Abstract: Family-based treatment (FBT) is widely regarded as the leading evidence-based treatment for adolescents with anorexia nervosa and has demonstrated benefits, particularly in supporting weight restoration. FBT is organised around five central tenets: an agnostic view of illness causation, a non-authoritarian therapeutic stance, parental empowerment, externalisation of the eating disorder and a pragmatic focus on symptom reduction. However, evidence of group-level efficacy does not establish that treatment is appropriate, accessible or psychologically safe for every young person and family. This lived-experience-informed critical analysis brings treatment research into conversation with qualitative evidence, recovery-oriented scholarship and emerging understandings of treatment-related harm. It examines how the five tenets distribute authority, credibility and responsibility and how their enactment may shape psychological safety and recovery. Agnosticism may become incuriosity about the functions and contexts of eating-disorder behaviours; non-authoritarianism may redistribute rather than reduce clinical authority; parental empowerment may displace the young person’s voice; externalisation may support separation from the eating disorder for some while invalidating the testimony of others; and pragmatism may privilege physical restoration without adequately addressing psychological and relational needs. These tensions may be particularly consequential for young people whose presentations include neurodivergence, trauma, chronic illness or other intersecting needs. Rather than rejecting FBT, we propose five corresponding shifts: from agnosticism to explorative curiosity, from non-authoritarianism to meaningful patient choice, from parental empowerment to shared empowerment, from externalisation to individualised differentiation, and from pragmatism to person-centred responsiveness. These shifts may preserve the urgency of nutritional care while strengthening collaboration, dignity and psychological safety.

Abstract: Mapping host activity to the MITRE ATT&CK framework usually needs manual work or a cloud-hosted large language model, which raises privacy and cost concerns for small organisations that cannot send their logs off-host. We ask whether a small language model (SLM), running locally and without fine-tuning, can classify Linux auditd records at the technique level of MITRE ATT&CK. We built a deterministic cascade that captures, filters, scores, and aggregates auditd events into session narratives, discarding most records before any model call. Only these narratives reach a quantised SLM (Qwen 2.5 7B, served by Ollama). Candidate techniques are calibrated only on public sources (MITRE ATT&CK, GTFOBins, Sigma), never on the evaluation data. We evaluated the full pipeline on the CAM-LDS dataset, across 29 attack variants, with deterministic and reproducible model outputs under a fixed seed and a temperature of zero. The system reaches a macro-averaged incident-level F1 of 54.2%. Per-variant scores range from 19% to 78%, driven by how much of the attack auditd can observe: scripted attacks leave few visible events, while interactive ones expose far more. An initial pilot with Phi-3 Mini 3.8B produced invalid, hallucinated technique IDs and missed obvious techniques, motivating the switch to Qwen 2.5 7B, which gives the best accuracy–latency trade-off. Our central finding is where the loss happens: the deterministic cascade preserves almost all of the observable on-host activity, so the binding constraints are downstream. The model under-confirms techniques it is given 71% of missed techniques were already available to it and over-broad signatures, not hallucinations, cause most false positives. A local, fine-tuning-free SLM gives a usable technique-level baseline; the ceiling is the model and the catalogue, not the filtering stage.

Abstract: Psychological control — the extent to which an organization shapes employees’ inner thoughts, feelings, and self-expression rather than merely their outward behavior — is a pervasive but understudied feature of Korean workplace culture. This study examined how psychological control is experienced by Korean employees, how it is perceived to function (as a management tool) and dysfunction (as a source of psychological strain), and how these perceptions relate to the broader Korean organizational-cultural context. Using a descriptive-correlational design, a structured online questionnaire was completed by 403 Korean employees across a range of occupations and tenure levels. Respondents reported experiencing a high degree of psychological control at work (M = 4.00, “Agree”), most strongly reflected in having sacrificed personal values to meet organizational expectations and having hidden opinions for fear of negative consequences. Respondents also strongly endorsed the dysfunctional consequences of psychological control (M = 4.20, “Strongly Agree”) — increased competition and tension, suppressed emotion, reduced trust, and higher turnover intention — while agreeing only moderately that control brings functional benefits such as focus, responsibility, and efficiency (M = 3.54, “Agree”). Experienced psychological control was significantly and positively correlated with perceived dysfunctional outcomes (ρ = .569, p < .001) and with the perceived cultural normalization of organizational conformity (ρ = .307, p < .001), but only weakly correlated with perceived functional benefits (ρ = .240, p < .001), and functional and dysfunctional perceptions were themselves statistically unrelated (ρ = .080, p = .111). These findings suggest that, in the Korean workplace context studied, psychological control is experienced primarily as a source of strain rather than a motivational tool, and that beliefs about its benefits operate largely independently of beliefs about its costs.

