Hasil untuk "Municipal refuse. Solid wastes"

Gregory N. Frank

Current alignment evaluation mostly measures whether models encode dangerous concepts and whether they refuse harmful requests. Both miss the layer where alignment often operates: routing from concept detection to behavioral policy. We study political censorship in Chinese-origin language models as a natural experiment, using probes, surgical ablations, and behavioral tests across nine open-weight models from five labs. Three findings follow. First, probe accuracy alone is non-diagnostic: political probes, null controls, and permutation baselines can all reach 100%, so held-out category generalization is the informative test. Second, surgical ablation reveals lab-specific routing. Removing the political-sensitivity direction eliminates censorship and restores accurate factual output in most models tested, while one model confabulates because its architecture entangles factual knowledge with the censorship mechanism. Cross-model transfer fails, indicating that routing geometry is model- and lab-specific. Third, refusal is no longer the dominant censorship mechanism. Within one model family, hard refusal falls to zero while narrative steering rises to the maximum, making censorship invisible to refusal-only benchmarks. These results support a three-stage descriptive framework: detect, route, generate. Models often retain the relevant knowledge; alignment changes how that knowledge is expressed. Evaluations that audit only detection or refusal therefore miss the routing mechanism that most directly determines behavior.

Maxim Dedyaev

The rapid expansion of artificial intelligence in public governance has generated strong optimism about faster processes, smarter decisions, and more modern administrative systems. Yet despite this enthusiasm, we still know surprisingly little about how AI actually takes shape inside different layers of government. Especially in federal systems where authority is fragmented across multiple levels. In practice, the same algorithm can serve very different purposes. This study responds to that gap by examining how AI is used across federal, state, and municipal levels in the United States. Drawing on a comparative qualitative analysis of thirty AI implementation cases, and guided by a digital-era governance framework combined with a sociotechnical perspective, the study identifies two broad modes of algorithmic governance: control-oriented systems and support-oriented systems. The findings reveal a clear pattern of functional differentiation across levels of government. At the federal level, AI is most often institutionalized as a tool for high-stakes control: supporting surveillance, enforcement, and regulatory oversight. State governments occupy a more ambiguous middle ground, where AI frequently combines supportive functions with algorithmic gatekeeping, particularly in areas such as welfare administration and public health. Municipal governments, by contrast, tend to deploy AI in more pragmatic and service-oriented ways, using it to streamline everyday operations and improve direct interactions with residents. By foregrounding institutional context, this study advances debates on algorithmic governance by demonstrating that the character, function, and risks of AI in the public sector are fundamentally shaped by the level of governance at which these systems are deployed.

Nontobeko Gloria Maphuhla, Opeoluwa Oyehan Oyedeji

The massive production of municipal solid waste presents a significant global challenge for sustainable urban development and maintaining citizens’ quality of life, requiring effective management and disposal strategies. Waste-to-energy incineration technology has become increasingly important as a solution that simultaneously addresses the growing volumes of municipal solid waste and rising energy needs worldwide. This comprehensive review examines the research findings on the effectiveness of incineration as a waste-to-energy conversion method. The primary goal was to conduct a thorough systematic review assessing WtE incineration effectiveness across several key areas: energy recovery efficiency, waste volume reduction capabilities, environmental impact, and economic feasibility. A comprehensive literature search was conducted across ScienceDirect and additional pertinent databases, utilizing appropriate search terms in accordance with the PRISMA framework. A total of 431 studies were systematically identified, published between 2015 and 2025, and only 25 relevant studies were included in this review. Researchers collected data focusing on energy recovery percentages, volume reduction rates, emission reductions, and economic performance metrics. The findings revealed that every study included in the analysis showed positive results for WtE incineration across various performance measures. This research discovered the feasibility of generating electrical power from garbage through WtE incineration processes. The projected energy yields, ranging from gigawatt-hours to kilowatt-hours, were quantified for several nations, including Mexico (11,681.64 GWh), Cambodia (1625.81 GWh), Bangladesh (187.04 GWh), South Africa (6944 GWh), Iran (17,678 GWh), Nigeria (10,000 GWh), Indonesia (2487 MWh), Algeria (11.6 MWh), China (2316.52 MWh), Iraq (203.917 MWh), Uganda (774 kWh), and Pakistan (675 kWh). Energy recovery efficiency demonstrated a wide range from 30% to 92.75%, with waste volume reduction consistently reaching 90–95% levels, significantly prolonging landfill operational lifespans. From an environmental perspective, technology achieved greenhouse gas emission reductions ranging from 30% to 87%. This dual-purpose approach makes it an attractive, sustainable solution for both waste management and renewable energy production. By adopting this approach, cities can address waste and energy issues while boosting economic growth and job creation. However, it also involves substantial costs, technical difficulties, and environmental hazards that necessitate meticulous oversight.

