Purpose This study aims to propose a conceptual framework for metaverse tourism from a sociotechnical perspective, highlighting the interactions between metaverse technological capabilities and travelers’ behaviors to enhance the understanding of tourism experiences in the metaverse, and suggesting research opportunities to further advance the discourse on metaverse tourism. Design/methodology/approach A framework for metaverse tourism is developed by applying a sociotechnical perspective to the contexts of hospitality and tourism. This framework is based on a review of the literature on the metaverse in the fields of hospitality and tourism, marketing and information systems. Findings Existing frameworks on metaverse tourism have emphasized experiential dimensions and motivations, often overlooking the technological factors that shape user decisions. The proposed framework identifies the key technological capabilities and digital infrastructure in metaverse tourism that influence travelers’ behaviors. It also highlights the interconnections between the virtual and physical worlds via smart tourism layers, shaping immersive and absorptive involvement. Research limitations/implications This study identifies new research opportunities and highlights potential applications of metaverse tourism in the field of hospitality and tourism. Originality/value This study introduces a novel sociotechnical framework for understanding metaverse tourism, offering insights into how tourism experiences emerge from the interplay between technological capabilities and traveler behaviors.

Artificial Intelligence (A.I.) is a multidisciplinary field whose goal is to automate activities that presently require human intelligence. Recent successes in A.I. include computerized medical diagnosticians and systems that automatically customize hardware to particular user requirements. The major problem areas addressed in A.I. can be summarized as Perception, Manipulation, Reasoning, Communication, and Learning. Perception is concerned with building models of the physical world from sensory input (visual, audio, etc.). Manipulation is concerned with articulating appendages (e.g., mechanical arms, locomotion devices) in order to effect a desired state in the physical world. Reasoning is concerned with higher level cognitive functions such as planning, drawing inferential conclusions from a world model, diagnosing, designing, etc. Communication treats the problem understanding and conveying information through the use of language. Finally, Learning treats the problem of automatically improving system performance over time based on the system's experience. Many important technical concepts have arisen from A.I. that unify these diverse problem areas and that form the foundation of the scientific discipline. Generally, A.I. systems function based on a Knowledge Base of facts and rules that characterize the system's domain of proficiency. The elements of a Knowledge Base consist of independently valid (or at least plausible) chunks of information. The system must automatically organize and utilize this information to solve the specific problems that it encounters. This organization process can be generally characterized as a Search directed toward specific goals. The search is made complex because of the need to determine the relevance of information and because of the frequent occurence of uncertain and ambiguous data

The metaverse is an innovative digital environment that aims to offer immersive experiences to consumers by integrating virtual and physical worlds. By providing opportunities across various fields, it enhances interaction levels, facilitates access to information, and improves service quality. However, it is anticipated that the metaverse will also bring about issues related to privacy, confidentiality, data security, and ethics. In particular, personal data collected within metaverse environments and the sensitivities associated with privacy are considered significant ethical risk areas. In this context, the ethical issues arising from the metaverse have increasingly become an important field of academic research. This study analyzes 388 academic publications published between 2008 and 2025 in the Web of Science (WoS) database that jointly address metaverse and ethics, using bibliometric analysis with the R and VOSviewer program. The research findings indicate that academic interest in metaverse and ethics has increased markedly, especially since 2022. In terms of publication and citation counts, China, the United States, and the United Kingdom are identified as leading countries, while China and the United States stand out in international academic collaborations. In addition, concepts such as virtual reality, artificial intelligence, data privacy, security, and ethics are found to be at the core of metaverse research. This study reveals the current academic structure of the metaverse and ethics field and provides a guiding perspective for future research on metaverse ethics.

Analog electronics, i.e., circuits that process continuously varying signals, have quietly powered the backbone of agricultural automation long before the advent of modern digital technologies. Yet, the accelerating focus on digitalization, IoT, and AI in precision agriculture has largely overshadowed the enduring, indispensable role of analog components in sensing, signal conditioning, power conversion, and actuation. This paper provides a comprehensive state-of-the-art review of analog electronics applied to agricultural systems. It revisits historical milestones, from early electroculture and soil-moisture instrumentation to modern analog front-ends for biosensing and analog electronics for alternatives source of energy and weed control. Emphasis is placed on how analog electronics enable real-time, low-latency, and energy-efficient interfacing with the physical world, a necessity in farming contexts where ruggedness, simplicity, and autonomy prevail. By mapping the trajectory from electroculture experiments of the 18th-century to 21st-century transimpedance amplifiers, analog sensor nodes, and low-noise instrumentation amplifiers in agri-robots, this work argues that the true technological revolution in agriculture is not purely digital but lies in the symbiosis of analog physics and biological processes.