Abstract: Robust Optimization Over Time (ROOT) addresses dynamic optimization problems where the goal is not to identify the best solution at each instant, but to find solutions that remain effective across multiple future environments. Despite recent algorithmic advances, the theoretical foundations of ROOT remain limited, especially regarding its formulation in terms of survival time. Additionally, discrete problem settings have received little attention. This paper addresses these limitations from a theoretical perspective by analyzing survival time in discrete ROOT under stochastic k-bit-flip environmental dynamics. Using drift analysis and martingale techniques, we derive tight analytical bounds and identify phase transitions for survival time on dynamic OneMax, capturing gradual fitness degradation. For dynamic LeadingOnes, which models catastrophic structural failure, we prove that survival time follows an exact geometric distribution. Monte Carlo simulations closely match all theoretical predictions. The results connect ROOT with classical runtime analysis and formalize the mechanisms governing solution durability, providing a mathematical foundation for predictive robustness models, survival-aware heuristics, and deployment policies tailored to combinatorial environments.

Abstract: We review what is known about finite-time singularity formation ("blow-up") in the unsteady Prandtl boundary-layer equations, from the numerical discovery of the van Dommelen--Shen singularity in 1980 to the rigorous description of its self-similar structure completed in 2018--2022. The survey is organized around five questions: (i) what the singularity is physically (unsteady separation and eruption of the boundary layer); (ii) where it is known to occur (the impulsively started cylinder, vortex-induced separation, colliding wall streams); (iii) what has been proved (the E--Engquist theorem, the Kukavica--Vicol--Wang theorem, the Collot--Ghoul--Ibrahim--Masmoudi description of the stable blow-up pattern, the classification of inviscid singularities); (iv) how blow-up relates to the ill-posedness of the Prandtl equations without monotonicity and to the inviscid limit of the Navier--Stokes equations; and (v) how the reduced one-dimensional system on the symmetry axis, which carries the rigorous theory, sits inside the recent similarity transformation of Sun [Phys.\ Fluids \textbf{36}, 083616 (2024)] as its \(m=1\) member. We also record the explicit affine blow-up solution \(u=-x/(t_*-t)\), \(v=y/(t_*-t)\), which is exact for both Prandtl and Navier--Stokes, and explain why Prandtl blow-up bears no relation to the finite-energy Navier--Stokes blow-up announced in September 2026. A list of open problems closes the survey.

Abstract: The Rubin survey is expected to extend the halo white-dwarf luminosity function into a regime where current data neither establish nor exclude remnants with cooling ages beyond the standard cosmic age. Public Montreal thick-hydrogen DA tracks give dwell times in the window M_bol = 17–20 from 4.84 Gyr at 0.5 M_☉ to 0.70 Gyr at 1.2 M_☉. Under an illustrative uniform cooling-age density on [T₀, 2T₀], with T₀ = 13–14 Gyr, the Montreal-only occupancy is 0.10–0.19 at 0.6 M_☉, at most 0.21 on the grid, and zero at 0.95 M_☉ and above. BaSTI-calibrated bright-edge shifts widen the 0.6 M_☉ range to 0.10–0.27; no analyzed channel exceeds 0.3720. A robust-within-set candidate whose conservative cooling-age lower bound exceeds the standard ceiling under the youngest age assignment in the declared model set is a chronology anomaly under that set, regardless of abundance. A null remains parametric: the population inputs are underived, and the frozen two-stage Rubin operator has not been executed on Rubin observations. Supplementary Material 1 regenerates the calculations and figure.