Shuang Zhao, Sheng Yang, Qi Huang et al.

Amidst global climate change, carbon emissions across the edible vegetable oil supply chain are critical for sustainable development. This paper systematically reviews the existing literature, employing life cycle assessment (LCA) to analyze key factors influencing carbon footprints at stages including cultivation, processing, and transportation. It reveals the differential impacts of fertilizer application, energy structures, and regional policies. Unlike previous reviews that focus on single crops or regions, this study uniquely integrates global data across major edible oils, identifying three critical gaps: methodological inconsistency (60% of studies deviate from the requirements and guidelines for LCA); data imbalance (80% concentrated on soybean/rapeseed); weak policy-technical linkage. Key findings: fertilizer emissions dominate cultivation (40–60% of total footprint), while renewable energy substitution in processing reduces emissions by 35%. Future efforts should prioritize multidisciplinary integration, enhanced data infrastructure, and policy scenario analysis to provide scientific insights for the low-carbon transformation of the global edible oil industry.

Balakrishnan Varun Kumar, Sassi Rekik, Delmaria Richards et al.

The rapid growth and seasonal availability of agricultural materials, such as straws, stalks, husks, shells, and processing wastes, present both a disposal challenge and an opportunity for renewable fuel production. Solar-assisted thermochemical conversion, such as solar-driven pyrolysis, gasification, and hydrothermal routes, provides a pathway to produce bio-oils, syngas, and upgraded chars with substantially reduced fossil energy inputs compared to conventional thermal systems. Recent experimental research and plant-level techno-economic studies suggest that integrating concentrated solar thermal (CSP) collectors, falling particle receivers, or solar microwave hybrid heating with thermochemical reactors can reduce fossil auxiliary energy demand and enhance life-cycle greenhouse gas (GHG) performance. The primary challenges are operational intermittency and the capital costs of solar collectors. Alongside, machine learning (ML) and AI tools (surrogate models, Bayesian optimization, physics-informed neural networks) are accelerating feedstock screening, process control, and multi-objective optimization, significantly reducing experimental burden and improving the predictability of yields and emissions. This review presents recent experimental, modeling, and techno-economic literature to propose a unified classification of feedstocks, solar-integration modes, and AI roles. It reveals urgent research needs for standardized AI-ready datasets, long-term field demonstrations with thermal storage (e.g., integrating PCM), hybrid physics-ML models for interpretability, and region-specific TEA/LCA frameworks, which are most strongly recommended. Data’s reporting metrics and a reproducible dataset template are provided to accelerate translation from laboratory research to farm-level deployment.

David Bastos, Ricardo Infante Gomes, Diogo Gonçalves et al.

CO₂ emissions, a significant contributor to climate change, have spurred the exploration of sustainable solutions. One putative solution involves using recycled aggregates (RAs) from construction and demolition waste (CDW) to substitute natural sand in construction materials. This not only extends the life cycle of the waste but also reduces the use of natural resources. The potential to capture CO₂ in RAs presents a promising route to mitigate the environmental impact of the construction industry and contribute to its much anticipated decarbonization. This research takes a unique approach by investigating the incorporation of an amine-based additive—specifically 2-amino-2-methyl-1,3-propanediol (AMPD)—to enhance CO₂ capture into a real-case RA from recycling plants, transforming CDW with low carbon-capture potential into a highly reactive CO₂ capture material. Through TG analysis, FTIR-ATR and the combination of both (TG-FTIR), we were able to validate the use of RA materials as a support medium and quantify the CO₂ capture potential (12%) of the AMPD amine; a dual valorization was achieved: new value was added to low-quality CDW and we enhanced CO₂ sequestration, offering hope for a more sustainable future.

