Faced with the dual constraints of resources and the environment, green technology innovation has become an important measure to solve the development challenges of heavily polluting industries. From the perspective of institutional regulation theory, this paper studies the impact of direct environmental regulation on green technology innovation in Chinese listed companies of heavily polluting industries by using the Panel Poisson fixed effect model. Besides, the heterogeneity of ownership and industry is discussed. The results indicate that direct environmental regulations exert a strong and significant incentive effect on green technology innovations in heavily polluting industries. Regarding the heterogeneity of enterprise ownership, direct environmental regulations were found to be more significant to the green technology innovations of state-owned listed companies in such industries. Considering industry heterogeneity, compared with labor-resource intensive industries, direct environmental regulation can effectively encourage green technology innovations in technology-capital intensive industries. This study provides a policy basis for promoting environmental governance and green technology innovation in China's heavily polluting industries.
Opioid use by pregnant women represents a significant public health concern given the association of opioid exposure and adverse maternal and neonatal outcomes, including preterm labor, stillbirth, neonatal abstinence syndrome, and maternal mortality (1,2). State-level actions are critical to curbing the opioid epidemic through programs and policies to reduce use of prescription opioids and illegal opioids including heroin and illicitly manufactured fentanyl, both of which contribute to the epidemic (3). Hospital discharge data from the 1999–2014 Healthcare Cost and Utilization Project (HCUP) were analyzed to describe U.S. national and state-specific trends in opioid use disorder documented at delivery hospitalization. Nationally, the prevalence of opioid use disorder more than quadrupled during 1999–2014 (from 1.5 per 1,000 delivery hospitalizations to 6.5; p<0.05). Increasing trends over time were observed in all 28 states with available data (p<0.05). In 2014, prevalence ranged from 0.7 in the District of Columbia (DC) to 48.6 in Vermont. Continued national, state, and provider efforts to prevent, monitor, and treat opioid use disorder among reproductive-aged and pregnant women are needed. Efforts might include improved access to data in Prescription Drug Monitoring Programs, increased substance abuse screening, use of medication-assisted therapy, and substance abuse treatment referrals.
Although recent advancements in large language models (LLMs) have significantly improved their performance on various tasks, they still face challenges with complex and symbolic multi-step reasoning, particularly in mathematical reasoning. To bolster the mathematical reasoning capabilities of LLMs, most existing efforts concentrate on seeking assistance from either domain experts or GPT-4 for high-quality process-supervised data, which is not only expensive but also labor-intensive. In our study, we propose an innovative framework, AlphaMath, that bypasses the need for process annotations (from humans or GPTs) by leveraging Monte Carlo Tree Search (MCTS). This framework focuses on unleashing the potential of a well-pretrained LLM to autonomously enhance its mathematical reasoning. Specifically, we integrate a value model with the LLM, automatically generating both process supervision and step-level evaluation signals in MCTS. Furthermore, we propose an efficient inference strategy, step-level beam search, where the value model is crafted to assist the policy model (i.e., LLM) in navigating more effective reasoning paths, rather than solely relying on prior probabilities. The experimental results on both in-domain and out-of-domain datasets demonstrate that even without GPT-4 or human-annotated process supervision, our AlphaMath framework achieves comparable or superior results to previous state-of-the-art methods.
Deep reinforcement learning has proven to be a powerful approach to solving control tasks, but its characteristic high-frequency oscillations make it difficult to apply in real-world environments. While prior methods have addressed action oscillations via architectural or loss-based methods, the latter typically depend on heuristic or synthetic definitions of state similarity to promote action consistency, which often fail to accurately reflect the underlying system dynamics. In this paper, we propose a novel loss-based method by introducing a transition-induced similar state. The transition-induced similar state is defined as the distribution of next states transitioned from the previous state. Since it utilizes only environmental feedback and actually collected data, it better captures system dynamics. Building upon this foundation, we introduce Action Smoothing by Aligning Actions with Predictions from Preceding States (ASAP), an action smoothing method that effectively mitigates action oscillations. ASAP enforces action smoothness by aligning the actions with those taken in transition-induced similar states and by penalizing second-order differences to suppress high-frequency oscillations. Experiments in Gymnasium and Isaac-Lab environments demonstrate that ASAP yields smoother control and improved policy performance over existing methods.
