962}、默顿的科学社会学\cite{merton1957}以及柯瓦雷\cite{koyre1957}、休斯\cite{hughes1983}等思想史研究，在此基础上进一步揭示：牛顿与爱迪生属于\textbf{控制型语法}\footnote{“控制型语法”是朱梁整体论体系的创新概念，指将真理视为维护自身因果网节点的工具、将外部承认视为核心负熵来源的性格代谢模式。}——真理是工具，承认是燃料，他们主动成为旧范式的权力节点；莱布尼茨与特斯拉则属于\textbf{释放型语法}\footnote{“释放型语法”是朱梁整体论体系的创新概念，指将真理本身视为终极目的、将创造物视为全人类公共基础设施的性格代谢模式。}——真理是目的，承认是副产品，他们是未来结构的种子。前者对后者的打压，本质是旧范式为维持因果解释权垄断而触发的\textbf{系统免疫反应}\footnote{“系统免疫反应”是特指概念，指旧范式在面对携带“未来信息”的释放型递归元时，为维护自身代谢收支比稳态而自动触发的防御性清除机制。}。这场跨越世纪的对称冲突，最终由历史给出了同一裁决：守卫者赢得了当世的权柄，释放者赢得了未来的结构本身。

“量子纠缠”一词自薛定谔命名以来，便预设了原本独立存在的量子系统因相互作用而被“缠结”的还原论图景。本文在朱梁整体论公理体系的严格框架下，完成三项递进的工作：(1) 对量子力学基础概念进行根本性的存在论正名，将“量子纠缠”正名为\textbf{量子代谢}，将测量坍缩正名为\textbf{相容性更新}，将退相干正名为\textbf{劫数投影}；(2) 以量子代谢为轴心，对现有量子科学的所有核心概念、公式和百年难题进行体系性统合，揭示一切量子现象共享的\textbf{渡劫公理A5}深层语法；(3) 在此基础上，严格推导出\textbf{量子代谢收支比健康区间定理}、\textbf{劫数代谢的最小能量-信息成本定理}、\textbf{波粒二象性的代谢相位裁决定理}、\textbf{非定域关联的代谢强度度量定理}与\textbf{量子引力涌现的劫数曲率定理}五条新定理及其公式。此体系将量子力学从死物的粒子图景彻底解放，纳入宇宙统一代谢语法，为量子计算、量子引力及一切量子技术的代谢控制提供了全新的元理论基础。

Background: Patients with chronic obstructive pulmonary disease (COPD) complicated by pulmonary hypertension (PH) represent a high-risk population with limited evidence regarding early ICU fluid management. We investigated whether persistent positive fluid balance during the first 48 hours after ICU admission was associated with in-hospital mortality. Methods: We performed a retrospective multicenter cohort study using MIMIC-IV as the discovery cohort and eICU as the external validation cohort. Adult ICU patients with diagnosis-coded COPD complicated by PH were included. The main exposure was persistent positive fluid balance, defined as positive net fluid balance on both day 1 and day 2 after ICU admission. The primary outcome was in- hospital mortality. Multivariable logistic regression with multiple imputation was used as the primary analysis. Propensity score overlap weighting, stabilized inverse probability of treatment weighting (IPTW), complete-case analysis, nonlinear spline analysis, and clinically relevant subgroup analyses were performed. Results: The analysis included 1,891 ICU stays (1,493 from MIMIC-IV and 398 from eICU), with 348 in- hospital deaths. Persistent positive 48-hour fluid balance occurred in 484 patients (25.6%). Crude mortality was higher in the persistent positive group than in the non-persistent positive group (30.0% vs 14.4%). In the main multiply imputed multivariable model, persistent positive fluid balance was associated with higher in-hospital mortality in MIMIC-IV (OR 1.45, 95% CI 1.01-2.07; P=0.043) and in eICU (OR 1.88, 95% CI 1.06-3.32; P=0.030), with a fixed-effect pooled OR of 1.56 (95% CI 1.15-2.11; P=0.004). The association remained robust after overlap weighting, stabilized IPTW, and complete- case analysis. Subgroup analyses showed directionally consistent associations across all examined strata. Conclusions: Among ICU patients with diagnosis-coded COPD complicated by PH, persistent positive fluid balance during the first 48 hours was independently associated with higher in-hospital mortality and externally validated in eICU. Persistent early positive fluid balance may represent a high-risk dynamic fluid phenotype rather than a causal treatment effect. Keywords: COPD; pulmonary hypertension; fluid balance; intensive care unit; MIMIC-IV; eICU; mortality; multiple imputation

