Hasil untuk "Municipal refuse. Solid wastes"

Ningjing Tang, Alice Qian, Qiaosi Wang et al.

Content Warning: This paper contains participant quotes and discussions related to mental health challenges, emotional distress, and suicidal ideation. Large language models (LLMs) are increasingly used for mental health support, yet the model safeguards -- particularly refusals to engage with sensitive content -- remain poorly understood from the perspectives of users and mental health professionals (MHPs) and have been reported to cause real-world harms. This paper presents findings from a sequential mixed-methods study examining how LLM refusals are experienced and interpreted in mental health support interactions. Through surveys (N=53) and in-depth interviews (N=16) with individuals using LLMs for mental health support and MHPs, we reveal that refusals are not isolated, single-turn system behaviors, but rather constitute dynamic, multi-phase experiences: pre-refusal expectation formation, refusal triggering and encounter, refusal message framing, resource referral provision, and post-refusal outcomes. We contribute a multi-phase framework for evaluating refusals beyond binary policy compliance accuracy and design recommendations for future refusal mechanisms. These findings suggest that understanding LLM refusals requires moving beyond single-turn interactions toward recognizing them as holistic experiential processes embedded within the entire LLM design pipeline and the broader realities of mental health access.

Peeyush Phogat

This review explores the potential of waste-derived fuels (WDF) and refuse-derived fuels (RDF) as sustainable alternatives to conventional fossil fuels, addressing both energy recovery and waste management challenges. WDF and RDF are generated from diverse feedstocks—such as municipal solid waste (MSW), industrial by-products, and biomass—via mechanical, thermochemical, and biochemical processing. These fuels can be utilized in cement kilns, power plants, and industrial furnaces, contributing to reduced fossil fuel consumption. The manuscript classifies WDF and RDF into solid, liquid, and gaseous forms, examining their energy characteristics and combustion properties in comparison to coal, diesel, and natural gas. A detailed analysis of calorific value, moisture and ash content, and emissions profiles is provided to assess their performance. The environmental benefits, including decreased landfill usage, methane mitigation, and lower greenhouse gas emissions, are emphasized through life cycle assessments (LCA). Economic aspects such as production costs, energy substitution potential, and global market adoption trends are also discussed. Technological pathways including shredding, drying, gasification, pyrolysis, and anaerobic digestion are analyzed for their efficiency, capital investment requirements, and environmental impacts. The review also highlights current challenges—such as feedstock heterogeneity, emissions control, and public perception—and outlines future prospects enabled by technological advancements, regulatory support, and integration into circular economy frameworks. This comprehensive evaluation positions WDF and RDF as viable contributors to a low-carbon energy future, offering pathways for sustainable waste valorization and renewable energy generation.

Mahavir Dabas, Si Chen, Charles Fleming et al.

Safety alignment is crucial for large language models (LLMs) to resist malicious instructions but often results in over-refusals, where benign prompts are unnecessarily rejected, impairing user experience and model utility. We introduce ACTOR (Activation-Based Training for Over-Refusal Reduction), a robust and compute- and data-efficient training framework that minimizes over-refusals by leveraging internal activation patterns from diverse queries. ACTOR precisely identifies and adjusts the activation components that trigger refusals, providing stronger control over the refusal mechanism. By fine-tuning only a single model layer, ACTOR effectively reduces over-refusals across multiple benchmarks while maintaining the model's ability to handle harmful queries and preserve overall utility.

Begoña González, Diego Rossit, Mariano Frutos et al.

Few activities are as crucial in urban environments as waste management. Mismanagement of waste can cause significant economic, social, and environmental damage. However, waste management is often a complex system to manage and therefore where computational decision-support tools can play a pivotal role in assisting managers to make faster and better decisions. In this sense, this article proposes, on the one hand, a unified optimization model to address two common waste management system optimization problem: the determination of the capacity of waste bins in the collection network and the design and scheduling of collection routes. The integration of these two problems is not usual in the literature since each of them separately is already a major computational challenge. On the other hand, two improved exact formulations based on mathematical programming and a genetic algorithm (GA) are provided to solve this proposed unified optimization model. It should be noted that the GA considers a mixed chromosome representation of the solutions combining binary and integer alleles, in order to solve realistic instances of this complex problem. Also, different genetic operators have been tested to study which combination of them obtained better results in execution times on the order of that of the exact solvers. The obtained results show that the proposed GA is able to match the results of exact solvers on small instances and, in addition, can obtain feasible solutions on large instances, where exact formulations are not applicable, in reasonable computation times.

