Emerging large language model (LLM) applications involve diverse reasoning strategies and agentic workflows, straining the capabilities of existing serving systems built on a monolithic token generation loop. This paper introduces Pie, a programmable LLM serving system designed for flexibility and efficiency. Pie decomposes the traditional generation loop into fine-grained service handlers exposed via an API and delegates control of the generation process to user-provided programs, called inferlets. This enables applications to implement new KV cache strategies, bespoke generation logic, and seamlessly integrate computation and I/O-entirely within the application, without requiring modifications to the serving system. Pie executes inferlets using WebAssembly, benefiting from its lightweight sandboxing. Our evaluation shows Pie matches state-of-the-art performance on standard tasks (3-12% latency overhead) while significantly improving latency and throughput (1.3x-3.4x higher) on agentic workflows by enabling application-specific optimizations.

Long-horizon robot manipulation tasks remain challenging for Vision-Language-Action (VLA) policies due to drift and exposure bias, often denoise the entire trajectory with fixed hyperparameters, causing small geometric errors to compound across stages and offering no mechanism to allocate extra test-time compute where clearances are tight. To address these challenges, we introduce FORGE-Tree, a plug-in control layer that couples a stage-aligned Diffusion Forcing (DF) head with test-time Monte Carlo Tree Diffusion (MCTD). With a frozen VLA encoder, DF aligns timesteps to subtask stages; during inference we partially denoise only a target segment while keeping other tokens frozen, turning trajectory refinement into a sequence of local edits. We then apply Monte Carlo Tree Diffusion to select the next segment to refine. A scene graph supplies priors for expansion and geometry relation-aware scoring for rollouts, yielding tree-structured denoising whose performance scales with search budget while preserving the executed prefix. Evaluation on LIBERO, FORGE-Tree improves success rate by 13.4 to 17.2 pp over the native VLA baselines with both OpenVLA and Octo-Base. Gains remain consistent under comparable compute budgets, especially on long-horizon variants. Videos available at: https://taco-group.github.io/FORGE-Tree/

Distributed microgrids are conventionally dependent on communication networks to achieve secondary control objectives. This dependence makes them vulnerable to stealth data integrity attacks (DIAs) where adversaries may perform manipulations via infected transmitters and repeaters to jeopardize stability. This paper presents a physics-guided, supervised Artificial Neural Network (ANN)-based framework that identifies communication-level cyberattacks in microgrids by analyzing whether incoming measurements will cause abnormal behavior of the secondary control layer. If abnormalities are detected, an iteration through possible spanning tree graph topologies that can be used to fulfill secondary control objectives is done. Then, a communication network topology that would not create secondary control abnormalities is identified and enforced for maximum stability. By altering the communication graph topology, the framework eliminates the dependence of the secondary control layer on inputs from compromised cyber devices helping it achieve resilience without instability. Several case studies are provided showcasing the robustness of the framework against False Data Injections and repeater-level Man-in-the-Middle attacks. To understand practical feasibility, robustness is also verified against larger microgrid sizes and in the presence of varying noise levels. Our findings indicate that performance can be affected when attempting scalability in the presence of noise. However, the framework operates robustly in low-noise settings.

The Generalized Sliding-Tile Puzzle (GSTP), allowing many square tiles on a board to move in parallel while enforcing natural geometric collision constraints on the movement of neighboring tiles, provide a high-fidelity mathematical model for many high-utility existing and future multi-robot applications, e.g., at mobile robot-based warehouses or autonomous garages. Motivated by practical relevance, this work examines a further generalization of GSTP called the Colored Generalized Sliding-Tile Puzzle (CGSP), where tiles can now assume varying degrees of distinguishability, a common occurrence in the aforementioned applications. Our study establishes the computational complexity of CGSP and its key sub-problems under a broad spectrum of possible conditions and characterizes solution makespan lower and upper bounds that differ by at most a logarithmic factor. These results are further extended to higher-dimensional versions of the puzzle game.

Understanding collective behavior and how it evolves is important to ensure that robot swarms can be trusted in a shared environment. One way to understand the behavior of the swarm is through collective behavior reconstruction using prior demonstrations. Existing approaches often require access to the swarm controller which may not be available. We reconstruct collective behaviors in distinct swarm scenarios involving shared environments without using swarm controller information. We achieve this by transforming prior demonstrations into features that sufficiently describe multi-agent interactions before behavior reconstruction with multi-agent generative adversarial imitation learning (MA-GAIL). We show that our approach outperforms existing algorithms in all investigated swarm scenarios, and can be used to observe and reconstruct a swarm's behavior for further analysis and testing, which might be impractical or undesirable on the original robot swarm.