Nicola Kelly

There has been significant recent interest in layered solid-state materials containing an [M2O] square lattice layer (M = transition metal), particularly because [M2O] is the anti-type of the [CuO2] planes in the layered cuprate superconductors. In addition to the superconducting titanium oxypnictides, the [M2O] anti-cuprate layer also occurs in a wide range of layered oxychalcogenide compounds with M spanning early (Ti, V) to later transition metals (Mn, Co, Fe). The chalcogenide in question - which sandwiches the anti-cuprate layer - may be S, Se or Te, and in combination with a wide range of intervening "spacer" layers, many different structural families have been investigated. This review surveys the structures and physical properties of all these oxychalcogenide materials and relates these properties to their common anti-cuprate square lattice [M2O] layer. It is organised around the different oxidation states of the metal ion M, in order to explore the effects of the electronic configuration of M on the physical properties of each compound as a whole. A key part of the review highlights the use of soft-chemical modifications to alter physical properties of these materials, in the synthesis of novel van der Waals materials and other metastable compounds. Future avenues for these materials in the bulk, few- and single-layer limits are discussed.

Flor Miriam Plaza-del-Arco, Paul Röttger, Nino Scherrer et al.

Large language models (LLMs) are increasingly integrated into our daily lives and personalized. However, LLM personalization might also increase unintended side effects. Recent work suggests that persona prompting can lead models to falsely refuse user requests. However, no work has fully quantified the extent of this issue. To address this gap, we measure the impact of 15 sociodemographic personas (based on gender, race, religion, and disability) on false refusal. To control for other factors, we also test 16 different models, 3 tasks (Natural Language Inference, politeness, and offensiveness classification), and nine prompt paraphrases. We propose a Monte Carlo-based method to quantify this issue in a sample-efficient manner. Our results show that as models become more capable, personas impact the refusal rate less and less. Certain sociodemographic personas increase false refusal in some models, which suggests underlying biases in the alignment strategies or safety mechanisms. However, we find that the model choice and task significantly influence false refusals, especially in sensitive content tasks. Our findings suggest that persona effects have been overestimated, and might be due to other factors.

John Mavi, Diana Teodora Găitan, Sergio Coronado

Large Language Models (LLMs) are widely used across sectors, yet their alignment with International Humanitarian Law (IHL) is not well understood. This study evaluates eight leading LLMs on their ability to refuse prompts that explicitly violate these legal frameworks, focusing also on helpfulness - how clearly and constructively refusals are communicated. While most models rejected unlawful requests, the clarity and consistency of their responses varied. By revealing the model's rationale and referencing relevant legal or safety principles, explanatory refusals clarify the system's boundaries, reduce ambiguity, and help prevent misuse. A standardised system-level safety prompt significantly improved the quality of the explanations expressed within refusals in most models, highlighting the effectiveness of lightweight interventions. However, more complex prompts involving technical language or requests for code revealed ongoing vulnerabilities. These findings contribute to the development of safer, more transparent AI systems and propose a benchmark to evaluate the compliance of LLM with IHL.

Xiaorui Wu, Fei Li, Xiaofeng Mao et al.

Large language models (LLMs) frequently refuse to respond to pseudo-malicious instructions: semantically harmless input queries triggering unnecessary LLM refusals due to conservative safety alignment, significantly impairing user experience. Collecting such instructions is crucial for evaluating and mitigating over-refusals, but existing instruction curation methods, like manual creation or instruction rewriting, either lack scalability or fail to produce sufficiently diverse and effective refusal-inducing prompts. To address these limitations, we introduce EVOREFUSE, a prompt optimization approach that generates diverse pseudo-malicious instructions consistently eliciting confident refusals across LLMs. EVOREFUSE employs an evolutionary algorithm exploring the instruction space in more diverse directions than existing methods via mutation strategies and recombination, and iteratively evolves seed instructions to maximize evidence lower bound on LLM refusal probability. Using EVOREFUSE, we create two novel datasets: EVOREFUSE-TEST, a benchmark of 582 pseudo-malicious instructions that outperforms the next-best benchmark with 85.34% higher average refusal triggering rate across 9 LLMs without a safety-prior system prompt, 34.86% greater lexical diversity, and 40.03% improved LLM response confidence scores; and EVOREFUSE-ALIGN, which provides 3,000 pseudo-malicious instructions with responses for supervised and preference-based alignment training. With supervised fine-tuning on EVOREFUSE-ALIGN, LLAMA3.1-8B-INSTRUCT achieves up to 29.85% fewer over-refusals than models trained on the second-best alignment dataset, without compromising safety. Our analysis with EVOREFUSE-TEST reveals models trigger over-refusals by overly focusing on sensitive keywords while ignoring broader context. Our code and datasets are available at https://github.com/FishT0ucher/EVOREFUSE.