Nirmalendu Prakash, Yeo Wei Jie, Amir Abdullah et al.

Refusal on harmful prompts is a key safety behaviour in instruction-tuned large language models (LLMs), yet the internal causes of this behaviour remain poorly understood. We study two public instruction-tuned models, Gemma-2-2B-IT and LLaMA-3.1-8B-IT, using sparse autoencoders (SAEs) trained on residual-stream activations. Given a harmful prompt, we search the SAE latent space for feature sets whose ablation flips the model from refusal to compliance, demonstrating causal influence and creating a jailbreak. Our search proceeds in three stages: (1) Refusal Direction: find a refusal-mediating direction and collect SAE features near that direction; (2) Greedy Filtering: prune to a minimal set; and (3) Interaction Discovery: fit a factorization machine (FM) that captures nonlinear interactions among the remaining active features and the minimal set. This pipeline yields a broad set of jailbreak-critical features, offering insight into the mechanistic basis of refusal. Moreover, we find evidence of redundant features that remain dormant unless earlier features are suppressed. Our findings highlight the potential for fine-grained auditing and targeted intervention in safety behaviours by manipulating the interpretable latent space.

Andrea Marelli, Luca Magri, Federica Arrigoni et al.

In industrial settings, weakly supervised (WS) methods are usually preferred over their fully supervised (FS) counterparts as they do not require costly manual annotations. Unfortunately, the segmentation masks obtained in the WS regime are typically poor in terms of accuracy. In this work, we present a WS method capable of producing accurate masks for semantic segmentation in the case of video streams. More specifically, we build saliency maps that exploit the temporal coherence between consecutive frames in a video, promoting consistency when objects appear in different frames. We apply our method in a waste-sorting scenario, where we perform weakly supervised video segmentation (WSVS) by training an auxiliary classifier that distinguishes between videos recorded before and after a human operator, who manually removes specific wastes from a conveyor belt. The saliency maps of this classifier identify materials to be removed, and we modify the classifier training to minimize differences between the saliency map of a central frame and those in adjacent frames, after having compensated object displacement. Experiments on a real-world dataset demonstrate the benefits of integrating temporal coherence directly during the training phase of the classifier. Code and dataset are available upon request.

Yuanbo Xie, Yingjie Zhang, Tianyun Liu et al.

Jailbreak attacks pose persistent threats to large language models (LLMs). Current safety alignment methods have attempted to address these issues, but they experience two significant limitations: insufficient safety alignment depth and unrobust internal defense mechanisms. These limitations make them vulnerable to adversarial attacks such as prefilling and refusal direction manipulation. We introduce DeepRefusal, a robust safety alignment framework that overcomes these issues. DeepRefusal forces the model to dynamically rebuild its refusal mechanisms from jailbreak states. This is achieved by probabilistically ablating the refusal direction across layers and token depths during fine-tuning. Our method not only defends against prefilling and refusal direction attacks but also demonstrates strong resilience against other unseen jailbreak strategies. Extensive evaluations on four open-source LLM families and six representative attacks show that DeepRefusal reduces attack success rates by approximately 95%, while maintaining model capabilities with minimal performance degradation.

Tuan T. Nguyen, John Le, Thai T. Vu et al.

Large language models (LLMs) achieve impressive performance across diverse tasks yet remain vulnerable to jailbreak attacks that bypass safety mechanisms. We present RAID (Refusal-Aware and Integrated Decoding), a framework that systematically probes these weaknesses by crafting adversarial suffixes that induce restricted content while preserving fluency. RAID relaxes discrete tokens into continuous embeddings and optimizes them with a joint objective that (i) encourages restricted responses, (ii) incorporates a refusal-aware regularizer to steer activations away from refusal directions in embedding space, and (iii) applies a coherence term to maintain semantic plausibility and non-redundancy. After optimization, a critic-guided decoding procedure maps embeddings back to tokens by balancing embedding affinity with language-model likelihood. This integration yields suffixes that are both effective in bypassing defenses and natural in form. Experiments on multiple open-source LLMs show that RAID achieves higher attack success rates with fewer queries and lower computational cost than recent white-box and black-box baselines. These findings highlight the importance of embedding-space regularization for understanding and mitigating LLM jailbreak vulnerabilities.

