Cooperation is vital to our survival and progress. Evolutionary game theory offers a lens to understand the structures and incentives that enable cooperation to be a successful strategy. As artificial intelligence agents become integral to human systems, the dynamics of cooperation take on unprecedented significance. The convergence of human-agent teaming, contract theory, and decentralized frameworks like Web3, grounded in transparency, accountability, and trust, offers a foundation for fostering cooperation by establishing enforceable rules and incentives for humans and AI agents. We conceptualize Incentivized Symbiosis as a social contract between humans and AI, inspired by Web3 principles and encoded in blockchain technology, to define and enforce rules, incentives, and consequences for both parties. By exploring this paradigm, we aim to catalyze new research at the intersection of systems thinking in AI, Web3, and society, fostering innovative pathways for cooperative human-agent coevolution.

Following the first part of our project, this paper comprehensively studies two types of extragradient-based methods: anchored extragradient and Nesterov's accelerated extragradient for solving [non]linear inclusions (and, in particular, equations), primarily under the Lipschitz continuity and the co-hypomonotonicity assumptions. We unify and generalize a class of anchored extragradient methods for monotone inclusions to a wider range of schemes encompassing existing algorithms as special cases. We establish $\mathcal{O}(1/k)$ last-iterate convergence rates on the residual norm of the underlying mapping for this general framework and then specialize it to obtain convergence guarantees for specific instances, where $k$ denotes the iteration counter. We extend our approach to a class of anchored Tseng's forward-backward-forward splitting methods to obtain a broader class of algorithms for solving co-hypomonotone inclusions. Again, we analyze $\mathcal{O}(1/k)$ last-iterate convergence rates for this general scheme and specialize it to obtain convergence results for existing and new variants. We generalize and unify Nesterov's accelerated extra-gradient method to a new class of algorithms that covers existing schemes as special instances while generating new variants. For these schemes, we can prove $\mathcal{O}(1/k)$ last-iterate convergence rates for the residual norm under co-hypomonotonicity, covering a class of nonmonotone problems. We propose another novel class of Nesterov's accelerated extragradient methods to solve inclusions. Interestingly, these algorithms achieve both $\mathcal{O}(1/k)$ and $o(1/k)$ last-iterate convergence rates, and also the convergence of iterate sequences under co-hypomonotonicity and Lipschitz continuity. Finally, we provide a set of numerical experiments encompassing different scenarios to validate our algorithms and theoretical guarantees.

Advancements in artificial intelligence (AI) and deep learning have led to neural networks being used to generate lightning-speed answers to complex questions, to paint like Monet, or to write like Proust. Leveraging their computational speed and flexibility, neural networks are also being used to facilitate fast, likelihood-free statistical inference. However, it is not straightforward to use neural networks with data that for various reasons are incomplete, which precludes their use in many applications. A recently proposed approach to remedy this issue inputs an appropriately padded data vector and a vector that encodes the missingness pattern to a neural network. While computationally efficient, this "masking" approach can result in statistically inefficient inferences. Here, we propose an alternative approach that is based on the Monte Carlo expectation-maximization (EM) algorithm. Our EM approach is likelihood-free, substantially faster than the conventional EM algorithm as it does not require numerical optimization at each iteration, and more statistically efficient than the masking approach. This research represents a prototype problem that indicates how improvements could be made in AI by introducing Bayesian statistical thinking. We compare the two approaches to missingness using simulated incomplete data from two models: a spatial Gaussian process model, and a spatial Potts model. The utility of the methodology is shown on Arctic sea-ice data and cryptocurrency data.