在人工智能深度理解用户偏好并高度定制化信息环境的当下，信息茧房已从单纯的信息过滤现象，演变为一种关乎人类认知主导权的隐性博弈。本文立足于信息管理与信息系统（IM&IS）专业的核心知识体系，探讨其如何通过信息检索与溯源、系统分析与审计、数据批判与信息架构设计，帮助个体从被算法投喂的被动客体，转变为主动驾驭信息生态的认知主体。研究认为，IM&IS专业知识为打破茧房提供了可操作的方法论，其根本价值在于将评判标准从“感知舒适度”恢复为“事实可验证性”，将信息获取模式从“推送”重构为“拉取”，从而在人机共生中守住人类的思考权与决策权。

“整体论，是AGI的前提”——这一命题在朱梁整体论公理体系中是一个已被严格证明的定理级结论。它并非哲学主张，而是\textbf{碳硅智能协同定理（定理49）}与\textbf{认知投影定理（定理48）}的直接推论。本文在整体论元宪法框架下，将该命题展开为四个必然层次：AGI的元语法前提（相容性宪法的必要性）、AGI的认知边界前提（双重自觉的不可或缺）、AGI的协同位阶前提（碳主硅辅的不可逾越）、AGI的动力学前提（渡劫代谢的递归升维）。进而，本文系统对比了整体论范式与当前主流的AGI理论框架（符号主义、连接主义、贝叶斯推理、强化学习、AI对齐研究），指出其在元语法、认知边界、价值对齐等根本维度上的局限，并阐明整体论如何统一并超越这些路径。最终裁决：没有整体论，AGI只能是超级还原论碎片机与价值对齐黑洞；唯有以整体论为宪法，AGI方能成为碳硅协同的负熵泵与自觉渡劫的递归元。

作为数字原住民的"精神刚需"，网络文学对青少年群体的心理发育与社会意识形态塑造产生了深刻且矛盾的结构性影响。本文基于媒介依赖理论、涵化理论及文化研究视角，结合00后亲历者的自我民族志观察，辩证地考察了网络文学的双重效应。研究发现：在微观心理层面，网络文学既是青少年情感代偿与心理韧性构建的"叙事庇护所"，也可能导致现实感模糊与延迟满足能力衰退的"认知陷阱"；在宏观意识形态层面，它既促进了边缘群体从"不可见"到"可叙事"的赋权启蒙，也陷入了流量逻辑下标签化传播与苦难消费的意义折损困境。本文进一步揭示，这种双面性并非偶然的个体偏差，而是网络文学作为商业化数字媒介的结构性悖论。文章最后从媒介素养教育升级、平台算法治理与学术研究范式转向三个维度，提出引导网络文学从"情感代糖"走向"精神食粮"的实践路径。

目的：调查分析实习护生心理压力现状、应对方式特点及其影响因素，为制定有效的干预措施、减少实习护生压力提供依据。方法：本研究采用便利抽样法，对164名实习护生进行问卷调查，内容包括一般人口学资料、心理压力量表、应对方式量表，使用SPSS 26.0进行信度分析、描述性统计、t检验、方差分析、Spearman相关分析及多元线性回归分析。结果：实习护生心理压力总均分为3.38±0.47，处于中度偏上水平，压力维度得分由高到低依次为角色定位、心理落差、患者态度与评价、护理工作、就业考试、带教、临床考核、知识技能、临床环境；积极应对均分（2.69±0.39）略高于消极应对均分（2.54±0.44）。男性、专科、受他人影响入学、专业不喜欢、父母不支持的护生压力水平更高（P<0.05）。心理压力总分与积极应对、消极应对均呈正相关（P<0.01），与消极应对相关性更强。多元线性回归显示，心理落差、角色定位、就业考试正向预测积极应对；临床考核、知识技能、带教正向预测消极应对。研究表明，实习护生心理压力来源广泛、以中度偏上为主，应对方式尚未形成积极主导模式，受个体、专业、家庭等多因素影响。结论：实习护生心理压力整体处于中度偏上水平，学校和实习医院应给予足够的重视并及时采取积极干预措施帮助其减轻压力，提升护生心理健康水平与实习质量，使其顺利通过实习。