Utsav Sharma, Dayanand Sharma, Amit Kumar et al.

Urbanization and population growth in India have quickened, leading to an annual generation of around 62 million tonnes of municipal solid waste (MSW). Improper management of organic waste presents a major environmental problem due to air and water pollution, soil contamination and greenhouse gas production. This research aims to develop refuse-derived fuel (RDF) as a viable option, converting waste into a high-calorific energy carrier for industrial use. The RDF samples were collected from five strategic locations in Uttar Pradesh: Morta Site, Pipeline Site, and Sector 146 Noida, covering various waste compositions found at these landfill sites. Proximate and ultimate analyses of the RDF prepared from these sources were conducted, followed by in-depth Thermogravimetric Analysis (TGA) to validate its suitability as a potential feedstock. Careful waste segregation and treatment for better fuel quality can help minimize the difference in calorific values between different sites. Based on RDF tests, the waste-to-energy technology can divert over 30 % of solid waste from landfills and cut greenhouse gas emissions by as much as 25 % compared to traditional disposal methods. Unlike RDF, which is part of the replacement line for coal in industrial furnaces such as thermal power plants, it eliminates over 15 % and 20 % of sulfur dioxide (SO2) and nitrogen oxides (NOx). Ensuring that RDFs support sustainable energy technologies and align with circular economy principles, the study's results could enhance energy efficiency in waste management and complement environmental policy goals across all states in India and worldwide.

Bruce W. Lee, Inkit Padhi, Karthikeyan Natesan Ramamurthy et al.

LLMs have shown remarkable capabilities, but precisely controlling their response behavior remains challenging. Existing activation steering methods alter LLM behavior indiscriminately, limiting their practical applicability in settings where selective responses are essential, such as content moderation or domain-specific assistants. In this paper, we propose Conditional Activation Steering (CAST), which analyzes LLM activation patterns during inference to selectively apply or withhold activation steering based on the input context. Our method is based on the observation that different categories of prompts activate distinct patterns in the model's hidden states. Using CAST, one can systematically control LLM behavior with rules like "if input is about hate speech or adult content, then refuse" or "if input is not about legal advice, then refuse." This allows for selective modification of responses to specific content while maintaining normal responses to other content, all without requiring weight optimization. We release an open-source implementation of our framework at github.com/IBM/activation-steering .

Shangyu Wu, Ying Xiong, Yufei Cui et al.

Retrieval-based augmentations (RA) incorporating knowledge from an external database into language models have greatly succeeded in various knowledge-intensive (KI) tasks. However, integrating retrievals in non-knowledge-intensive (NKI) tasks is still challenging. Existing works focus on concatenating retrievals with inputs to improve model performance. Unfortunately, the use of retrieval concatenation-based augmentations causes an increase in the input length, substantially raising the computational demands of attention mechanisms. This paper proposes a new paradigm of RA named \textbf{ReFusion}, a computation-efficient Retrieval representation Fusion with bi-level optimization. Unlike previous works, ReFusion directly fuses the retrieval representations into the hidden states of models. Specifically, ReFusion leverages an adaptive retrieval integrator to seek the optimal combination of the proposed ranking schemes across different model layers. Experimental results demonstrate that the proposed ReFusion can achieve superior and robust performance in various NKI tasks.

Suman Kunwar

Computer vision methods have shown to be effective in classifying garbage into recycling categories for waste processing, existing methods are costly, imprecise, and unclear. To tackle this issue, we introduce MWaste, a mobile application that uses computer vision and deep learning techniques to classify waste materials as trash, plastic, paper, metal, glass or cardboard. Its effectiveness was tested on various neural network architectures and real-world images, achieving an average precision of 92\% on the test set. This app can help combat climate change by enabling efficient waste processing and reducing the generation of greenhouse gases caused by incorrect waste disposal.