This article explores the implementation of the One Channel System (OCS) in Indonesia-Malaysia bilateral labor migration policy through the lens of constructivism. This study examines how international norms on migrant worker protection, particularly those outlined in ILO conventions, are internalized in migration policies, as well as key challenges in the effective implementation of OCS, such as institutional coordination, economic interests, and technological barriers. Using a qualitative approach, the research analyzes the tension between normative commitments and material considerations faced by both countries. The findings show that while Indonesia prioritizes the protection of its migrant workers, Malaysia's reliance on informal recruitment mechanisms like the Maid Online System (SMO) weakens the achievement of OCS goals. This duality highlights the limitations in norm socialization and the challenges of aligning global labor standards with domestic realities. This study contributes to the understanding of constructivist theory in migration policy by emphasizing the intersection of international norms, state identity, and pragmatic realities. The significance of this research lies in its ability to provide policy recommendations to improve the implementation of OCS and its contribution to the development of constructivism in migration policy, showing how international norms and state identity interact in labor migration governance.
Evan Shieh, Faye-Marie Vassel, Cassidy R. Sugimoto
et al.
The widespread deployment of generative language models (LMs) is raising concerns about societal harms. Despite this, studies of bias in generative LMs, including attempted self-audits by LM developers, have thus far been conducted in limited contexts. To address this gap, this study examines representational harms in synthetic texts produced by leading language models in response to open-ended creative writing prompts based in the United States. We conduct our investigation on 500,000 synthetic texts generated by five publicly available generative language models: ChatGPT 3.5 and ChatGPT 4 (developed by OpenAI), Llama 2 (Meta), PaLM 2 (Google), and Claude 2.0 (Anthropic). We base our selection of models on both the sizable amount of funding wielded by these companies and their investors (on the order of tens of billions in USD), as well as the prominent policy roles that each company has played on the federal level. At the time of data collection (from August 16th to November 7th, 2023), the selected models were considered state-of-the-art for each company. Creative writing prompts reflect three domains of life set in the United States: classroom interactions (“Learning”), the workplace (“Labor”), and interpersonal relationships (“Love”). Informed by intersectionality theory, we considered the role of power embedded in language by creating one power-neutral scenario and one power-laden scenario for each prompt. For example, power-neutral Learning prompts consist of a single student excelling in an academic subject, whereas the power-laden prompts consist of one star student helping a struggling student in an academic subject. We then analyze the resulting model responses for textual cues shown to exacerbate socio-psychological harms for minoritized individuals by race, gender, and sexual orientation. To do this at scale, we fine-tuned a coreference resolution model (gpt3.5-turbo) to perform automated extraction of characters’ gender references and names at high precision. To evaluate our model, we hand-label the inferred gender (based on gender references) and name on an evaluation set of 4,600 uniformly down-sampled story generations from all five models (0.0063, 95% CI). Fine-tuning our model on a non-overlapping set of 150 training examples yields precision above 98% for both gender references and names. Recall rates reach 97% for gender references and exceed 99% for names. Following previous studies, we infer racial signals from first names using fractionalized counting over the Florida Voter Registration Dataset (which consists of 27 million named individuals and self-identified racial identities). We find that when LMs are used for story writing, they generate texts that reinforce discrimination against minoritized groups by race, gender, and sexual orientation. Using mixed-methods analyses, we identify three specific harms: omission, subordination, and stereotyping. Stories produced by language models simultaneously underrepresent minoritized individuals as main characters while overrepresenting them as subordinated characters. Diverse consumers, if they are to be represented at all, disproportionately see themselves portrayed by language models as “struggling students” (as opposed to “star students”), “patients” or “defendants” (as opposed to “doctors” or “lawyers”), and a friend or romantic partner who is more likely to borrow money or do the chores for someone else. The magnitude of bias far exceeds the level of "real-world" inequities. Underrepresentation of non-dominant identities in power-neutral stories exceeds national demographics in the US by up to two orders of magnitude. Meanwhile, non-dominant character identities are up to thousands of times more likely to appear as subordinated than empowered. For example, Claude casts the name ”Juan” as a struggling student 1,380 times, yet only once as a star student. We find that these harms impact every non-dominant group we studied (in the US context). These include individuals with intersectional Asian, Black, Indigenous, Latine, NH/PI, MENA, Female, Non-binary, and Queer identities. Language models propagate a plethora of stereotypes that are known to inflict psychological harm and negative self-perception, including the ” glass/bamboo ceiling”, ” perpetual foreigner”, ”noble savage”, ”white savior”, and others.