The rapid convergence of artificial intelligence (AI), cloud computing, and 5G communication has positioned extended reality (XR) as a core technology bridging the physical and virtual worlds. Encompassing virtual reality (VR), augmented reality (AR), and mixed reality (MR), XR has demonstrated transformative potential across sectors such as healthcare, industry, education, and defense. However, the compact architecture and limited computational capabilities of XR devices render conventional cryptographic authentication schemes inefficient, while the real-time transmission of biometric and positional data introduces significant privacy and security vulnerabilities. To overcome these challenges, this study introduces PXRA (PUF-based XR authentication), a lightweight and secure authentication and key distribution protocol optimized for cloud-assisted XR environments. PXRA utilizes a physically unclonable function (PUF) for device-level hardware authentication and offloads elliptic curve cryptography (ECC) operations to the cloud to enhance computational efficiency. Authenticated encryption with associated data (AEAD) ensures message confidentiality and integrity, while formal verification through ProVerif confirms the protocol’s robustness under the Dolev–Yao adversary model. Experimental results demonstrate that PXRA reduces device-side computational overhead by restricting XR terminals to lightweight PUF and hash functions, achieving an average authentication latency below 15 ms sufficient for real-time XR performance. Formal analysis verifies PXRA’s resistance to replay, impersonation, and key compromise attacks, while preserving user anonymity and session unlinkability. These findings establish the feasibility of integrating hardware-based PUF authentication with cloud-assisted cryptographic computation to enable secure, scalable, and real-time XR systems. The proposed framework lays a foundation for future XR applications in telemedicine, remote collaboration, and immersive education, where both performance and privacy preservation are paramount. Our contribution lies in a hybrid PUF–cloud ECC architecture, context-bound AEAD for session-splicing resistance, and a noise-resilient BCH-based fuzzy extractor supporting up to 15% BER.

Reducing the material cost of inorganic solid-state electrolytes is crucial to advancing all-solid-state batteries (ASSBs) for next-generation energy storage applications. The halospinel Li2Sc2/3Cl4 solid electrolyte (SE) possesses a high ionic conductivity of 1.5 mS cm-1 and good cycling stability up to 4.6 V. However, the high cost of Sc limits its practical application. In this study, we combine M3GNET universal machine learning interatomic potential (UMLIP) and density functional theory (DFT) for efficient screening of lower-cost cation-substituted halospinel compositions for synthesis. As a cost-mitigation strategy, predicted Mg2+-, Al3+-, and Zr4+-substituted Li2Sc2/3Cl4 spinels with substitution fractions ranging from 20.9% to 37.5% were experimentally synthesized with only minor impurities, achieving room-temperature ionic conductivities as high as 1.85 mS cm-1. Substitution of Fe3+ was also achieved, albeit with a 7% Fe2+ impurity. Molecular dynamics simulations (MD) using highly accurate moment tensor potentials (MTPs) indicate that Li+/Sc3+/Mn+ ordering plays a crucial role in determining the conductivity of disordered substituted compositions. ASSBs operating at 3.8 mAh cm-2 capacity with Li1.75Sc0.416Zr0.25Cl4 at a high current density of 2 mA cm-2 exhibited 80% of the capacity of more moderately loaded ASSBs cycled at a low rate. This work provides a foundational methodology for predicting the thermodynamic stability and ion transport of disordered lithium solid electrolytes and accelerating the discovery of novel materials for a range of applications.

