One of the apostolic fields in which the Society of Jesus was involved since its foundation in 1540 was the missiones ad gentes [missions among non-Christians], which produced a constant flow of European missionaries to other continents. Specifically, the Jesuit provinces of Spain sent many missionaries beyond their borders, creating administrative units that were initially dependent on the metropolis and later became autonomous Jesuit territories. There are many partial studies of many of the realities related to the Jesuit missions; We now intend to take a brief historical overview to illustrate this centrifugal trend in Spain within the Jesuit sphere, both in the old Society (before its suppression by Pope Clement XIV in 1773) and in the contemporary one (since its restoration by Pope Pius VII in 1814). To this end, we will briefly review demographic and geographical data, which provide overall figures and territorial configurations throughout the history of the Society of Jesus.
Video world models have shown immense potential in simulating the physical world, yet existing memory mechanisms primarily treat environments as static canvases. When dynamic subjects hide out of sight and later re-emerge, current methods often struggle, leading to frozen, distorted, or vanishing subjects. To address this, we introduce Hybrid Memory, a novel paradigm requiring models to simultaneously act as precise archivists for static backgrounds and vigilant trackers for dynamic subjects, ensuring motion continuity during out-of-view intervals. To facilitate research in this direction, we construct HM-World, the first large-scale video dataset dedicated to hybrid memory. It features 59K high-fidelity clips with decoupled camera and subject trajectories, encompassing 17 diverse scenes, 49 distinct subjects, and meticulously designed exit-entry events to rigorously evaluate hybrid coherence. Furthermore, we propose HyDRA, a specialized memory architecture that compresses memory into tokens and utilizes a spatiotemporal relevance-driven retrieval mechanism. By selectively attending to relevant motion cues, HyDRA effectively preserves the identity and motion of hidden subjects. Extensive experiments on HM-World demonstrate that our method significantly outperforms state-of-the-art approaches in both dynamic subject consistency and overall generation quality. Code is publicly available at https://github.com/H-EmbodVis/HyDRA.
Generative sequence models are typically trained on sample sequences from natural or formal languages. It is a crucial question whether -- or to what extent -- sample-based training is able to capture the true structure of these languages, often referred to as the ``world model''. Theoretical results indicate that we can hope for soundness at best, that is, generating valid sequences, but not necessarily all of them. However, it is still important to have practical tools that are able to verify whether a given sequence model is sound. In this study, we focus on chess, as it is a domain that provides enough complexity while having a simple rule-based world model. We propose adversarial sequence generation for verifying the soundness of the sequence model. Our adversaries generate valid sequences so as to force the sequence model to generate an invalid next move prediction. Apart from the falsification of soundness, this method is also suitable for a more fine-grained analysis of the failure modes and the effects of different choices during training. To demonstrate this, we propose a number of methods for adversarial sequence generation and evaluate the approach on a large set of chess models. We train models on random as well as high-quality chess games, using several training recipes. We find that none of the models are sound, but some training techniques and dataset choices are able to improve soundness remarkably. We also investigate the potential application of board state probes in both our training and attack methods. Our findings indicate that the extracted board states have no causal role in next token prediction in most of the models.
Fleur B Nijdam, Marieke A. J. Hof, H. Blokzijl
et al.