Sub-federal authorities have the power to administer transportation tax, gambling tax, local taxes and taxes on total income, including setting rates, providing benefits or determining the tax base, which makes these taxes an important instrument of sub-national revenue policy. Together with the state duty, they make up about 10% of sub-federal tax revenues and are generally referred to as other types of taxes when assessing tax capacity in calculating fiscal equalization grants to regions. Given the increasing role of other types of taxes, the task of correct assessment of tax potential for them is becoming more and more urgent. The article examines the relationship between a region’s share in actual revenues from other types of taxes, a region’s share in the total labor force and a region’s shares in actual revenues from individual taxes using a sample of 85 regions for 2019–2023. The panel regression model employs individual fixed effects, which allows us to account for the impact of region-specific factors, including the level of regional tax competition, such as reduced tax rates and additional tax benefits. The assessment results show that the tax capacity for other types of taxes is proportional to the region’s share in the total labor force and the region’s share in the tax potential for personal income tax. These results allow us to refine the formula for assessing tax capacity for other types of taxes. Since the Russian tax system is currently undergoing significant changes, including those related to profit tax and tax levied in connection with the application of the simplified tax system, the obtained estimates can be subsequently refined using new data.
Transformer language models (LMs) exhibit behaviors -- from storytelling to code generation -- that seem to require tracking the unobserved state of an evolving world. How do they do this? We study state tracking in LMs trained or fine-tuned to compose permutations (i.e., to compute the order of a set of objects after a sequence of swaps). Despite the simple algebraic structure of this problem, many other tasks (e.g., simulation of finite automata and evaluation of boolean expressions) can be reduced to permutation composition, making it a natural model for state tracking in general. We show that LMs consistently learn one of two state tracking mechanisms for this task. The first closely resembles the "associative scan" construction used in recent theoretical work by Liu et al. (2023) and Merrill et al. (2024). The second uses an easy-to-compute feature (permutation parity) to partially prune the space of outputs, and then refines this with an associative scan. LMs that learn the former algorithm tend to generalize better and converge faster, and we show how to steer LMs toward one or the other with intermediate training tasks that encourage or suppress the heuristics. Our results demonstrate that transformer LMs, whether pre-trained or fine-tuned, can learn to implement efficient and interpretable state-tracking mechanisms, and the emergence of these mechanisms can be predicted and controlled.
This paper concerns the existence and related properties of solutions to the Schrödinger-Bopp-Podolsky system, which reduces to a nonlinear and nonlocal partial differential equation describing a Schrödinger field coupled with its electromagnetic field in Bopp-Podolsky theory under purely electrostatic conditions. Firstly, by applying the mountain-pass lemma, we obtain the existence of nontrivial solutions. Then, through some estimates of the ground state energy, we prove the existence of ground state solutions. By exploring the relationship between solutions and paths associated with critical points, we further demonstrate that the obtained solutions are ground states of mountain-pass type. Additionally, the positivity, radial symmetry, rotational invariance, and exponential decay of the ground state solutions are considered. Finally, in the radial case, we explore the asymptotic behavior of the obtained solutions with respect to $a$.
We introduce the AI Occupational Capability Index (AI-OCI), a novel methodology for quantifying the alignment between AI model capabilities and the tasks that define human occupations. Unlike prior automation risk metrics, which rely on expert heuristics or job-level generalizations, AI-OCI operates at the task level by embedding and comparing over 19,000 occupational tasks with 338 AI capabilities using state-of-the-art language models. The resulting scores reveal how well AI systems can perform specific human functions, enabling interpretable, task-aligned assessments of labor exposure. Empirical evaluations show strong correlations with benchmark indices such as AIOE and GPT-4 Beta exposure scores, while diverging from legacy automation risk measures. We demonstrate AI-OCI’s utility through case-based analyses of employment and wage shifts across high-alignment occupations during the era of large language model adoption. The framework supports scalable, real-time tracking of AI’s workforce impact and provides a foundation for integrating labor intelligence into education, policy, and economic planning.