Relevance. The article provides an essential overview of sports phygitalogy, a new scientific field that combines sports science, digital pedagogy, cyber-pedagogy, phygital pedagogy and psychology. It is based on the theory of human cybersocialization, which explains the emergence of Homo sapiens cyberus, a person socialized in cyberspace. This has led to the formation of Homo sapiens phygitalis, a rational human being who is socialized in both the physical and virtual worlds. Sports phygitalogy studies the development of sports in the context of this convergence, exploring the patterns, principles, and technologies that shape modern sports. It aims to understand how sports are influenced by the digital and virtual world, and how this impacts athletes, coaches, and fans. The subject and object of sports digitalology, its methodological principles, and its socio-pedagogical significance in the context of digitalization of sports activities are defined. Methods and research organization. The study used a retrospective analysis of scientific articles from the scientific electronic library Elibrary.ru, using the following queries: “phygital”, “phygitology”, “phygitalogy”, “sports phygitology”, and ‘sports phygitalogy”. The research also employed methods such as induction, deduction, abstraction, synthesis, and generalization; as well as pedagogical modeling and design methods. Results. The article discusses the author’s idea of sports phygitalogy, which is a system of knowledge about phygital sports and their role in the development of the individuals and society. It outlines the subject, object, main sections, principles, and immediate research prospects of sports phygitalogy. The author provides definitions for terms such as “sports phygitalogy”, “sports phygital psychology”, “sports phygital pedagogy”, “ethnophygitalogy”, and “sports ethnophygitalogy”. Conclusion. Sports phygitalogy is a theoretical discipline that also acts as a practical mechanism for ensuring the sustainable development of sports in today’s world, where the physical and digital realms are converging. This field has the potential to influence the way we think about the values, technologies, and social aspects of modern sports on a systemic level.

Abstract The currently known compact extreme emission-line galaxies (the “Green Peas” (GPs)) in the Sloan Digital Sky Survey (SDSS) are rare and were mostly found among serendipitous spectroscopic targets, thus leaving open the possibility that a substantial population of GPs is missed. A significantly larger number of identified GPs in the Local Universe might provide a better characterization of their high-redshift analogs and Lyman continuum escape. In this paper, we confront the challenges of robustly identifying GPs without spectroscopic information, a needed approach considering the incompleteness of spectroscopic surveys for compact sources. The principal difficulty stems from a significant contamination of photometric candidates by stars and quasars of similar color. To solve this, we introduce a spectral energy distribution matching method, which separates candidate GPs from contaminants on the basis of SDSS and Wide-field Infrared Survey Explorer photometry of spectroscopically confirmed stars, quasars, and galaxies. The method has an effectiveness of 85% and a contamination rate of ∼10%. With it, we identify ∼9600 GP candidates expected to lie in the 0.12 < z < 0.36 range—a tenfold increase over what would be selected using SDSS DR18 spectra. Some of the new GPs are as bright as r ∼ 19, and 1200 are predicted to have [O III ]5007 equivalent widths in excess of 500 Å. The new population contains many “Extended Peas,” which are absent among known GPs and possibly represent merging systems. We provide catalogs containing 8313 newly identified GP candidates, as well as 917 GPs confirmed using SDSS spectroscopy and 521 GPs with spectroscopic redshifts from LAMOST and other sources.

Driven by the concepts of digital twin and metaverse, constructing a high-fidelity, semantic-rich, and interactive digital copy of the physical world has become a key issue in the field of surveying, mapping, and geographic information. However, in typical complex landforms such as urban canyons and mountainous forest areas, the single-sensor data acquisition methods (such as UAV oblique images and lidar) has inherent information blind spots and accuracy bottlenecks. Traditional data fusion approaches predominantly focus on shallow geometric alignment and splicing at the geometric level, ignoring the heterogeneity of different data sources in semantic connotations, leads to common problems such as geometric distortion, detail loss, and semantic inconsistency in the fusion products. To break through this dilemma, this paper proposes an adaptive fusion framework for multi-source data of complex landforms (SAAF-Net) with deeply coupled semantic information. Centered on computer vision, this framework constructs a full-link technical process from raw data to high-precision semantic 3D models: Two-stream parallel semantic parsing: A two-stream deep semantic segmentation network for images and point clouds (based on SegFormer and PointNeXt) is designed to achieve fine-grained classification of scene features (the average intersection over union mIoU exceeds 90%), providing high-dimensional semantic priors for fusion. Semantic-guided cross-source registration: A semantic weighted iterative closest point algorithm (SW-ICP) is proposed. By constraining the corresponding point search space through cross-source semantic consistency and combining with the significance weighting of local geometric structures, the robustness problem of heterogeneous data registration is solved. Neural adaptive fusion modeling: A multi-factor driven neural network model is constructed to dynamically evaluate the confidence of data sources under different semantic categories and observation conditions, achieving the optimal fusion of pixel-level elevation and texture. Experiments in the city center and mountainous forest areas show that compared with mainstream methods, the root mean square error (RMSE) of SAAF-Net is reduced by 35% - 48%, and the completeness is improved to over 99%. Especially, the reconstruction quality in building edges, vegetation-covered areas, and light-shadow areas is significantly improved.with a substantial enhancement in visual realism. This research provides theoretical and technical support for the construction of a high-precision 3D base for digital twin cities.