Background/Objectives: Pharmacogenomics (PGx) has revolutionized personalized medicine, notably by predicting drug responses through the study of the metabolic genotype of drug-metabolizing enzymes. However, these genotypes rely heavily on the availability and completeness of drug metabolism information and do not account for (all) “phenoconversion” factors, like drug–drug interactions and comorbidities. To address these limitations, a more phenotypic approach would be desirable, for which pharmacometabolomics (PMx) could be useful by studying and elucidating drug metabolism in patient samples, such as blood and urine. Methods: This study explored the potential of PMx to analyze real-world drug metabolite profiles of the extensively studied drug cyclosporine (CsA) using 24-h urine samples from 732 kidney and 350 liver transplant recipients included in the TransplantLines Biobank and Cohort Study (NCT identifier NCT03272841). Detected metabolites were matched with existing information on CsA metabolism gathered through a comprehensive literature review, aiming to confirm previously reported metabolites and identify potentially unreported ones. Results: Our analyses confirmed the urinary presence of CsA and six known metabolites. Additionally, we detected three known metabolites not previously reported in urine and identified one unreported metabolite, potentially suggesting the involvement of glutathione conjugation. Lastly, the observed metabolic patterns showed no notable differences between kidney and liver transplant recipients. Conclusions: Our findings demonstrate the potential of PMx to enhance the understanding of drug metabolism, even for well-studied compounds such as CsA. Moreover, this study highlights the value of PMx in real-world drug metabolism research and its potential to complement PGx in advancing personalized medicine.
Simple Summary Patients with sarcoma—the rare cancers of soft tissue and bone—do not all have the same clinical course once the tumor is removed. By following nearly 300 adults in a nationwide, real-time database, we discovered that metastases (new tumors appearing elsewhere) differed when they developed, where they landed, and how many appeared. About one-third of patients already had—or quickly developed—metastases within six months, while others stayed free of spread for many years. Many tumors went only to the lungs and stayed there, but some jumped rapidly to the bones, liver, or lymph nodes. Roughly half the patients had only a few small spots that could be treated with surgery or focused radiation; the rest had many lesions in several organs and needed immediate drug therapy. These patterns suggest that everyone should not get the same scan schedule or the same first-line treatment. Tailoring follow-up and therapy to a tumor’s timing, destination, and tumor-load could improve care while sparing low-risk patients unnecessary tests.
The article is devoted to the problem of studying the activities of revolutionary and ideological virtuosi of the XX century by means of the historical sociology of modernity. It attempts both to identify the main trends in the participation of intellectually active strata of modern societies in the radical transformation of social reality, and to clarify the reasons that pushed them to do so. It is noted that the ideological background of such forms of socio-political radicalism were the symbolic means of challenging the old forms of social reality and constructing new forms through secular ideologies of various types (socialism, communism, anarcho-syndicalism, fascism, national socialism, etc.). The role of conductors of the ideological message of these secular religions of salvation to the masses was usually performed by radical-minded intellectuals and students, for whom the worship of new idols and symbols of faith became the true meaning of life and struggle. In the XX century, it was the intellectuals who acted as the main creators and operators of collective salvation religions, and the social communities created on their basis. Moreover, they not only tried to save themselves with their help, but at the same time laid claim to the legitimate monopoly of determining who would be saved, and who would not. Thus, radical intellectuals acted as both the creators of the new norm and as agents of the policy of rationalization. The article pays special attention to the characteristics of the creators and social carriers of radical ideological views — and the worldviews based on them — as well as the refraction of these utopian visions in the socio-political practice of the last century. In conclusion, it is suggested that Max Weber’s historical sociology and Emile Durkheim’s symbolic theory of the “sacred”, supplemented by modern developments in the field of political sociology and socio-historical anthropology, can serve as a relevant system of theoretical coordinates for the study of ideological and revolutionary virtuosi of the XX century.
Rakhymzhan Rashimbetov, Nauryzbay Utpinov, Gilimbek Mazhiyev
et al.