S. Gerassimidou, C. Velis, Paul T. Williams et al.

Thermogravimetric analysis (TGA) is the most widespread thermal analytical technique applied to waste materials. By way of critical review, we establish a theoretical framework for the use of TGA under non-isothermal conditions for compositional analysis of waste-derived fuels from municipal solid waste (MSW) (solid recovered fuel (SRF), or refuse-derived fuel (RDF)). Thermal behaviour of SRF/RDF is described as a complex mixture of several components at multiple levels (including an assembly of prevalent waste items, materials, and chemical compounds); and, operating conditions applied to TGA experiments of SRF/RDF are summarised. SRF/RDF mainly contains cellulose, hemicellulose, lignin, polyethylene, polypropylene, and polyethylene terephthalate. Polyvinyl chloride is also used in simulated samples, for its high chlorine content. We discuss the main limitations for TGA-based compositional analysis of SRF/RDF, due to inherently heterogeneous composition of MSW at multiple levels, overlapping degradation areas, and potential interaction effects among waste components and cross-contamination. Optimal generic TGA settings are highlighted (inert atmosphere and low heating rate (⩽10°C), sufficient temperature range for material degradation (⩾750°C), and representative amount of test portion). There is high potential to develop TGA-based composition identification and wider quality assurance and control methods using advanced thermo-analytical techniques (e.g. TGA with evolved gas analysis), coupled with statistical data analytics.

Laraib Ehtasham

Municipal solid waste (MSW), including refuse and non-hazardous waste, is generated mainly from households. This article adopts the approach of describing the MSW collection systems in three Asian countries that are Singapore, Mongolia, and Nepal; by reviewing the literature related to this particular subject. Therefore, by going through the literature, it has been found that the authorities responsible for waste collection are: Public Waste Collectors (PWCs) appointed by National Environment Agency (NEA) in Singapore, the municipal government, and Tohijilt Uilchilgeenii Kompani (TKUs) in Mongolia, and 270 municipalities in Nepal. Along with this, the waste collection fee in Singapore is paid to Singapore Power (combined with utility bills); in Mongolia to TKUs (based on their performance); and in Nepal, it is fixed on ad hoc basis. Overall, based on the practices of waste collection discussed in this article, Singapore has a better waste collection system than that Mongolia and Nepal.

I. M. W. Widyarsana, D. Saraswati

Municipal solid waste in Indonesia has a high potential to be used as an alternative energy source. One of the methods is by producing refuse-derived fuel (RDF) as coal substitution. Pre-treatment is needed to reduce high water content in domestic waste, especially in Indonesia, to produce high-quality RDF. This study compares and analyzes the optimal pre-treatment method to produce RDF from domestic waste based on each RDF product characteristic. The research reviewed and analyzed the data of RDF characteristics of various pre-treatment methods applied in Indonesia from pilot-scale experiments as primary data and previous research as secondary data. The methods compared are fermentation method, sun-drying, and bio drying. RDF from each method was tested to collect the proximate and ultimate characteristics data. Statistical analysis of RDF characteristics is carried out to determine whether there is a significant difference from different pre-treatment methods. This analysis shows no significant effect on the difference in the characteristics of the RDF. RDF from bio drying provides the highest average number of calorific values and lowest moisture content. The study concluded that the bio drying method is recommended in more waste treatment facilities in Indonesia to increase waste as an alternative energy resource.

Nguyen Huu Hoang, C. Fogarassy

According to decision no. 491/QD-TTg signed in 2018 by the Vietnamese Prime Minister approving adjustments to the national strategy for the general management of solid waste until 2025 with a vision toward 2050, Vietnam has committed itself to move toward collecting, transporting, and treating 100% of non-household waste by 2025 and 85% of waste discharged by households by 2025. This paper aims to determine which is the best sustainable solid waste management system out of those that have been formulated by World Bank experts for Hanoi until 2030 for implementing the national strategy. The paper compares four distinct solid waste management enhancement alternatives, namely, “Improving the current system for waste collection and transportation”; “Reducing, reusing, and recycling waste at source”; “Mechanical–biological treatment (MBT) plants for classifying, composting, and refuse-derived fuel (RDF) for the cement industry”; and “MBT plants for classifying, composting, and RDF for waste-to-energy/incineration plants”. The comparison was made using an analytic hierarchy process. As a result, the research indicated that “MBT plants for classifying, composting, and RDF for waste-to-energy/incineration plants” has the highest ranking in terms of a sustainable solution for the municipal solid waste management system. Therefore, it should be applied for managing the current situation in Hanoi. At the same time, the sustainable development of the system must seek to decrease the waste-to-energy ratio continuously and significantly through the planned reuse of materials that can be recycled to industry. According to the literature, in major cities in Asia and Africa, development programs are moving toward waste-to-energy solutions. The EU’s circular innovation programs and action plan may be in the opposite direction to this trend.