Abstract: To address the issues of lengthy vibration durability test cycles for seat frames and poor consistency of damage equivalence at different locations under multiaxial excitation in existing acceleration methods, this paper proposes a frequency-segmented accelerated vibration test method based on damage rate equivalence. The method derives an analytical relationship between damage rate and stress spectral moments based on the Dirlik broadband damage estimation, enabling frequency-selective amplification of the acceleration PSD, thereby compressing test duration while maintaining a constant damage rate ratio at critical nodes. The proposed method is implemented in three strategies—triaxial full-band, triaxial frequency-segmented, and uniaxial frequency-segmented—and compared with the Inverse Power Law (IPL) and Fatigue Damage Spectrum (FDS) methods. Through finite element analysis of a seat frame and dynamic force measurements on a six-degree-of-freedom hydraulic shaker, the damage rate ratios at four structural nodes are used as indicators to evaluate each method. Results show that when the triaxial frequency-segmented strategy (DRR-TP) achieves the target damage rate ratio of 0.25, the thrust increase is only 17.1 %, substantially lower than the 21.0 % increase from full-band acceleration, while displacement requirements remain nearly unchanged. At the calibrated node, DRR-TP provides the best accuracy, comparable to FDS and superior to IPL and the uniaxial strategy. However, deviations exist across all methods at non-calibrated nodes, with IPL and the uniaxial strategy being particularly pronounced. Furthermore, accelerated high-intensity road spectra may produce stresses approaching the yield limit, necessitating amplitude constraints. Overall, the DRR-TP method offers a favorable balance between efficiency and equipment feasibility, providing a practical framework for accelerated durability testing of seat frames.

Abstract: Project-type production in the Heating, Ventilation, and Air Conditioning (HVAC) sector is highly vulnerable to scheduling disruptions caused by high product variety, complex routing structures, and uncertainty in the delivery of critical externally sourced components. In the investigated manufacturing environment, supplier-related delivery variability, rather than internal capacity limitations, constitutes the dominant source of schedule instability. This study develops a data-driven, event-based dynamic rescheduling framework that integrates supplier delay characterization with a Mixed-Integer Linear Programming (MILP) model. Historical procurement records obtained from the enterprise resource planning system were analyzed using K-Means clustering, identifying four representative delay classes of 5, 26, 51, and 127 days. These empirically derived patterns were incorporated into the rescheduling logic. The framework is activated at the disruption information time, t_info, when completed and ongoing operations are fixed to preserve schedule feasibility. A rolling-horizon mechanism freezes the subsequent five-day production window while reoptimizing the remaining horizon. Delays exceeding 40 days trigger an escalation rule that enforces the fastest available logistics mode. The multi-objective formulation simultaneously minimizes service-level agreement penalties, logistics expenditures, and operational energy costs. Implemented in Python 3.9 and solved with Gurobi 11.0, the model attained a 0.00% optimality gap within 1.2-2.4 s across the tested instances. Results show that early disruption information limits logistics costs to approximately 2,397 USD, whereas delayed information increases them to approximately 2,723 USD. The proposed framework provides a computationally efficient and transferable decision-support mechanism for coordinating production rescheduling and logistics expediting under supplier uncertainty, while indicating that expediting is economically justified when its marginal cost per recovered day remains below the corresponding dynamic delay penalty.