The purpose of the study was to analyse the mechanisms of regulation of Islamic education in different countries and their influence on the religious orientation of students at Islamic schools (madrasas). The study examined internal and external motivational factors determining the choice of religious education in the context of state policy, as well as the key conceptual approaches explaining the transformation of the religious identity of students in Islamic educational institutions. The study was based on a systematic analysis of scientific publications, a comparative analysis of educational strategies in different countries, and a content analysis of sources reflecting the specifics of the interaction between the secular state and religious institutions. The findings of the study revealed that Islamic education is regulated according to three principal models: strict state control (France, China), partial autonomy (Turkey, Kazakhstan), and complete independence of religious educational institutions (Pakistan, Saudi Arabia). In countries with strict state control, religious education gradually lost popularity among young people, while in societies with partial autonomy, religious schools adapted to modern educational standards while preserving traditional elements of teaching. In states with full autonomy, madrasa graduates faced limited opportunities for professional adaptation, which affected their integration into the secular environment. The analysis revealed a major influence of social and economic factors on the choice of Islamic education, including family traditions, social stereotypes, funding levels, and state support. Existing theoretical models (the theory of secularisation, modernisation of education, and the confessional approach) do not provide an exhaustive explanation for the transformation of religious education. In this regard, the study substantiated the need to develop hybrid theoretical approaches that accommodate the political, economic, and cultural aspects of educational policy. The practical significance of the study lies in the possibility of using its findings in the development of educational reforms aimed at balancing secular and religious standards in the educational systems of different countries.
Introduction. The appeal to Marxism as a system of ideas explaining and predicting the ways of development of mankind and the fate of the state is still relevant. Marxism is popular not only in postsocialist countries, but also in the consistently developed bourgeois world. It is not overlooked as an object of research by representatives of leftist intellectuals at leading foreign and Russian universities. The purpose of this paper is to identify or clarify the reasons for the discrepancy between the social class ideas of Marxism and the practice of the modern state, the prospects for the theoretical and practical development of this, perhaps, the most influential doctrine after the world religions. Methodology and materials. Within the framework of a short article, which does not specifically touch upon the problems of foundations (sources), political economy, the concept of man, the interpretation of Marxism in the countries that today call themselves socialist, and many others, the author uses the methodological technique of interpreting Marx’s key theses concerning the driving forces of the historical development of society and the fate of the state through the comparison of the thinker’s ideas (insights) with contemporary reality. Research results and their discussion. The author reveals inconsistencies between the social class theory of Marxism and its practical implementation in the Soviet socialist state of the twentieth century; discovers the shortcomings of this theory in explaining the European society and state of the twentieth and early twenty-first centuries; reveals its potential in modern state-organized society. Conclusions. The study argues the point of view according to which the transformation of the social class theory of Marxism into a discourse devoid of practical possibilities of overthrowing the state (as an institution) has led to the paradox of defense and improvement of the system of institutions of the modern state by the supporters of the once revolutionary doctrine.
The article considers history, prerequisites and consequences of the spread of New Age ideas and practices in contemporary society as a new form of spirituality that contributes to the emergence and spread of new religious movements (NRMs). The authors note that in the post-secular society, religious consciousness, (para) religious ideas and practices have become significant factors in social processes. While traditional religions lose their significance, there is an increasing influence on everyday consciousness of various esoteric and occult ideas and practices due to the impact of New Age spirituality on contemporary culture. The origins of such processes can be traced back to the second half of the 19th century - fashion for occultism and esotericism among the European and American intelligentsia and emergence of international religious movements of a new type (Baha’is in Iran, Mormons and Jehovah’s Witnesses in the USA, etc.). These forms of spirituality received a new impetus in the second half of the 1960s-1970s under the countercultural revolution which gave rise to mass fascination with “alternative spirituality” among students. In Western sociology, these processes were called “occult revival”. The authors consider features of nontraditional spirituality in Russia - from the emergence of such fashion among the intelligentsia in the 1970s to the mass wave of esotericism that swept over the Russian society in the 1990s. When speaking about the current situation, the authors emphasize the role of the Internet in the spread and formation of new religious movements, communities and ideologies and note the network nature and the lack of clear institutional boundaries in (para) religious communities. The article presents grounds for classifying NRMs and explains their influence on various spheres of culture and mass consciousness, which is the growing de-rationalization of everyday consciousness as eclectically combining elements of scientific, religious and occult pictures of the world.
