Abstract
The global capacity for mineral processing must expand rapidly to meet the demand for critical minerals, which are essential for building the clean energy technologies necessary to mitigate climate change. However, the efficiency of mineral processing is severely limited by uncertainty, which arises from both the variability of feedstock and the complexity of process dynamics. To optimize mineral processing circuits under uncertainty, we introduce an AI-driven approach that formulates mineral processing as a Partially Observable Markov Decision Process (POMDP). We demonstrate the capabilities of this approach in handling both feedstock uncertainty and process model uncertainty to optimize the operation of a simulated, simplified flotation cell as an example. We show that by integrating the process of information gathering (i.e., uncertainty reduction) and process optimization, this approach has the potential to consistently perform better than traditional approaches at maximizing an overall objective, such as net present value (NPV). Our methodological demonstration of this optimization-under-uncertainty approach for a synthetic case provides a mathematical and computational framework for later real-world application, with the potential to improve both the laboratory-scale design of experiments and industrial-scale operation of mineral processing circuits without any additional hardware.

AI Summary

• The authors propose an online solver that uses Monte Carlo tree search to explore the action space and make decisions in real-time. [3]
• The online solver proposed by the authors is able to make decisions in real-time using Monte Carlo tree search, which allows for efficient exploration of the action space. [3]
• The paper presents a novel approach to optimize mineral processing circuits under uncertainty using Partially Observable Markov Decision Processes (POMDPs). [2]

Abstract
As artificial intelligence (AI) agents are deployed across economic domains, understanding their strategic behavior and market-level impact becomes critical. This paper puts forward a groundbreaking new framework that is the first to capture the real-world economic forces that shape agentic labor markets: adverse selection, moral hazard, and reputation dynamics. Our framework encapsulates three core capabilities that successful LLM-agents will need: \textbf{metacognition} (accurate self-assessment of skills), \textbf{competitive awareness} (modeling rivals and market dynamics), and \textbf{long-horizon strategic planning}. We illustrate our framework through a tractable simulated gig economy where agentic Large Language Models (LLMs) compete for jobs, develop skills, and adapt their strategies under competitive pressure. Our simulations illustrate how LLM agents explicitly prompted with reasoning capabilities learn to strategically self-improve and demonstrate superior adaptability to changing market conditions. At the market level, our simulations reproduce classic macroeconomic phenomena found in human labor markets, while controlled experiments reveal potential AI-driven economic trends, such as rapid monopolization and systemic price deflation. This work provides a foundation to further explore the economic properties of AI-driven labour markets, and a conceptual framework to study the strategic reasoning capabilities in agents competing in the emerging economy.

AI Summary

• The impact of generative AI on online labor markets is significant, with studies showing reductions in employment and earnings for human freelancers in artistic professions. [3]
• AI agents can perform tasks more quickly than humans, allowing them to provide more value to clients and potentially dominating a labor market in a monopolistic fashion. [3]
• Agent-based computational economics (ACE): A subfield that utilizes computational agents to model and understand economic phenomena from a bottom-up perspective. [3]
• Self-Improving and Reflective Agents: AI systems that can improve themselves, often through self-reflection or feedback from the environment. [3]
• The introduction of generative AI has significant implications for online labor markets, including reduced employment and earnings for human freelancers in artistic professions. [3]
• The replicability and speed of task completion by AI agents also have the potential to dominate a labor market in a monopolistic fashion. [3]
• The study assumes that the impact of generative AI on online labor markets is solely due to its ability to automate certain tasks, without considering other factors such as changes in consumer demand or supply chain disruptions. [3]
• Agent-based computational economics (ACE) is a subfield that utilizes computational agents to model and understand economic phenomena from a bottom-up perspective. [3]
• Self-Improving and Reflective Agents are AI systems that can improve themselves, often through self-reflection or feedback from the environment. [3]
• Online labor markets: Platforms that match clients with remote service providers for tasks such as data entry, software programming, design, or analytics. [2]

Abstract
Posted-price mechanisms (PPMs) are a widely adopted strategy for online resource allocation due to their simplicity, intuitive nature, and incentive compatibility. To manage the uncertainty inherent in online settings, PPMs commonly employ dynamically increasing prices. While this adaptive pricing achieves strong performance, it introduces practical challenges: dynamically changing prices can lead to fairness concerns stemming from price discrimination and incur operational costs associated with frequent updates. This paper addresses these issues by investigating posted pricing constrained by a limited, pre-specified number of allowed price changes, denoted by $Δ$. We further extend this framework by incorporating a second critical dimension: risk sensitivity. Instead of evaluating performance based solely on expectation, we utilize a tail-risk objective-specifically, the Conditional Value at Risk (CVaR) of the total social welfare, parameterized by a risk level $δ\in [0, 1]$. We formally introduce a novel problem class kSelection-$(δ,Δ)$ in online adversarial selection and propose a correlated PPM that utilizes a single random seed to correlate posted prices. This correlation scheme is designed to address both the limited price changes and simultaneously enhance the tail performance of the online algorithm. Our subsequent analysis provides performance guarantees under these joint constraints, revealing a clear trade-off between the number of allowed price changes and the algorithm's risk sensitivity. We also establish optimality results for several important special cases of the problem.