Abstract: Habitat degradation caused by anthropogenic disturbance is a major threat to global biodiversity, yet the spatiotemporal dynamics of the ecological intactness of giant panda habitat and the associated population responses remain poorly quantified. Focusing on the Sichuan section of the Giant Panda National Park and the Liangshan Mountains, using multi-source anthropogenic disturbance data and three national giant panda surveys, we integrated fuzzy overlay analysis, Theil–Sen trend analysis, Geodetector, and the Mann–Whitney U test to analyze the spatiotemporal dynamics, driving factors, and population responses of habitat intactness. The mean habitat intactness in 2020 was 6.70. Intactness declined slowly during 1980–2020, and 73.7% of the habitat remained stable. Land-use type and road construction were the dominant driving factors, and the explanatory power of any two interacting factors exceeded that of either factor alone. The intactness of giant panda occurrence sites was higher than that of random background points (significant in five of the six survey-by-region comparisons) and was higher in the Liangshan Mountains than in the national park. However, intactness was not significantly correlated with population density. These results indicate that wild giant pandas prefer high-intactness habitat, although intactness alone does not predict population density, and they provide empirical evidence for intactness-oriented conservation management.

Abstract: Local information can be lost rapidly while a collective mode survives for a long time. We study this distinction on a physical–observation joint state space. The second variation of relative entropy under a local record map gives a positive Fisher-loss operator; the least loss over normalized noninvariant scores is the spectral gap of the associated posterior-resampling process. The local strength and the compatibility of surviving directions enter separately. A fixed-width interacting ladder has a positive gap at every fixed finite coupling, as follows from an explicit block comparison, although strong correlation makes the bound small. We then reverse the construction. For a given irreducible real Hamiltonian with nonpositive off-diagonal entries, its ground amplitude determines pair-record probabilities and energy-valued loss weights, reproducing the physical excitation gap exactly. A positive trial amplitude gives an error interval controlled by the oscillation of its local energy, without requiring the exact ground state. When signs remain, the energy form is a difference of positive information forms and an explicit cancellation constant controls the gap. Sparse diagonalization of open transverse-field chains up to 65 536 configurations agrees with both the reconstructed information spectra and an independent free-fermion calculation. The critical law n∆_n/J → π, the gapped regime, and the ordered finite-chain splitting are distinguished. The result is an explicit connection for specified channels and Hamiltonians, with the requirements of positivity, locality, and energy normalization kept separate.

Abstract: Climate change affects livelihoods across Africa, but awareness of climate change and support for government action may have different social correlates. This study identifies the social correlates of self-reported climate change awareness and of conditional support for immediate government action to limit climate change across 39 African countries, using Afrobarometer Round 9 data. We estimate weighted logistic and cumulative-logit models with country fixed effects and country-clustered standard errors. The awareness model includes respondents with valid awareness responses, while the policy-support model is restricted to respondents who reported awareness. Education is strongly associated with awareness: relative to no or informal education; post-secondary and post-graduate education correspond to a 22.6 and 26.6 percentage-point higher probability of reporting awareness respectively. Media exposure, community-meeting attendance, and collective action have smaller but still significant associations with awareness, corresponding to increases of 4.3, 3.3, and 5.9 percentage points, respectively. Among respondents who report awareness, perceived freedom of expression corresponds to a 2.4 percentage-point higher probability of strongly supporting government action, whereas perceived religious-leader corruption corresponds to a 3.2 percentage-point lower probability. Awareness and conditional policy support are therefore stratified by different factors: because education and information access shape who hears about climate change, educational and communication policies remain central to extending climate information across Africa, while institutional perceptions matter for sustaining public backing for costly government climate action among those already aware.

Abstract: Road infrastructure maintenance is a critical challenge in urban environments, where potholes pose safety risks and raise operational costs. Existing pothole-detection systems mainly report road damage and provide no integrated repair mechanism, leaving a delay between identification and action. This paper presents Roadronix, an AI-based semi-autonomous system that detects potholes from image input, classifies their severity, selects a repair action through a rule-based decision engine, and simulates a step-by-step repair pipeline (aligning, filling, compacting, completion) during road preparation stages. The system offers AI and manual operating modes and a dashboard that reports detection counts, repair stages, and performance metrics. A simulation-based evaluation demonstrates the complete detection-to-repair workflow without physical hardware. The approach aims to shift road maintenance from a reactive to a proactive process.