Isaac Slavitt

Lead pipe remediation budgets are limited and ought to maximize public health impact. This goal implies a non-trivial optimization problem; lead service lines connect water mains to individual houses, but any realistic replacement strategy must batch replacements at a larger scale. Additionally, planners typically lack a principled method for comparing the relative public health value of potential interventions and often plan projects based on non-health factors. This paper describes a simple process for estimating child health impact at a parcel level by cleaning and synthesizing municipal datasets that are commonly available but seldom joined due to data quality issues. Using geocoding as the core record linkage mechanism, parcel-level toxicity data can be combined with school enrollment records to indicate where young children and lead lines coexist. A harm metric of estimated exposure-years is described at the parcel level, which can then be aggregated to the project level and minimized globally by posing project selection as a 0/1 knapsack problem. Simplifying further for use by non-experts, the implied linear programming relaxation is solved intuitively with the greedy algorithm; ordering projects by benefit cost ratio produces a priority list which planners can then consider holistically alongside harder to quantify factors. A case study demonstrates the successful application of this framework to a small U.S. city's existing data to prioritize federal infrastructure funding. While this paper focuses on lead in drinking water, the approach readily generalizes to other sources of residential toxicity with disproportionate impact on children.

V. Fischer, M. Pacheco Paneque, A. Legrain et al.

In most Swiss municipalities, a curbside system consisting of heavy trucks stopping at almost each household is used for non-recoverable waste collection. Due to the many stops of the trucks, this strategy causes high fuel consumption, emissions and noise. These effects can be alleviated by reducing the number of stops performed by collection vehicles. One possibility consists of locating collection points throughout the municipality such that residents bring their waste to their most preferred location. The optimization problem consists of selecting a subset of candidate locations to place the points such that each household disposes the waste at the most preferred location. Provided that the underlying road network is available, we refer to this optimization problem as the capacitated multi-vehicle covering tour problem on a road network (Cm-CTP-R). We introduce two mixed-integer linear programming (MILP) formulations: a road-network-based formulation that exploits the sparsity of the network and a customer-based formulation typically used in vehicle routing problems (VRP). To solve large instances, we propose a two-phased heuristic approach that addresses the two subproblems the Cm-CTP-R is built on: a set covering problem to select the locations and a split-delivery VRP to determine the routes. Computational experiments on both small and real-life instances show that the road-network-based formulation is better suited. Furthermore, the proposed heuristic provides good solutions with optimality gaps below 0.5% and 3.5% for 75% of the small and real-life instances respectively and is able to find better solutions than the exact method for many real-life instances.

Jash Shah, Sagar Kamat

Segregation of garbage is a primary concern in many nations across the world. Even though we are in the modern era, many people still do not know how to distinguish between organic and recyclable waste. It is because of this that the world is facing a major crisis of waste disposal. In this paper, we try to use deep learning algorithms to help solve this problem of waste classification. The waste is classified into two categories like organic and recyclable. Our proposed model achieves an accuracy of 94.9%. Although the other two models also show promising results, the Proposed Model stands out with the greatest accuracy. With the help of deep learning, one of the greatest obstacles to efficient waste management can finally be removed.

M. M. Azis, Jonas Kristanto, C. Purnomo

Municipal solid waste (MSW) processing is still problematic in Indonesia. From the hierarchy of waste management, it is clear that energy recovery from waste could be an option after prevention and the 5R (rethink, refuse, reduce, reuse, recycle) processes. The Presidential Regulation No 35/2018 mandated the acceleration of waste-to-energy (WtE) plant adoption in Indonesia. The present study aimed to demonstrate a techno-economic evaluation of a commercial WtE plant in Indonesia by processing 1000 tons of waste/day to produce ca. 19.7 MW of electricity. The WtE electricity price is set at USD 13.35 cent/kWh, which is already higher than the average household price at USD 9.76 cent/kWh. The capital investment is estimated at USD 102.2 million. The annual operational cost is estimated at USD 12.1 million and the annual revenue at USD 41.6 million. At this value, the internal rate of return (IRR) for the WtE plant is 25.32% with a payout time (PoT) of 3.47 years. In addition, this study also takes into account electricity price sales, tipping fee, and pretreatment cost of waste. The result of a sensitivity analysis showed that the electricity price was the most sensitive factor. This study reveals that it is important to maintain a regulated electricity price to ensure the sustainability of the WtE plant in Indonesia.