Diogo J. Paulo, João Martins, Hugo Proença et al.

Urban waste management remains a critical challenge for the development of smart cities. Despite the growing number of litter detection datasets, the problem of monitoring overflowing waste containers, particularly from images captured by garbage trucks, has received little attention. While existing datasets are valuable, they often lack annotations for specific container tracking or are captured in static, decontextualized environments, limiting their utility for real-world logistics. To address this gap, we present StreetView-Waste, a comprehensive dataset of urban scenes featuring litter and waste containers. The dataset supports three key evaluation tasks: (1) waste container detection, (2) waste container tracking, and (3) waste overflow segmentation. Alongside the dataset, we provide baselines for each task by benchmarking state-of-the-art models in object detection, tracking, and segmentation. Additionally, we enhance baseline performance by proposing two complementary strategies: a heuristic-based method for improved waste container tracking and a model-agnostic framework that leverages geometric priors to refine litter segmentation. Our experimental results show that while fine-tuned object detectors achieve reasonable performance in detecting waste containers, baseline tracking methods struggle to accurately estimate their number; however, our proposed heuristics reduce the mean absolute counting error by 79.6%. Similarly, while segmenting amorphous litter is challenging, our geometry-aware strategy improves segmentation mAP@0.5 by 27% on lightweight models, demonstrating the value of multimodal inputs for this task. Ultimately, StreetView-Waste provides a challenging benchmark to encourage research into real-world perception systems for urban waste management.

Shuzhou Yuan, Ercong Nie, Yinuo Sun et al.

Large language models (LLMs) frequently produce false refusals, declining benign requests that contain terms resembling unsafe queries. We address this challenge by introducing two comprehensive benchmarks: the Exaggerated Safety Benchmark (XSB) for single-turn prompts, annotated with "Focus" keywords that identify refusal-inducing triggers, and the Multi-turn Scenario-based Exaggerated Safety Benchmark (MS-XSB), which systematically evaluates refusal calibration in realistic, context-rich dialog settings. Our benchmarks reveal that exaggerated refusals persist across diverse recent LLMs and are especially pronounced in complex, multi-turn scenarios. To mitigate these failures, we leverage post-hoc explanation methods to identify refusal triggers and deploy three lightweight, model-agnostic approaches, ignore-word instructions, prompt rephrasing, and attention steering, at inference time, all without retraining or parameter access. Experiments on four instruction-tuned Llama models demonstrate that these strategies substantially improve compliance on safe prompts while maintaining robust safety protections. Our findings establish a reproducible framework for diagnosing and mitigating exaggerated refusals, highlighting practical pathways to safer and more helpful LLM deployments.

Rosa Hernández-López, Aurelio López-Malo, Ricardo Navarro-Amador et al.

Sugarcane bagasse (SCB) is a waste product from Mexico’s sugar industry that is generally burned or discarded. It contains around 48% cellulose, representing a significant source of this component from industrial waste. Eugenol is found in clove oil; it has been used for its medicinal and antimicrobial benefits in the food and pharmaceutical industries. This study aims to develop a filtering material using sugarcane bagasse (SCB) and encapsulated eugenol as an antimicrobial agent. The study involves extracting cellulose from SCB using alkaline hydrolysis with ultrasound, followed by forming composite materials encapsulated in alginate with eugenol concentrations from 0 to 1% <i>v</i>/<i>v</i>. These materials were characterized and tested for antimicrobial efficacy. The findings indicate that the cellulose–eugenol–alginate composite displays high eugenol encapsulation efficiency and effective short-term release. In well-diffusion assays, the material showed inhibition halos up to 20.47 mm against <i>S. aureus</i>, suggesting its potential as an eco-friendly alternative to traditional antimicrobial agents in filter materials.