Abstract: This mixed-methods study examines the pedagogical efficacy of AI-driven leadership simulations in undergraduate leadership education. The author served as course instructor. Twenty-nine students completed six sequential simulations paired with a reflective writing assignment over one six-week summer term. Qualitative thematic analysis reveals three primary outcomes: (1) simulations produced characteristic dip-then-climb learning trajectories consistent with productive failure frameworks, (2) personalized career profiles functioned as "mirrors with teeth" surfacing gaps between self-perception and demonstrated behavior, and (3) students underwent epistemological shifts from certainty-seeking to ambiguity tolerance. Quantitative self-report data indicates significant increases in adaptive leadership orientation (17% to 62%), listening-first approaches (24% to 76%), and comfort with uncertainty (21% to 66%). Findings suggest that consequence-driven simulation environments, when paired with individualized feedback and structured reflection, can accelerate the transition from declarative leadership knowledge to enacted judgment, particularly by surfacing relational skill gaps among analytically strong students.

Abstract: Reinforcement learning from human feedback (RLHF) treats annotators as preference oracles. This is expensive and can encode annotator-pool bias into the reward signal: rater demographics, positional artifacts, and misspecified helpfulness criteria can become part of what the model is trained to produce. A growing body of work—including AI-labeled feedback on summarization and dialogue, direct preference optimization, self-rewarding models evaluated on AlpacaEval 2.0, and iterative self-evolution for multimodal models—shows that competitive benchmark performance can, in several evaluated settings, be obtained without dense human pair labels. Whether annotator-induced bias decreases when humans leave the loop is a separate question that these works do not directly measure. We propose a protocol called SPARSE-ALIGN in which humans remain in the alignment loop but perform a different task. A policy generates its own candidate pairs. A frozen LLM judge, operating inside a debate structure intended to reduce reward hacking, evaluates the pairs for consistency. High-consistency pairs are accepted automatically. A small, stratified sample of routed pairs—at most 5% of training pairs in any run—is sent to human auditors. Auditors never directly provide an A/B preference or ranking. Instead, they flag non-factual bias artifacts (length preference, positional bias, sycophancy, demographic stereotyping, self-preference) or, when objective correctness is at stake, provide a sparse factual reference. A factual reference may condition the automated judge and therefore indirectly affect the machine-generated preference label; direct human pairwise preference labels remain excluded from the training process. We hypothesize that this arrangement can match RLHF on instruction-following quality while reducing annotator-induced bias relative to dense RLHF and ungrounded self-rewarding. The paper specifies the experiments, statistical tests, ablations, and failure conditions that would confirm or falsify those predictions. No experimental results are reported here.

Abstract: Generative artificial intelligence (GenAI) is increasingly used in corporate reporting, where accounting policies and estimates require professional judgment. This conceptual study examines how GenAI may assist preparers without displacing human responsibility or reinforcing managerial bias. Using theory adaptation, it integrates research on professional judgment, motivated reasoning and automation reliance. It identifies five bias-amplification mechanisms: anchoring, automation bias, outcome-directed prompting, unverified justification and diffusion of responsibility. The paper develops TRACE, a proposed framework comprising task boundary, regulatory grounding, alternative assessment, challenge and contradiction, and evidence trail. Its intended operation is illustrated through four constructed IFRS cases covering IAS 8 and IAS 16, IAS 36, IAS 37 and IFRS 15, three involving climate-transition assumptions. The AI proposals are stylised illustrations, not actual model outputs or empirical data. The analysis suggests that GenAI is most defensible when used to identify alternatives and contradictory evidence and to challenge a preferred treatment, and least defensible when asked to select or justify a conclusion. TRACE may make AI-assisted judgments more attributable and contestable, but it cannot guarantee unbiased or technically correct outcomes; independent human review remains necessary. The framework has not been empirically validated. It considers governance, control, reporting reliability and sustainable management.