How can we attribute the behaviors of machine learning models to their training data? While the classic influence function sheds light on the impact of individual samples, it often fails to capture the more complex and pronounced collective influence of a set of samples. To tackle this challenge, we study the Most Influential Subset Selection (MISS) problem, which aims to identify a subset of training samples with the greatest collective influence. We conduct a comprehensive analysis of the prevailing approaches in MISS, elucidating their strengths and weaknesses. Our findings reveal that influence-based greedy heuristics, a dominant class of algorithms in MISS, can provably fail even in linear regression. We delineate the failure modes, including the errors of influence function and the non-additive structure of the collective influence. Conversely, we demonstrate that an adaptive version of these heuristics which applies them iteratively, can effectively capture the interactions among samples and thus partially address the issues. Experiments on real-world datasets corroborate these theoretical findings and further demonstrate that the merit of adaptivity can extend to more complex scenarios such as classification tasks and non-linear neural networks. We conclude our analysis by emphasizing the inherent trade-off between performance and computational efficiency, questioning the use of additive metrics such as the Linear Datamodeling Score, and offering a range of discussions.

The field of Artificial Intelligence (AI) continues to drive transformative innovations, with significant progress in conversational interfaces, autonomous vehicles, and intelligent content creation. Since the launch of ChatGPT in late 2022, the rise of Generative AI has marked a pivotal era, with the term Large Language Models (LLMs) becoming a ubiquitous part of daily life. LLMs have demonstrated exceptional capabilities in tasks such as text summarization, code generation, and creative writing. However, these models are inherently limited by their token-level processing, which restricts their ability to perform abstract reasoning, conceptual understanding, and efficient generation of long-form content. To address these limitations, Meta has introduced Large Concept Models (LCMs), representing a significant shift from traditional token-based frameworks. LCMs use concepts as foundational units of understanding, enabling more sophisticated semantic reasoning and context-aware decision-making. Given the limited academic research on this emerging technology, our study aims to bridge the knowledge gap by collecting, analyzing, and synthesizing existing grey literature to provide a comprehensive understanding of LCMs. Specifically, we (i) identify and describe the features that distinguish LCMs from LLMs, (ii) explore potential applications of LCMs across multiple domains, and (iii) propose future research directions and practical strategies to advance LCM development and adoption.

As AI systems become more capable, widely deployed, and increasingly autonomous in critical areas such as cybersecurity, biological research, and healthcare, ensuring their safety and alignment with human values is paramount. Machine unlearning -- the ability to selectively forget or suppress specific types of knowledge -- has shown promise for privacy and data removal tasks, which has been the primary focus of existing research. More recently, its potential application to AI safety has gained attention. In this paper, we identify key limitations that prevent unlearning from serving as a comprehensive solution for AI safety, particularly in managing dual-use knowledge in sensitive domains like cybersecurity and chemical, biological, radiological, and nuclear (CBRN) safety. In these contexts, information can be both beneficial and harmful, and models may combine seemingly harmless information for harmful purposes -- unlearning this information could strongly affect beneficial uses. We provide an overview of inherent constraints and open problems, including the broader side effects of unlearning dangerous knowledge, as well as previously unexplored tensions between unlearning and existing safety mechanisms. Finally, we investigate challenges related to evaluation, robustness, and the preservation of safety features during unlearning. By mapping these limitations and open challenges, we aim to guide future research toward realistic applications of unlearning within a broader AI safety framework, acknowledging its limitations and highlighting areas where alternative approaches may be required.

Multimodal Large Language Models (MLLMs) have achieved impressive performance and have been put into practical use in commercial applications, but they still have potential safety mechanism vulnerabilities. Jailbreak attacks are red teaming methods that aim to bypass safety mechanisms and discover MLLMs' potential risks. Existing MLLMs' jailbreak methods often bypass the model's safety mechanism through complex optimization methods or carefully designed image and text prompts. Despite achieving some progress, they have a low attack success rate on commercial closed-source MLLMs. Unlike previous research, we empirically find that there exists a Shuffle Inconsistency between MLLMs' comprehension ability and safety ability for the shuffled harmful instruction. That is, from the perspective of comprehension ability, MLLMs can understand the shuffled harmful text-image instructions well. However, they can be easily bypassed by the shuffled harmful instructions from the perspective of safety ability, leading to harmful responses. Then we innovatively propose a text-image jailbreak attack named SI-Attack. Specifically, to fully utilize the Shuffle Inconsistency and overcome the shuffle randomness, we apply a query-based black-box optimization method to select the most harmful shuffled inputs based on the feedback of the toxic judge model. A series of experiments show that SI-Attack can improve the attack's performance on three benchmarks. In particular, SI-Attack can obviously improve the attack success rate for commercial MLLMs such as GPT-4o or Claude-3.5-Sonnet.