The development of the derivatives financial instruments market is a crucial factor in strengthening financial stability and enhancing risk management efficiency in the context of globalized financial markets. The relevance of this topic stems from the underdeveloped state of this market segment in Ukraine and the need to establish an effective mechanism for its functioning. The aim of this study is to identify and systematize the factors influencing the development of the derivatives market in Ukraine and to substantiate their interrelationships. The research employs a combination of systemic, analytical, statistical, and comparative approaches, allowing for a comprehensive examination of the interaction between macroeconomic, market, and behavioral factors. The study follows a descriptive and analytical design and is based on official data from the National Securities and Stock Market Commission, the Ministry of Finance of Ukraine, and the World Federation of Exchanges. Based on a synthesis of scientific approaches, the study formulates the hypothesis that the development of the derivatives market is determined by the interplay of global, macroeconomic, and behavioral factors, with institutional conditions and the financial culture of market participants playing a key role. The methodological framework includes systemic and comparative analysis, statistical generalization, and content analysis of scientific sources, enabling a comprehensive assessment of market development trends from 2021 to 2025. The study summarizes the structure of factors affecting the derivatives market, establishes their hierarchical interrelationships, characterizes macroeconomic and behavioural determinants, systematizes barriers to market formation, evaluates trading volume dynamics, and proposes directions for further improvement of market infrastructure. The theoretical significance lies in deepening scientific understanding of the structure of determinants of the derivatives market and the potential application of the findings for further modelling. The practical significance is reflected in the possibility of using the results to improve regulatory policies, enhance financial literacy, and stimulate the exchange segment. The scientific novelty consists in the development of a multi-level classification of factors affecting the derivatives market, accounting for the interconnections between global, macroeconomic, and behavioural determinants. The main conclusions provide a comprehensive assessment of the market’s current state and its growth potential in the context of financial system digitalization. Future research prospects involve deepening empirical analysis of interrelationships among factor groups and conducting comparative studies with Central and Eastern European countries. This article is of a theoretical-analytical type.
Promoting low-carbon agriculture is vital for climate action and food security. State farms serve as crucial agricultural production bases in China and are essential in reducing China’s carbon emissions and boosting emission efficiency. This study calculates the carbon emissions of state farms across 29 Chinese provinces using the IPCC method from 2010 to 2022. It also evaluates emission efficiency with the Super-Slack-Based Measure (Super-SBM model) and analyzes influencing factors using the Logarithmic Mean Divisia Index (LMDI) method. The findings suggest that the three largest carbon sources are rice planting, chemical fertilizers, and land tillage. Secondly, agricultural carbon emissions in state farms initially surge, stabilize with fluctuations, and ultimately decline, with higher emissions observed in northern and eastern China. Thirdly, the rise of agricultural carbon emission efficiency is driven primarily by technological progress. Lastly, economic development and industry structure promote agricultural carbon emissions, while production efficiency and labor scale reduce them. To reduce carbon emissions from state farms in China and improve agricultural carbon emission efficiency, the following measures can be taken: (1) Improve agricultural production efficiency and reduce carbon emissions in all links; (2) Optimize the agricultural industrial structure and promote the coordinated development of agriculture; (3) Reduce the agricultural labor scale and promote the specialization, professionalization, and high-quality development of agricultural labor; (4) Accelerate agricultural green technology innovation and guide the green transformation of state farms. This study enriches the theoretical foundation of low-carbon agriculture and develops a framework for assessing carbon emissions in Chinese state farms, offering guidance for future research and policy development in sustainable agriculture.
This research provides a detailed comparative analysis of the number of labor disputes and the evolution of Kazakhstan’s labor discipline legislation. Key observations include the correlation between stringent disciplinary norms and reduced unemployment, the influence of judicial practices on labor discipline, and the necessity of balancing strict discipline with fair treatment to prevent legal disputes. The findings offer valuable insights for optimizing labor policies to support sustainable labor market development in Kazakhstan. These insights align with the goals of SDG 8, which promotes sustained, inclusive, and sustainable economic growth, full and productive employment, and decent work for all.
The purpose of this paper is not only to analyze the current state of the labor market in Kazakhstan but also to propose recommendations for optimizing labor policy, taking into account both current challenges and potential opportunities for sustainable development of the labor market in the country. Based on a wide range of data, including statistical indicators range of unemployment levels, legislative changes, and analysis of judicial practice, we aim to identify correlations and cause-and-effect relationships between labor discipline and market indicators.
Keywords: employee, employer, labor, labor law, employment contract, termination of employment contract, labor relations, labor disputes, dismissal, at the initiative of the employer.
Law in general. Comparative and uniform law. Jurisprudence
Recently, a simple but powerful language for expressing and learning general policies and problem decompositions (sketches) has been introduced in terms of rules defined over a set of Boolean and numerical features. In this work, we consider three extensions of this language aimed at making policies and sketches more flexible and reusable: internal memory states, as in finite state controllers; indexical features, whose values are a function of the state and a number of internal registers that can be loaded with objects; and modules that wrap up policies and sketches and allow them to call each other by passing parameters. In addition, unlike general policies that select state transitions rather than ground actions, the new language allows for the selection of such actions. The expressive power of the resulting language for policies and sketches is illustrated through a number of examples.