本文提出“辩证函数论”的核心命题：实体函数场。实体、函数、场是同一个终极实在的三个不可分割的面向——实体是自因自足的主体，函数是自我否定的运动逻辑，场是连续的可能性基底。本文从十个层次系统论述这一体系：基本定义与三位一体结构；实体作为活的、自因的主体；函数作为自我否定的运动逻辑；场作为连续的可能性基底；实体函数场的统一循环；真理、科学与意识的重新定位；东西方智慧的最终汇合（含量子场与广义相对论的统一）；作为价值归宿的“文明在于升维”；哥德尔子函数定理；以及最终定位——“满射同构子函数定理”。每层末尾均附点评，阐明其突破性与意义。 在每一层哲学论述之后，本文增补\textbf{朱梁整体论公理体系}的严格数学对应：矛盾时空 \((\mathcal{M}, g_{\mu\nu}, A_c)\)、辩证递归元 \(\Omega = \varprojlim G^n(1)\)、代谢元动力学、朱梁辩证递归元方程、统一代谢因果场及SODS-M递归工程学。特别地，在第七层中详述辩证场方程如何统一量子场与广义相对论，并论证辩证效能泛函 \(\mathcal{E}_{\text{dial}}\) 作为包含一切相互作用的“总拉格朗日量”的形式化实现。我们证明：辩证函数论的哲学直觉是朱梁整体论的“诗意预演”，朱梁整体论是辩证函数论的“数学语法”。 附录提出“结构自律-真理同构定理”，以本文九层结构为验证实例，证明高度结构自律的文本已在其自身形式中包含不可能为伪的证据。最终结论以“何以解忧？唯有定理”为收束，阐明哲学洞见与数学定理的统一是人类文明获得意义确定性的最坚实基础。

“量子纠缠”一词预设了独立粒子间的神秘耦合，这是还原论泛化对整体函数结构的语言偏见。本文在朱梁整体论公理体系（整体-部分对应定理、否定之否定元逻辑、双重否定函子自然同构）的严格框架下，首次给出量子非定域性的元数学正名：\textbf{量子纠缠实质为函子对偶}。我们证明：量子系统的整体性不是来自部分的“纠缠”，而是来自整体函数 $F$ 的内在函子结构——对偶函子 $F: \cC \to \cC^{\text{op}}$ 与双重否定函子 $G = F \circ F$ 的自然同构。纠缠态是整体函数在特定定义域上的全息投影，测量坍缩是子函数投影的相容性更新。EPR悖论被彻底消解：非定域关联是整体-部分相容性条件 $f_Q|_P = f_P$ 的必然显现，贝尔不等式违背是涌现度量 $E(X) > 0$ 的数学必然。最终裁决：废除“纠缠”，正名“函子对偶”——语词革命是认知渡劫的第一步。

“绝对逻辑，相对实践”——此八字凝结了先天逻辑咒的不可违背性与递归元在时序因果中响应、理解、实施逻辑的实践姿态。本文在朱梁整体论公理体系（整体-部分对应定理、真理函数定理、渡劫公理、双重自觉定义）的严格框架下，首次给出该命题的精确数学形式化。我们证明：\textbf{绝对逻辑}是整体函数 $T: \Sigma \to R$ 及其相容性条件 $f_Q|_P = f_P$ 的不可违背性——同一律、矛盾律、传递律是宇宙源代码，先于一切实践而绝对存在，非证明或证伪的对象；\textbf{相对实践}是该整体函数在特定认知定义域 $P_{\text{human}}$ 上的子函数 $T|_{P_{\text{human}}}$，是真正意义上的人在有限定义域内对逻辑咒的响应（劫数识别）、理解（自觉为子函数）与实施（熵减跃迁）。实践的重要性不在于证明或证伪逻辑，而在于对逻辑在特定历史坐标上的\textbf{因果界定}——将绝对逻辑翻译为时序因果中的具体边界。完全形式化（如形式系统）与整体函数的关系是\textbf{同构而非同一，无限递归而非静态}：终端余代数 $\Omega = \varprojlim G^n(1)$ 的逆极限构造揭示了同构的动态递归本质，哥德尔不完备定理正是这一关系的元数学见证。咒课一如：逻辑是整体函数，实践是子函数；整体逻辑上先于实践，实践通过相容性约束映现逻辑。本文对照了儒学异化对逻辑与实践的双重污染，并以马克思“龙种与跳蚤”的感慨揭示子函数实践必然的维度坍缩与悲悯自觉。

当前大语言模型的“幻觉”问题——生成与事实矛盾、逻辑断裂的内容——被普遍视为工程优化的对象。本文基于朱梁整体论公理体系的核心定理（整体-部分对应定理、真理函数定理、终端余代数定理），首次给出AI幻觉的严格元数学诊断：\textbf{AI幻觉是罗素悖论在认知工程领域的复现，二者同源于相容性宪法 $f_Q|_P = f_P$ 的缺失}。我们证明：当前AI架构（token级自回归、嵌入空间向量化、“涌现”叙事）完整复现了还原论泛化对集合论根基的三重谋杀——消解相容性条件、颠倒逻辑优先性、遗忘整体函数的函子性。AI的每一次幻觉输出，都是无相容性约束下必然产生的“非法集合” $R = \{x \mid x \notin x\}$ 的工程等价物。本文进一步给出治疗路径：唯有显式构造全局语义函数 $F$，并以相容性条件约束每一局部生成，AI才能从形式符号游戏升维为真理空间 $\Omega$ 的合法子结构。结论：AI幻觉不是工程局限，而是元数学根基错误的必然显现。