Current and upcoming generations of visible-shortwave infrared (VSWIR) imaging spectrometers promise unprecedented capacity to quantify Earth System processes across the globe. However, reliable cloud screening remains a fundamental challenge for these instruments, where traditional spatial and temporal approaches are limited by cloud variability and limited temporal coverage. The Spectroscopic Transformer (SpecTf) addresses these challenges with a spectroscopy-specific deep learning architecture that performs cloud detection using only spectral information (no spatial or temporal data are required). By treating spectral measurements as sequences rather than image channels, SpecTf learns fundamental physical relationships without relying on spatial context. Our experiments demonstrate that SpecTf significantly outperforms the current baseline approach implemented for the EMIT instrument, and performs comparably with other machine learning methods with orders of magnitude fewer learned parameters. Critically, we demonstrate SpecTf's inherent interpretability through its attention mechanism, revealing physically meaningful spectral features the model has learned. Finally, we present SpecTf's potential for cross-instrument generalization by applying it to a different instrument on a different platform without modifications, opening the door to instrument agnostic data driven algorithms for future imaging spectroscopy tasks.

HotFlip is a topical gradient-based word substitution method for attacking language models. Recently, this method has been further applied to attack retrieval systems by generating malicious passages that are injected into a corpus, i.e., corpus poisoning. However, HotFlip is known to be computationally inefficient, with the majority of time being spent on gradient accumulation for each query-passage pair during the adversarial token generation phase, making it impossible to generate an adequate number of adversarial passages in a reasonable amount of time. Moreover, the attack method itself assumes access to a set of user queries, a strong assumption that does not correspond to how real-world adversarial attacks are usually performed. In this paper, we first significantly boost the efficiency of HotFlip, reducing the adversarial generation process from 4 hours per document to only 15 minutes, using the same hardware. We further contribute experiments and analysis on two additional tasks: (1) transfer-based black-box attacks, and (2) query-agnostic attacks. Whenever possible, we provide comparisons between the original method and our improved version. Our experiments demonstrate that HotFlip can effectively attack a variety of dense retrievers, with an observed trend that its attack performance diminishes against more advanced and recent methods. Interestingly, we observe that while HotFlip performs poorly in a black-box setting, indicating limited capacity for generalization, in query-agnostic scenarios its performance is correlated to the volume of injected adversarial passages.

Providing a comprehensive view of the city operation and offering useful metrics for decision making is a well known challenge for urban risk analysis systems. Existing systems are, in many cases, generalizations of previous domain specific tools and or methodologies that may not cover all urban interdependencies and makes it difficult to have homogeneous indicators. In order to overcome this limitation while seeking for effective support to decision makers, this article introduces a novel hybrid simulation framework for risk mitigation. The framework is built on a proposed city concept that considers urban space as a Complex Adaptive System composed by interconnected Critical Infrastructures. In this concept, a Social System, which models daily patterns and social interactions of the citizens in the Urban Landscape, drives the CIs demand to configure the full city picture. The frameworks hybrid design integrates agent based and network based modeling by breaking down city agents into system dependent subagents, to enable both inter and intra system interaction simulation, respectively. A layered structure of indicators at different aggregation levels is also developed, to ensure that decisions are not only data driven but also explainable. Therefore, the proposed simulation framework can serve as a DSS tool that allows the quantitative analysis of the impact of threats at different levels. First, system level metrics can be used to get a broad view on the city resilience. Then, agent level metrics back those figures and provide better explainability. On implementation, the proposed framework enables component reusability (for eased coding), simulation federation (enabling the integration of existing system oriented simulators), discrete simulation in accelerated time (for rapid scenario simulation) and decision oriented visualization (for informed outputs).