In the active landscape of 6G wi-fi systems, combining knowledge graph technology and advanced data analysis strategies presents a transformative approach to monitoring and managing relay protection status in intelligent substations. This paper implements a risk identification framework for relay protection equipment operation in smart substations. Based on this, the paper proposes an innovative Knowledge Graph (KG) -based risk identification model that effectively combines the remote tracking abilities of KG with advanced predictive and analytical ability to ensure efficient, accurate, and reliable control over substation equipment. Our model facilitates real-time data collection and exchange by combining interconnected devices with sensors, actuators, and network connectivity, improving substations' operational performance and risk management. The LSTM component was also introduced, which is well-suited for analysing the time series data and also applied to monitor operational patterns, detect anomalies, and predict the future risk of failures. Simultaneously, the KG-based system is effectively used for local fault cases to identify risky equipment and become aware of fault paths within the substation's network. This approach effectively minimises human intervention by automating the monitoring process and significantly reducing the dependence on traditional, costly solutions. The proposed design consists of three key modules - KG-LSTM-FD for fault detection, KG-LSTM-RI for risk identification, and KG-LSTM-FL for fault location estimation. Our model proved exceptional sensitivity and reliability, achieving a 99.98% successful rate throughout diverse fault scenarios with location estimation errors within 1%. The KG-LSTM-based risk identification model, examined through a prototype system, marks a tremendous advancement in combining the physical world with computer-based systems, providing a costeffective, highly efficient, and scalable solution for the next generation of smart substations.

The inception of modern science, in the 17th century, was accompanied by epistemological analyses that consider its foundation as laid on observation and experiment — a stance often regarded as excluding (or, at least, devaluating) metaphysics, especially in the English-speaking world. Qualms about metaphysics are already noticeable in Locke’s Essay (1690), and were supposedly deepened by Hume, in the following century. For almost two hundred years, Hume’s philosophy was regarded as radically sceptical concerning metaphysics generally, but particularly about causality and the very existence of an external, objective reality. In this paper, we argue, following a more recent interpretive vein in Hume’s scholarship, that Hume’s scepticism about these basic metaphysical issues was effectively “mitigated” (in his own words) by his pioneering adoption of a form of naturalised realism. According to it, belief in both causal relations and in the external world is taken as resulting from the natural operations of the human mind and, in this condition, as being justified, in an epistemologically non-ordinary sense of this notion. As a consequence, epistemology was seen by him as an undertaking akin to the natural sciences, both employing similar empirical methods and metaphysical hypotheses to explore the mind, in the former case, and the physical world, in the latter. Thus, instead of interpreting Hume as one of the forerunners of the anti-metaphysical trend within empiricism, we believe that his wide-ranging project of instituting a “science of man” made some room for metaphysics, as cultivated within the same fallibilistic, naturalistic approach of empirical science itself.

Fantasy can be defined as an idea having its own system governed by its own rules. Man attains fantasy as an undetectable reality in his physical existence. The structure of our perceived physical world shares a large part of fantasies, resting in various faculties belonging to man. Fantasy ideas breathe and grow in mutual co-spaces, shared by these faculties, the fantasy literature has a fundamental role in developing the fantasy ideas in different disciplines. In reference to architecture, fantasy dialogues with architecture on different levels, having various aspects. The objective of this research is to explore those aspects and attempt to establish the importance of fantasy imagination in architectural design activity as well as in the experience of architectural space. The findings reveal the spatial narratives, visionary drawings and the role of digital fantasy in the experience of contemporary public-private spaces are few aspects in the discussion. This research investigates the possible role of fantasy literature as a critical and imaginative framework for architectural thinking and design activity.