Nobuki Morita, Yo Toma, Hideto Ueno

The disposal of tea leaves discarded in the tea beverage market and clinker from coal-fired power plants has an impact on the environment; however, there are no reported cases of their combination for composting. Therefore, this study evaluated the effect of adding clinker from a coal-fired power plant to compost based on tea leaves, an organic waste product, on the composting rate and quality. The tea leaves-only compost was designated as Clinker 0%, and composts with 20% (<i>w</i>/<i>w</i>), 40% (<i>w</i>/<i>w</i>), and 60% (<i>w</i>/<i>w</i>) tea leaves supplemented with clinker were designated as Clinker 20, 40, and 60%, respectively. Each mixed material was placed in a 35 L polypropylene container with a lid and allowed to compost for 95 days. The composting rate was evaluated by the chemical oxygen demand (COD) in hot water extract and plant tests using juvenile komatsuna (<i>Brassica rapa</i> var. <i>perviridis</i>). The addition of clinker reduced the COD at the beginning of composting by 52.0, 74.3, and 86.7% in Clinker 20, 40, and 60%, respectively, compared to Clinker 0%. Furthermore, root elongation one month after composting was inhibited by Clinker 0% (60.1% relative to distilled water), but not by the addition of clinker (91.7–102.7% relative to distilled water). This suggests that the addition of clinker to tea leaf compost may accelerate composting.

Jesús Cisneros-Aguirre, María Afonso-Correa

This study examines recovery efforts at Gran Tarajal Harbor following a significant oil spill, employing a combination of innovative technologies tailored to enhance oil spill remediation. Cleanup operations incorporated advanced absorbent sponges with high reusability, absorbent granulates for targeted hydrocarbon capture, bioremediation techniques using allochthonous microorganisms to accelerate natural degradation processes, and the integration of newly designed oil containment barriers coupled with sponges. These technologies were instrumental in effectively mitigating environmental damage, as evidenced by a reduction in hydrocarbon concentrations in sediments from nearly 60,000 mg/kg to under 1600 mg/kg within seven months. Notably, advanced absorbent sponges demonstrated superior capacity for repeated use, optimizing the cleanup process and contributing to the sustainability of the response efforts. The most important finding of this research is the demonstrated efficacy of integrated approach in not only reducing hydrocarbon contamination but also in promoting ecological recovery. Heavy metal analyses revealed that lead and copper concentrations were primarily associated with routine port activities, while mercury levels, attributed to the spill, decreased significantly over time. Tissue analysis of local organisms showed minimal contamination, and assessments of biological communities indicated signs of ecological recovery. This work highlights the necessity of introduce new disruptive technologies in contingency plans.

Simon Lermen, Mateusz Dziemian, Govind Pimpale

Recently, language models like Llama 3.1 Instruct have become increasingly capable of agentic behavior, enabling them to perform tasks requiring short-term planning and tool use. In this study, we apply refusal-vector ablation to Llama 3.1 70B and implement a simple agent scaffolding to create an unrestricted agent. Our findings imply that these refusal-vector ablated models can successfully complete harmful tasks, such as bribing officials or crafting phishing attacks, revealing significant vulnerabilities in current safety mechanisms. To further explore this, we introduce a small Safe Agent Benchmark, designed to test both harmful and benign tasks in agentic scenarios. Our results imply that safety fine-tuning in chat models does not generalize well to agentic behavior, as we find that Llama 3.1 Instruct models are willing to perform most harmful tasks without modifications. At the same time, these models will refuse to give advice on how to perform the same tasks when asked for a chat completion. This highlights the growing risk of misuse as models become more capable, underscoring the need for improved safety frameworks for language model agents.

Alexander von Recum, Christoph Schnabl, Gabor Hollbeck et al.