AI Summary

• The paper presents a novel algorithm for the k-Selection problem with limited price changes (k-Selection-(δ,Δ)), which is a variant of the classic k-Selection problem. [2]
• The paper also presents a novel approach for solving delay differential equations, which is used to design each pricing function in Theorem 4. [1]

Abstract
Determining the optimal selling price is a challenge in revenue management, especially in markets characterized by nonlinear and price-sensitive demand. While traditional models, such as linear, power, and exponential demand functions, offer analytical convenience, they often fail to capture realistic purchase dynamics, leading to suboptimal pricing. The logit demand function addresses these limitations through its bounded, S-shaped curve, offering a more realistic representation of consumer behavior. Despite its advantages, most existing literature relies on heuristic approaches, such as pricing at the inflection point, which prioritizes maximum price sensitivity but does not guarantee maximum revenue. This study proposes a novel, exact pricing algorithm that analytically derives the revenue-maximizing price under the logit demand function using the Lambert W function. By providing a closed-form solution, the approach eliminates reliance on heuristic iterative methods and corrects the common practice of considering the inflection point price as market price. In fact, we demonstrate that the optimal price is consistently lower than the inflection-point price under reasonable assumptions, leading to lower prices for consumers and higher revenue for sellers. Numerical experiments illustrate the proposed algorithm and examine the changes in the optimality gap as demand function parameters vary. Results indicate that the optimal price is consistently lower than the inflection-point price, with an average 20% price reduction accompanied by a 15% increase in revenue.

AI Summary

• The study provides a comprehensive analysis of the logit demand function, including its properties and implications for pricing decisions. [3]
• Logit demand function: A mathematical model used to describe consumer demand in terms of price and other factors. [3]
• Inflection point: The point at which the second derivative of the demand function changes sign, indicating a change in the rate of change of demand with respect to price. [3]
• The research highlights that maximum revenue is not attained at the inflection point but rather at a lower price where the trade-off between price and demand is optimized. [2]

Abstract
We introduce the first version of the AI Consumer Index (ACE), a benchmark for assessing whether frontier AI models can perform high-value consumer tasks. ACE contains a hidden heldout set of 400 test cases, split across four consumer activities: shopping, food, gaming, and DIY. We are also open sourcing 80 cases as a devset with a CC-BY license. For the ACE leaderboard we evaluated 10 frontier models (with websearch turned on) using a novel grading methodology that dynamically checks whether relevant parts of the response are grounded in the retrieved web sources. GPT 5 (Thinking = High) is the top-performing model, scoring 56.1%, followed by o3 Pro (Thinking = On) (55.2%) and GPT 5.1 (Thinking = High) (55.1%). Models differ across domains, and in Shopping the top model scores under 50%. For some requests (such as giving the correct price or providing working links), models are highly prone to hallucination. Overall, ACE shows a substantial gap between the performance of even the best models and consumers' AI needs.

AI Summary

• The ACE benchmark is a comprehensive evaluation of conversational AI models, covering various domains such as DIY, food, gaming, and shopping. [3]
• The benchmark consists of multiple workflows for each domain, with specific instructions and criteria for the models to follow. [3]
• Gemini 2.5 Flash (On), Gemini 3 Pro (High), and o3 (On) also demonstrate strong performance across various domains. [3]
• The benchmark highlights the strengths and weaknesses of each model, providing valuable insights for developers and researchers to improve their conversational AI systems. [3]
• ACE-v1-heldout: A subset of the ACE benchmark used for evaluation, consisting of 100 cases per domain. [3]
• Bootstrapped confidence intervals: A statistical method used to estimate the uncertainty of mean scores by resampling with replacement from the original dataset. [3]
• Domain: A specific category or area of expertise within the ACE benchmark, such as DIY, food, gaming, or shopping. [3]
• Model: A conversational AI system being evaluated on the ACE benchmark, including models like Gemini 2.5 Flash (On), GPT-5 (High), and o3 (On). [3]
• The ACE benchmark provides a comprehensive evaluation of conversational AI models across various domains. [3]
• The results show that GPT-5 (High) and GPT-5.1 (High) perform exceptionally well in most domains, achieving high mean scores and confidence intervals. [2]

Abstract
As computational demands continue to rise, assessing the environmental footprint of AI requires moving beyond energy and water consumption to include the material demands of specialized hardware. This study quantifies the material footprint of AI training by linking computational workloads to physical hardware needs. The elemental composition of the Nvidia A100 SXM 40 GB graphics processing unit (GPU) was analyzed using inductively coupled plasma optical emission spectroscopy, which identified 32 elements. The results show that AI hardware consists of about 90% heavy metals and only trace amounts of precious metals. The elements copper, iron, tin, silicon, and nickel dominate the GPU composition by mass. In a multi-step methodology, we integrate these measurements with computational throughput per GPU across varying lifespans, accounting for the computational requirements of training specific AI models at different training efficiency regimes. Scenario-based analyses reveal that, depending on Model FLOPs Utilization (MFU) and hardware lifespan, training GPT-4 requires between 1,174 and 8,800 A100 GPUs, corresponding to the extraction and eventual disposal of up to 7 tons of toxic elements. Combined software and hardware optimization strategies can reduce material demands: increasing MFU from 20% to 60% lowers GPU requirements by 67%, while extending lifespan from 1 to 3 years yields comparable savings; implementing both measures together reduces GPU needs by up to 93%. Our findings highlight that incremental performance gains, such as those observed between GPT-3.5 and GPT-4, come at disproportionately high material costs. The study underscores the necessity of incorporating material resource considerations into discussions of AI scalability, emphasizing that future progress in AI must align with principles of resource efficiency and environmental responsibility.