As a key technology connecting the physical world and the digital world, 3D reconstruction has wide-ranging applications in fields such as autonomous driving, virtual reality, and industrial inspection. Traditional methods rely on handcrafted features and geometric constraints, and suffer from limitations like poor robustness and low integrity in complex scenarios. In recent years, deep learning technology, with its powerful feature learning and context modeling capabilities, has brought about revolutionary advancements to 3D reconstruction. This paper systematically reviews the research progress of deep learning-driven 3D reconstruction technology, focuses on analyzing core innovations including fusion architectures, attention mechanisms, and lightweight networks, and conducts an in-depth discussion on the breakthroughs and integration trends of cutting-edge representation technologies such as neural implicit fields and 3D Gaussian splatting. Furthermore, the paper points out that current technologies still face challenges such as generalization capability, dynamic scene processing, and computational overhead, and looks forward to future development directions including unified implicit-explicit modeling, lightweight deployment, and general-purpose 3D vision foundation models. It provides important references for in-depth understanding and promotion of the development of 3D reconstruction technology.

The ongoing exploration of the physical world has intensified the demand for intelligent computing in extreme environments. However, intelligent devices operating under extreme high-pressure conditions are limited by the pressure tolerance of the materials used for intelligent computing. A pressure-adaptive artificial synapse (PAAS) using VO2 (M1) nanoparticles is developed, leveraging the increased lattice rigidity during the M1-to-M1' phase transition (1 atm to 15.1GPa), which causes the photoinduced insulator-to-metal transition to be Mott dominated. The PAAS demonstrated a stable operating current, a superior biomimetic plasticity (maximum paired-pulse facilitation index from 109.6% to 155.4%), and an improved postsynaptic current linearity (Pearson's r from 0.64 to 0.97) from 1 atm to 15.1GPa. Furthermore, an artificial neural network mapped by PAAS under high pressure achieved a validation accuracy of 95%-97% in handwritten digit recognition. The PAAS is also applied to a convolutional autoencoder for denoising reconstruction of color images.

Many cultures use songs to influence the world around them (e.g. Levin 2006, Feld 1982). In Australian Aboriginal societies, song can be a tool to influence one’s environment. Such practices can affect the physical world—people, animals, plants— and the non-physical world, where spirit beings reside, a realm referred to as the Altyerre in Arandic languages of central Australia (Dobson 2007, Turner 2010, Wallace & Lovell 2009, Green 2012). Like the Tuvan people of Siberia who use singing and other verbal practices “to coexist peacefully with these spirit-masters and gain access to the resources under their control” (Levin 2006:28), many Indigenous Australians engage in ceremonial and other vocal practices to influence their world. In this article we explore song as a means to experience sentient ecology in two contrasting Aboriginal Australian contexts. In doing so, this contribution invites us to consider more broadly the role of humanly created sound in society.

This paper examines the evolving landscape of mobile robotics, focusing on challenges faced by roboticists working in industry when integrating robots into human-populated environments. Through interviews with sixteen industry professionals specializing in social mobile robotics, we examined two primary research questions: (1) What approaches to person detection and representation are used in industry? and (2) How does the relationship between industry and academia impact the research process? Our findings reveal diverse approaches to human detection, ranging from basic obstacle avoidance to advanced systems that differentiate among classes of humans. We suggest that robotic system design overall and human detection in particular are influenced by whether researchers use a framework of safety or sociality, how they approach building complex systems, and how they develop metrics for success. Additionally, we highlight the gaps and synergies between industry and academic research, particularly regarding commercial readiness and the incorporation of human-robot interaction (HRI) principles into robotic development. This study underscores the importance of addressing the complexities of social navigation in real-world settings and suggests that strengthening avenues of communication between industry and academia will help to shape a sustainable role for robots in the physical and social world.

Wide-field, multi-filter photometric surveys enable the reconstruction of the Milky Way’s star formation history (SFH) on Galactic scales and provide complementary insights into disc assembly. The 12-filter system of the Javalambre Photometric Local Universe Survey (J-PLUS) is particularly suitable, as its colours trace stellar chemical abundances and help mitigate the age–metallicity degeneracy in colour–magnitude diagram fitting. We aim to recover the SFH of the Milky Way disc and separate its chemically distinct components by combining J-PLUS DR3 photometry with Gaia astrometry. We also intend to test the potential of isochrone fitting to estimate ages and metallicities for individual stars as proxies for disc evolutionary trends. We fitted magnitudes and parallaxes of 1.38 łe4.2 mag) constrains stellar ages, while the faint region provides an empirical metallicity prior that mitigates the age–metallicity degeneracy. Both PARSEC and BaSTI isochrones, in solar-scaled and α-enhanced versions, were adopted. stars using a Bayesian multiple-isochrone technique. The bright region of the colour–absolute-magnitude diagram (M_ r The recovered SFH shows two sequences: an α-enhanced population forming rapidly between 12.5 and 8 Gyr ago, enriching from mathrm M/H ∼-0.6 to 0.1 dex; and a solar-scaled sequence emerging sim8 Gyr ago, dominating after sim7 Gyr with slower enrichment and reaching solar metallicity by 3 Gyr. Metal-rich (mathrm M/H stars are confined to $|z_ GC |łesssim1$ kpc, whereas metal-poor (mathrm M/H łesssim-0.5) stars reach $|z_ GC | kpc. Simultaneous fitting of solar-scaled and α-enhanced isochrones reveals distinct formation epochs for the thin and thick discs. J-PLUS multi-filter photometry, combined with Gaia parallaxes, effectively mitigates age–metallicity degeneracies and enables detailed mapping of the Milky Way’s temporal and chemical evolution.