The Driving World Model (DWM), which focuses on predicting scene evolution during the driving process, has emerged as a promising paradigm in the pursuit of autonomous driving (AD). DWMs enable AD systems to better perceive, understand, and interact with dynamic driving environments. In this survey, we provide a comprehensive overview of the latest progress in DWM. First, we review the DWM ecosystem, which is constructed using mainstream simulators, high-impact datasets, and various metrics that evaluate DWMs across multiple dimensions. We then categorize existing approaches based on the modalities of the predicted scenes, including video, point cloud, occupancy, latent feature, and traffic map, and summarize their specific applications in AD research. In addition, the performance of representative approaches across generating and driving tasks is presented. Finally, we discuss the potential limitations of current research and propose future directions. This survey provides valuable insights into the development and application of DWM, fostering its broader adoption in AD. The relevant papers are collected at https://github.com/LMD0311/Awesome-World-Model.
Romain Harang, Jason Naradowsky, Yaswitha Gujju
et al.
Large Language Models (LLMs) exhibit emergent capabilities in structured domains, suggesting they may implicitly internalize high-fidelity representations of world models. While probing techniques have shown promising signs of this in scientific and game-based settings, they rely on model-specific internal activations, which limit interpretability and generalizability. In this work, we propose a model-agnostic, state-based evaluation framework using chess as a benchmark to assess whether LLMs preserve the semantics of structured environments. Our method analyzes the downstream legal move distributions (state affordances) to estimate semantic fidelity between predicted and actual game states. This approach offers a more meaningful evaluation than conventional string-based metrics by aligning more closely with the strategic and rule-governed nature of chess. Experimental results demonstrate that our metrics capture deficiencies in state-tracking, highlighting limitations of LLMs in maintaining coherent internal models over long sequences. Our framework provides a robust tool for evaluating structured reasoning in LLMs without requiring internal model access, and generalizes to a wide class of symbolic environments.
We introduce World In Your Hands (WIYH), a large-scale open-source ecosystem comprising over 1,000 hours of human manipulation data collected in-the-wild with millimeter-scale motion accuracy. Specifically, WIYH includes (1) the Oracle Suite, a wearable data collection kit with an auto-labeling pipeline for accurate motion capture; (2) the WIYH Dataset, featuring over 1,000 hours of multimodal manipulation data across hundreds of skills in diverse real-world scenarios; and (3) extensive annotations and benchmarks supporting tasks from perception to action. Furthermore, experiments based on the WIYH ecosystem show that integrating WIYH's human-centric data improves robotic manipulation success rates from 8% to 60% in cluttered scenes. World In Your Hands provides a foundation for advancing human-centric data collection and cross-embodiment policy learning. All data and hardware design will be open-source.
Recent advances in agent development have focused on scaling model size and raw interaction data, mirroring successes in large language models. However, for complex, long-horizon multi-agent tasks such as robotic soccer, this end-to-end approach often fails due to intractable exploration spaces and sparse rewards. We propose that an effective world model for decision-making must model the world's physics and also its task semantics. A systematic review of 2024 research in low-resource multi-agent soccer reveals a clear trend towards integrating symbolic and hierarchical methods, such as Hierarchical Task Networks (HTNs) and Bayesian Strategy Networks (BSNs), with multi-agent reinforcement learning (MARL). These methods decompose complex goals into manageable subgoals, creating an intrinsic curriculum that shapes agent learning. We formalize this trend into a framework for Hierarchical Task Environments (HTEs), which are essential for bridging the gap between simple, reactive behaviors and sophisticated, strategic team play. Our framework incorporates the use of Large Language Models (LLMs) as generative world models of tasks, capable of dynamically generating this scaffolding. We argue that HTEs provide a mechanism to guide exploration, generate meaningful learning signals, and train agents to internalize hierarchical structure, enabling the development of more capable and general-purpose agents with greater sample efficiency than purely end-to-end approaches.