Abstract: Vision-language models (VLMs) have achieved remarkable progress in multimodal understanding and generation by integrating visual and linguistic representations. However, most current VLMs lack explicit mechanisms for persistent memory, limiting their ability to maintain contextual coherence, accumulate knowledge over time, and support long-term reasoning across extended interactions. To address these limitations, a diverse set of memory mechanisms has emerged, including latent memory, key-value caches, external memory stores, retrieval-augmented memory, and hybrid memory systems. Despite rapid advances, the design space of memory in VLMs remains fragmented, and a unified understanding of its design and application remains lacking. In this survey, we provide a comprehensive review of memory mechanisms in VLMs from a system-oriented perspective. We introduce a novel four-dimensional (4D) taxonomy that organizes existing approaches along four orthogonal aspects: when memory is maintained (temporal scope), where memory is stored (storage location), what memory encodes (information stored), and how memory is accessed, updated, and utilized (memory operations). Using this taxonomy as a unifying framework, we systematically analyze representative memory-enhanced VLM architectures, review evaluation protocols for memory capabilities, and summarize key application areas, including long-video understanding, multimodal dialogue, embodied agents, and robotic reasoning. We further discuss key challenges such as scalability, memory efficiency, continual updating and forgetting, and multimodal grounding, and outline promising directions toward adaptive and unified memory systems. Overall, this survey provides a comprehensive foundation for understanding memory in VLMs and serves as a roadmap for developing next-generation multimodal systems with persistent memory and long-term reasoning capabilities. A repository associated with this survey is available at https://github.com/Xzcv-hub/Awesome-Memory-in-VLM.

Abstract: We introduce a family of polynomial vector fields on the plane generated by factorization in the ring of radial invariants. This family, referred to as the Radial Factorization Family, is generated by coupling an arbitrary monic polynomial \( P \) in the radial invariant \( \Delta=x^{2}+y^{2} \) with the standard rotational generator. It admits an explicit reduction in polar coordinates, yielding a clean global unperturbed model whose periodic orbits are concentric circles with explicitly known radii. By adding a small polynomial perturbation that breaks rotational symmetry, we compute the first-order Melnikov function and establish that the displacement function of the Poincaré return map admits the expansion \( d(r,\varepsilon)=2\pi\varepsilon\,\mathcal{M}(r)+O(\varepsilon^{2}) \) on compact subintervals of the period annulus. Because the unperturbed orbits are circles, the resulting integrals admit explicit evaluation. For the linear center this recovers the classical bound \( \lfloor(d-1)/2\rfloor \) on the number of bifurcating limit cycles; near a hyperbolic invariant circle of a general member of the family the same framework shows unique persistence of a single limit cycle, while in the degenerate case of multiple roots the circles lose normal hyperbolicity and unfold under perturbation into at most \( m \) limit cycles controlled by the multiplicity. The central novelty of the construction lies in its inverse character: in higher dimensions it yields an optimal inverse-realization procedure that converts any prescribed collection of nested invariant spheres—together with a topologically admissible assignment of stability types—into a polynomial vector field of minimal degree within the radial-factorized equivariant class. This provides a systematic algebraic dictionary for topological synthesis, transforming the traditional paradigm of "analyse a given system" into "t design a system with prescribed dynamical skeleton". The framework provides a constructive algebraic perspective related to aspects of Hilbert's sixteenth problem and offers a practical tool for phase-space engineering in nonlinear dynamics.