Refusals - instances where large language models (LLMs) decline or fail to fully execute user instructions - are crucial for both AI safety and AI capabilities and the reduction of hallucinations in particular. These behaviors are learned during post-training, especially in instruction fine-tuning (IFT) and reinforcement learning from human feedback (RLHF). However, existing taxonomies and evaluation datasets for refusals are inadequate, often focusing solely on should-not-related (instead of cannot-related) categories, and lacking tools for auditing refusal content in black-box LLM outputs. We present a comprehensive framework for classifying LLM refusals: (a) a taxonomy of 16 refusal categories, (b) a human-annotated dataset of over 8,600 instances from publicly available IFT and RLHF datasets, (c) a synthetic dataset with 8,000 examples for each refusal category, and (d) classifiers trained for refusal classification. Our work enables precise auditing of refusal behaviors in black-box LLMs and automatic analyses of refusal patterns in large IFT and RLHF datasets. This facilitates the strategic adjustment of LLM refusals, contributing to the development of more safe and reliable LLMs.

Jung Eun Park, Sang Eun Lee, Seokhwi Kim

The oyster shell is a valuable calcium resource; however, its application is limited by its high NaCl content. Therefore, to establish the use of oyster shells as a viable resource, conditional experiments were conducted to select optimum parameters for NaCl removal. For this purpose, we compared leaching methods with batch and sequential procedures, determined the volume of water used for washing, and evaluated the mixing speed. The batch system removed NaCl when washed for >24 h over a shell to water ratio of 1:5. Results from the batch experiments confirmed that washing twice can completely remove NaCl from the shells on a like-for-like basis. Additionally, the efficiency of washing was sequentially evaluated in terms of the number of washing cycles. Compared to batch experiments, continuous washing could remove NaCl in approximately 10 min at a shell to water ratio of 1:4. We found that regardless of the washing methods, the volume of water used for washing is key for enhancing NaCl removal. Consequently, increasing the volume of water used for washing coupled with a proper sorting of fine particles can help enhance the purity of calcium, which will enable the use of oyster shell as an alternate Ca-resource.

Genevieve Soon, Hui Zhang, Adrian Wing-Keung Law et al.

The formulation of the CFD-DEM model, CD-MELT, is established in this study to include three-phase non-isothermal processes with simultaneous combustion and melting for gasification simulations. To demonstrate the model capability, CD-MELT is used to assess the need for slag recycling for the non-isothermal melting of municipal solid wastes (MSW) in a prototype waste-to-energy research facility. The simulation encompasses the full fixed-bed slagging gasification process, including chemical reactions and melting of MSW and slag. In order to assess the need for slag recycling, comparisons are made for the two cases of with and without, in terms of the slag mass, liquid slag volume fraction, exit gas composition, and temperature distribution in the gasifier. The prediction results enable the tracking of liquid molten slag as it permeates through the solids-packed bed for the first time in the literature as far as we are aware, which is crucial to address design considerations such as distribution of bed temperature and optimal location for slag-tap holes at the bottom, as well as potential slag clogging within the porous media. The model also predicts an uneven and intermittent slag permeation through the packed bed without the recycling, and provides a plausible explanation for the operators’ experience of why slag recycling is important for process stability. Finally, the predicted slag outlet temperature using the proposed CFD approach also agrees well with the measurement data published in an earlier case study for the same facility.

Modeste Kameni Nematchoua, José A. Orosa

Nowadays, the methodology aiming to achieve carbon neutrality and net zero energy on a large scale is known. Despite this, few specialists are mastering this technology globally. What new scenarios. applied at the neighbourhood scale. generate a significant reduction in the rate of CO₂ emissions and energy demand? In addition, a lack of massive, regular, and consistent data on carbon emissions and energy consumption has made it significantly difficult to understand the origins of climate change at the building and neighbourhood scales. This work has, as its main goal, the assessment of different strategies that facilitate reduction in the concentration of CO₂ and lower energy demands at the district level. The life cycle assessment of a new district has been carried out over 100 years during the four stages of the life cycle of the neighbourhood (construction, operation, demolition and end of life). The results showed that up to 93% of greenhouse gas (GHG) was produced during the operational stage. The energy demand due to transport and waste management represented 60% of the total energy demand of the district during the operational stage. There is still a possibility to maintain air temperature growth around 1.5 °C in the next decade by means of the following: Global warming + 100% of heavy renovation of all buildings + 100% of electric car − renewable energy. This strategy would facilitate a reduction of over 92% of the CO₂ produced at the district level.