This paper presents WorldPlay, a streaming video diffusion model that enables real-time, interactive world modeling with long-term geometric consistency, resolving the trade-off between speed and memory that limits current methods. WorldPlay draws power from three key innovations. 1) We use a Dual Action Representation to enable robust action control in response to the user's keyboard and mouse inputs. 2) To enforce long-term consistency, our Reconstituted Context Memory dynamically rebuilds context from past frames and uses temporal reframing to keep geometrically important but long-past frames accessible, effectively alleviating memory attenuation. 3) We also propose Context Forcing, a novel distillation method designed for memory-aware model. Aligning memory context between the teacher and student preserves the student's capacity to use long-range information, enabling real-time speeds while preventing error drift. Taken together, WorldPlay generates long-horizon streaming 720p video at 24 FPS with superior consistency, comparing favorably with existing techniques and showing strong generalization across diverse scenes. Project page and online demo can be found: https://3d-models.hunyuan.tencent.com/world/ and https://3d.hunyuan.tencent.com/sceneTo3D.
Remo Sasso, Michelangelo Conserva, Dominik Jeurissen
et al.
While reinforcement learning from scratch has shown impressive results in solving sequential decision-making tasks with efficient simulators, real-world applications with expensive interactions require more sample-efficient agents. Foundation models (FMs) are natural candidates to improve sample efficiency as they possess broad knowledge and reasoning capabilities, but it is yet unclear how to effectively integrate them into the reinforcement learning framework. In this paper, we anticipate and, most importantly, evaluate two promising strategies. First, we consider the use of foundation world models (FWMs) that exploit the prior knowledge of FMs to enable training and evaluating agents with simulated interactions. Second, we consider the use of foundation agents (FAs) that exploit the reasoning capabilities of FMs for decision-making. We evaluate both approaches empirically in a family of grid-world environments that are suitable for the current generation of large language models (LLMs). Our results suggest that improvements in LLMs already translate into better FWMs and FAs; that FAs based on current LLMs can already provide excellent policies for sufficiently simple environments; and that the coupling of FWMs and reinforcement learning agents is highly promising for more complex settings with partial observability and stochastic elements.
The combination of sofosbuvir/velpatasvir/voxilaprevir (SOF/VEL/VOX) is recommended as a salvage therapy for treatment-experienced chronic hepatitis C (CHC) patients. However, it is used in our country for treatment-naïve and treatment-experienced patients. This study aims to present real-world data from Türkiye on CHC patients treated with SOF/VEL/VOX. The present study was conducted by the Viral Hepatitis Study Group of the Turkish Society of Clinical Microbiology and Infectious Diseases (KLİMİK). It was a multicenter, retrospective, observational study. The data were collected from patients receiving SOF/VEL/VOX therapy at 12 medical centers in Türkiye between 1 June 2022 and 31 December 2024. The patients had received the treatment for 8 to 12 weeks. Of the 139 patients enrolled, 63.3% (n = 88) were male, with a mean age of 54.4 years. Most patients were non-cirrhotic (94.2%, n = 131) and treatment-naïve (92%, n = 128); 49.6% (n = 69) were infected with genotype 1b. Early virologic response (EVR) could be assessed in 126 patients, with an EVR rate of 82.5% (n = 104). End-of-treatment data were available for 113 patients, all achieving an end-of-treatment response. Among the 80 patients for whom week-12 post-treatment data were available, 97.5% sustained virologic response at week 12 (SVR12). Significant improvements were observed in AST, ALT, and platelet levels, along with reductions in APRI and FIB-4 scores (p = 0.001).” No serious adverse events leading to treatment discontinuation were reported. Mild adverse events included pruritus (2.1%, n = 3), fatigue (2.1%, n = 3), and nausea (1.4%, n = 2). The SOF/VEL/VOX combination is a highly effective and well-tolerated treatment option in treatment-naïve CHC patients, achieving an SVR12 rate of 97.5%.