Bio-inspired visuomorphic vision integrates multi-dimensional information (spectrum, spatial, temporal, and so on), providing an effective computational paradigm for sensing a visual scene in the physical world. Using photosensors with multi-dimensional information processing functionality to split complex optical information into visible and ultraviolet channels for separate perception and processing is the basis for constructing tetrachromatic vision systems. Here, by modulating the transport dynamics of photogenerated excitons between pentacene and ZnO thin films, both wavelength-dependent volatile positive photoconductance and non-volatile negative photoconductance characteristics are coupled into a single optoelectronic transistor. Utilizing the optoelectronic transistor as the tetrachromatic sensor, the constructed in-sensor computing system can effectively extract and identify the types of visible objects (99%) and the motion direction of ultraviolet objects (97%). This work provides a foundational hardware platform for intelligent artificial vision systems.

Classification constitutes a fundamental cognitive challenge for both biological and artificial intelligence systems. Here, we investigated how the brain categorizes stimuli that are not linearly separable in the physical world by analyzing the geometry of neural manifolds formed by macaque V2 neurons during a classification task involving motion-induced illusory contours. We identified two related but distinct neural manifolds: the sensory and perceptual manifolds. The sensory manifold was embedded in a three-dimensional subspace defined by three stimulus features, where contour orientations remained linearly inseparable. However, through a sequence of geometric transformations equivalent to twist operations, this three-dimensional sensory manifold expanded into a seven-dimensional perceptual manifold, enabling the linear separability of contour orientations. Computational modeling further demonstrated that this dimension expansion was facilitated by neurons exhibiting nonlinear mixed selectivity with heterogeneous response profiles. These findings provide insights into how biological neural networks enhance the dimensionality of representational spaces, illuminating the geometric mechanism underlying the transformation from sensation to perception.

We aim to characterise the mass-metallicity relation (MZR) and the 3D correlation between the stellar mass, metallicity, and star formation rate (SFR) known as the fundamental metallicity relation (FMR) for galaxies at 5<z<7. Using sim800 iii selected galaxies from deep NIRCam grism surveys, we present our stacked measurements of direct-̊m T_e metallicities, which we used to test recent strong-line metallicity calibrations. Our measured direct-̊m T_e metallicities (0.1--0.2,̊m Z_⊙ for M_⋆ M_⊙, respectively) match recent JWST/NIRSpec-based results. However, there are significant inconsistencies between observations and hydrodynamical simulations. We observe a flatter MZR slope than the SPHINX 20 and FLARES simulations, which cannot be attributed to selection effects. With simple models, we show that the effect of an O iii flux-limited sample on the observed shape of the MZR is strongly dependent on the FMR. If the FMR is similar to the one in the local Universe, the intrinsic high-redshift MZR should be even flatter than is observed. In turn, a 3D relation where SFR correlates positively with metallicity at fixed mass would imply an intrinsically steeper MZR. Our measurements indicate that metallicity variations at fixed mass show little dependence on the SFR, suggesting a flat intrinsic MZR. This could indicate that the low-mass galaxies at these redshifts are out of equilibrium and that metal enrichment occurs rapidly in low-mass galaxies. However, being limited by our stacking analysis, we are yet to probe the scatter in the MZR and its dependence on SFR. Large carefully selected samples of galaxies with robust metallicity measurements can put tight constraints on the high-redshift FMR and help us to understand the interplay between gas flows, star formation, and feedback in early galaxies.