Abstract: Artificial Intelligence (AI) is increasingly transforming education through personalized learning, intelligent tutoring, automated assessment, learning analytics, teacher support, and generative AI applications. However, the benefits of AI are not distributed equally, particularly in developing countries where digital infrastructure, teacher readiness, affordability, data governance, language diversity, and institutional capacity remain significant challenges. This study critically examines the opportunities, challenges, and future implications of AI in education, with particular attention to developing-country contexts and Bangladesh. A PRISMA-informed Systematic Critical Literature Review (SCLR) was conducted using peer-reviewed research, systematic reviews, meta-analyses, and authoritative institutional reports published primarily between 2018 and 2026. The review synthesizes evidence concerning personalized learning, teacher augmentation, generative AI, assessment, academic integrity, AI literacy, digital inequality, privacy, bias, and contextual adaptation. Bangladesh-specific evidence is incorporated to examine the country's readiness and the practical requirements for responsible AI adoption in education. The review indicates that AI can improve personalization, student engagement, feedback, learning efficiency, teacher productivity, and access to educational support. Recent experimental evidence also suggests that generative AI can produce positive learning effects when integrated with appropriate instructional design. Nevertheless, these benefits depend heavily on teacher involvement, infrastructure, institutional governance, and responsible implementation. Risks include misinformation, academic dishonesty, algorithmic bias, privacy violations, overreliance on automated systems, unequal access, and the marginalization of local languages and cultures. Based on the synthesis, this study proposes the Socio-Technical AI Integration Framework (STAIF), consisting of seven interconnected pillars: infrastructure-first deployment, teacher-in-the-loop implementation, local-language and cultural adaptation, offline-first accessibility, equity safeguards, responsible data governance, and continuous monitoring and evaluation. The framework provides a practical pathway for developing countries such as Bangladesh to move from fragmented AI experimentation toward equitable, sustainable, and human-centred AI integration in education.

Abstract: The methylxanthine caffeine is one of the most widely consumed neuroactive substances globally. Caffeine in the context of coffee consumption has been associated with diverse effects on human health and has been widely reviewed. Caffeine has drawn considerable interest due to its effects on cellular signalling, DNA damage sensitivity, Target of Rapamycin (TOR) activity, activation of AMP-activated Protein Kinase (AMPK), and extension of chronological lifespan in fission yeast and other model organisms. The cell cycle effects of caffeine have overshadowed its modulation of DNA damage and repair. Caffeine provides an interesting example of how scientific knowledge evolves and is influenced by trends within epochs. Despite recent advances in our understanding of how caffeine modulates cell cycle progression, these findings raise many new questions. These include its structural similarity to adenine and the mechanisms by which it affects DNA damage and repair, and its impact on TOR Complex 2 (TORC2) activity. Herein, I review our current knowledge on the cellular effects of caffeine and focus primarily on studies in fission yeast. It proposes that caffeine be considered a radiomimetic compound, outlining our current state of knowledge and future directions in applying its pharmacology in the prevention and treatment of disease.

Abstract: PURPOSE: Use of prosthesis to replace small hand joints, mainly in patients with de-generative osteoarthritis or with sequelae of fractures, is more and more common. As for other joints, use of prosthesis usually leads to good functional results, but is not free of complications such as implant infection. We discuss a possible solution presenting a single case treated by means of a custom-made spacer in our hospital comparing it to the very few options available in literature. METHODS: A 30-year-old active male, treated in another hospital two years before for an articular fracture of the middle phalanx of the ring finger of his left hand, came to our center with a rigid digit. He was therefore treated with a Swanson proximal interpha-langeal (PIP) joint prosthesis. At one month follow-up the patient presented clear infection signs of his implant. Authors removed the infected implant and replaced it with a small custom-made antibiotic impregnated cement spacer. The patient underwent a long lasting post-operative oral antibiotic therapy. Four months later a new Swanson prosthesis has been successfully implanted. Literature review has been performed. RESULTS: When the patient first came to our hospital, he presented with a rigid digit, reaching the day after our first surgery a PIP joint range of motion (ROM) in flex-ion-extension of 0-90°. At one-month follow-up, with the infected prosthesis, the ROM was 10-40°. After the last surgery, with the implant of a new Swanson prosthesis, at the six-month follow-up the patient was infection free, the PIP joint ROM was 10-90° and the patient was very satisfied. CONCLUSIONS: The literature discusses few cases of PIP prosthesis infection, and most of those are treated with implant removal and arthrodesis. The possibility of using a cus-tom-made antibiotic-impregnated cement spacer to achieve good results, as is universally accepted for other joints, should be kept in mind by all hand surgeons who face to prevent more drastic solutions.