Introduction Achieving and maintaining optimal glycemic control from the onset of type 1 diabetes (T1D) is crucial in pediatric care, especially in early childhood when the developing brain is highly vulnerable to both hypo- and hyperglycemia [1-3]. Hyperglycemia during early childhood increases the risk of long-term vascular complications, while severe hypoglycemia may impair neurocognitive development, causes family anxiety, and complicates social integration [4-5]. Although automated insulin delivery (AID) systems have demonstrated efficacy in controlled trials, real-world evidence in children under six years of age, particularly involving off-label use, remains limited. The Control-IQ (CIQ) algorithm, integrated into the Tandem t:slim X2 insulin pump, has shown benefits in adolescents and school-aged children [6-10]. However, few studies have evaluated its long-term use in children under six in routine clinical practice. Objective This study aimed to compare the real-world effectiveness and safety of the CIQ system in two pediatric age groups—children aged 0.5–5 years and children aged 6–10 years—over an 18-month follow-up period. We evaluated effectiveness in terms of glycemic control (% of time in glucose range 70–180 mg/dL [TIR], % of time in glucose range 70–140 mg/dL [TITR], and HbA1c) and safety in term of adverse events (diabetic ketoacidosis [DKA], hyperglycemia and severe hypoglycemia). Methods This prospective, multicenter observational study used retrospective data from 32 Italian pediatric diabetes centers. Eligible participants had T1D diagnosed ≥ 6 months, , were 10 years at CIQ start were excluded. Participants were stratified by age at CIQ initiation (0.5–5 and 6–10 years). At CIQ initiation (baseline) sex, presence of celiac disease or thyroiditis and parents’ age, nationality and education, were collected. HbA1c, BMI z-score, CGM-derived data (TIR, TITR, % of time spent in glucose ranges: 250 mg/dL, Glucose Monitoring Indicators and coefficient of variation of glucose), Glycemia, Standard Deviation of Glycemia [SD] and DKA episodes were assessed at baseline, 6, 12, and 18 months. Descriptive statistics were used for baseline comparisons. Chi-square or t tests evaluated group differences. Trend over time points in TIR, TITR, and HbA1c were analysed using mixed-effects models for repeated measures, adjusted by age group, sex, time from diagnosis to CIQ initiation, DKA at onset and parents’ socio-economic characteristics (at least one non-Italian parent, parents’ education). A sequential difference contrast was used to model time; interaction between time and age groups was evaluated. Only children with complete data on the outcomes at all four time points were included in these models. Safety outcomes included the proportions of DKA and severe hypoglycaemias occurring during 18-month follow-up. Results Of the 334 children enrolled, 253 (106 aged 0.5–5; 147 aged 6–10) had complete data on the outcomes and were included in longitudinal analyses. At T1D diagnosis, a higher prevalence of thyroiditis in the older group was found, and no significant sociodemographic differences. At CIQ initiation, younger children had a significantly shorter time from diagnosis to CIQ initiation (1.36 vs 2.61 years, p<0.001), higher HbA1c (8.3%% vs 7.7%, p=0.020) and higher glycaemic variability (SD 63.3 mg/dL vs 58.3 mg/dL, p = 0.023) while TIR, TITR, and the other CGM-derived data were comparable. Longitudinal analysis (Figure 1) showed significant improvement in both groups 6 months after CIQ initiation: TIR increased by 6.62% (95% CI: 4.89–8.36) and TITR by 5.63% (95% CI: 3.61–7.66), corresponding to over 80 additional minutes/day spent in target ranges. These improvements were sustained at 12 and 18 months. HbA1c decreased by an average of 0.82% (95% CI: –1.01 to –0.62) in the first 6 months, remaining stable thereafter. No significant interaction between time and age groups was observed, indicating similar trends in both cohorts. Having at least one non-Italian parent was significantly associated with lower TIR (-5.82%, 95% CI: -10.33 to -1.31) and higher HbA1c levels (0.31%, 95% CI: 0.01 to 0.63). A high parental education level (university or higher vs. up to lower secondary education) was associated with higher TIR (8.61%, 95% CI: 3.03–14.18) and lower HbA1c levels (−0.42%, 95% CI: −0.78 to −0.06). Age at CIQ initiation, time from diagnosis to CIQ initiation, DKA at diagnosis, and sex were not significant predictor. Regarding safety, no severe hypoglycaemia episodes were reported in the younger group, and only one occurred in the older group