The ability to generate sentiment-controlled feedback in response to multimodal inputs, comprising both text and images, addresses a critical gap in human-computer interaction by enabling systems to provide empathetic, accurate, and engaging responses. This capability has profound applications in healthcare, marketing, and education. To this end, we construct a large-scale Controllable Multimodal Feedback Synthesis (CMFeed) dataset and propose a controllable feedback synthesis system. The proposed system includes an encoder, decoder, and controllability block for textual and visual inputs. It extracts textual and visual features using a transformer and Faster R-CNN networks and combines them to generate feedback. The CMFeed dataset encompasses images, text, reactions to the post, human comments with relevance scores, and reactions to the comments. The reactions to the post and comments are utilized to train the proposed model to produce feedback with a particular (positive or negative) sentiment. A sentiment classification accuracy of 77.23% has been achieved, 18.82% higher than the accuracy without using the controllability. Moreover, the system incorporates a similarity module for assessing feedback relevance through rank-based metrics. It implements an interpretability technique to analyze the contribution of textual and visual features during the generation of uncontrolled and controlled feedback.

This paper focuses on the need for a rigorous theory of layered control architectures (LCAs) for complex engineered and natural systems, such as power systems, communication networks, autonomous robotics, bacteria, and human sensorimotor control. All deliver extraordinary capabilities, but they lack a coherent theory of analysis and design, partly due to the diverse domains across which LCAs can be found. In contrast, there is a core universal set of control concepts and theory that applies very broadly and accommodates necessary domain-specific specializations. However, control methods are typically used only to design algorithms in components within a larger system designed by others, typically with minimal or no theory. This points towards a need for natural but large extensions of robust performance from control to the full decision and control stack. It is encouraging that the successes of extant architectures from bacteria to the Internet are due to strikingly universal mechanisms and design patterns. This is largely due to convergent evolution by natural selection and not intelligent design, particularly when compared with the sophisticated design of components. Our aim here is to describe the universals of architecture and sketch tentative paths towards a useful design theory.

In this paper we present a computational modeling account of an active self in artificial agents. In particular we focus on how an agent can be equipped with a sense of control and how it arises in autonomous situated action and, in turn, influences action control. We argue that this requires laying out an embodied cognitive model that combines bottom-up processes (sensorimotor learning and fine-grained adaptation of control) with top-down processes (cognitive processes for strategy selection and decision-making). We present such a conceptual computational architecture based on principles of predictive processing and free energy minimization. Using this general model, we describe how a sense of control can form across the levels of a control hierarchy and how this can support action control in an unpredictable environment. We present an implementation of this model as well as first evaluations in a simulated task scenario, in which an autonomous agent has to cope with un-/predictable situations and experiences corresponding sense of control. We explore different model parameter settings that lead to different ways of combining low-level and high-level action control. The results show the importance of appropriately weighting information in situations where the need for low/high-level action control varies and they demonstrate how the sense of control can facilitate this.

The previous works on formalizing enterprise application integration (EAI) scenarios showed an emerging need for setting up formal foundations for integration patterns, the EAI building blocks, in order to facilitate the model-driven development and ensure its correctness. So far, the formalization requirements were focusing on more "conventional" integration scenarios, in which control-flow, transactional persistent data and time aspects were considered. However, none of these works took into consideration another arising EAI trend that covers social and multimedia computing. In this work we propose a Petri net-based formalism that addresses requirements arising from the multimedia domain. We also demonstrate realizations of one of the most frequently used multimedia patterns and discuss which implications our formal proposal may bring into the area of the multimedia EAI development.

The sheer amount of services and network complexity will require a step up of current network capabilities. Specifically, 5G networks will need to incorporate Artificial Intelligence (AI) and its offspring Machine Learning (ML). As AI/ML continue to gain steam and the rest of the business world gets on board, current networks are suffering from the lack of capabilities needed. To be honest, today's mostly manual, static networks are not suited for these advanced technologies. And while agile, self-organizing networks will exist in the future, service providers need to address their digital transformation efforts today, focusing on near-term solutions, to build the foundation for these networks of tomorrow.