after 12 months. A single DKA episode was recorded in a child under six. Moreover, CGM-derived data indicated that time spent in hypoglycaemia (<54 and 54–69 mg/dL) remained consistently below clinically relevant thresholds (<1% and <3%, respectively). Conclusion In this large real-world cohort of young children with T1D, the CIQ system demonstrated consistent and sustained improvements in glycaemic outcomes over 18 months, with minimal adverse events. Significant gains in TIR, TITR, and HbA1c were observed in both age groups, particularly in the first 6 months after CIQ initiation. These benefits were maintained long-term, regardless of initial glycaemic status and presence of DKA at diagnosis. The system proved safe even in children under six, supporting its current use in off-label settings with appropriate clinical oversight. Our findings reinforce the value of early AID adoption to optimize long-term metabolic outcomes in pediatric T1D.
Tangsholpan Zholmukhan, Murat Smagulov, Nurlan Kairbekov
et al.
The relevance of the research theme, lies in its importance for the effective formation of a harmoniously developed personality and society, as well as maintaining the cultural, social and economic prosperity of the modern society. The purpose of this article is to study the social aspects of the educational potential within the framework of Islamic studies, as well as to reveal the
conceptual approach to its understanding. The publication materials are the information from scientific and practical disciplines, highlighting the implementation of social aspects of the educational potential of the Islamic studies in its natural environment and in a foreign context, as well as monographic works on the topic and sources that are a fixation of research observations in the field of modernization of the educational and social systems of the society life. The following concepts are the result of its study: fixing the problem under study as a composite and inseparable and a large-scale system; determination of its leading categories; establishing the relationship
between its components; the authors’ interpretation of this topic as a panoramic phenomenon interconnected with many other spheres of society. The practical significance of the work was revealed in its novelty, relevance and demand among representatives of the leading areas of scientific thought and activity such as history and sociology; ethics and religious studies; cultural studies and economics; pedagogy and psychology. The prospect of the problem under consideration lies in the need to observe, fix and study its reflection and relevance in the life of modern society, as well as in the importance of expanding and enriching the boundaries of “Islamic studies” as a scientific direction that promotes awareness of the role of spiritual principle in the formation of
personality, as well as in saving the ethnic groups for their prosperity and progress.
A longstanding goal of artificial general intelligence is highly capable generalists that can learn from diverse experiences and generalize to unseen tasks. The language and vision communities have seen remarkable progress toward this trend by scaling up transformer-based models trained on massive datasets, while reinforcement learning (RL) agents still suffer from poor generalization capacity under such paradigms. To tackle this challenge, we propose Meta Decision Transformer (Meta-DT), which leverages the sequential modeling ability of the transformer architecture and robust task representation learning via world model disentanglement to achieve efficient generalization in offline meta-RL. We pretrain a context-aware world model to learn a compact task representation, and inject it as a contextual condition to the causal transformer to guide task-oriented sequence generation. Then, we subtly utilize history trajectories generated by the meta-policy as a self-guided prompt to exploit the architectural inductive bias. We select the trajectory segment that yields the largest prediction error on the pretrained world model to construct the prompt, aiming to encode task-specific information complementary to the world model maximally. Notably, the proposed framework eliminates the requirement of any expert demonstration or domain knowledge at test time. Experimental results on MuJoCo and Meta-World benchmarks across various dataset types show that Meta-DT exhibits superior few and zero-shot generalization capacity compared to strong baselines while being more practical with fewer prerequisites. Our code is available at https://github.com/NJU-RL/Meta-DT.