InfoPO: Information-Driven Policy Optimization for User-Centric Agents

Real-world user requests to LLM agents are often underspecified. Agents must interact to acquire missing information and make correct downstream decisions. However, current multi-turn GRPO-based methods often rely on trajectory-level reward computation, which leads to credit assignment problems and insufficient advantage signals within rollout groups. A feasible approach is to identify valuable interaction turns at a fine granularity to drive more targeted learning. To address this, we introduce InfoPO (Information-Driven Policy Optimization), which frames multi-turn interaction as a process of active uncertainty reduction and computes an information-gain reward that credits turns whose feedback measurably changes the agent's subsequent action distribution compared to a masked-feedback counterfactual. It then combines this signal with task outcomes via an adaptive variance-gated fusion to identify information importance while maintaining task-oriented goal direction. Across diverse tasks, including intent clarification, collaborative coding, and tool-augmented decision making, InfoPO consistently outperforms prompting and multi-turn RL baselines. It also demonstrates robustness under user simulator shifts and generalizes effectively to environment-interactive tasks. Overall, InfoPO provides a principled and scalable mechanism for optimizing complex agent-user collaboration. Code is available at https://github.com/kfq20/InfoPO.

AutoWebWorld: Synthesizing Infinite Verifiable Web Environments via Finite State Machines

The performance of autonomous Web GUI agents heavily relies on the quality and quantity of their training data. However, a fundamental bottleneck persists: collecting interaction trajectories from real-world websites is expensive and difficult to verify. The underlying state transitions are hidden, leading to reliance on inconsistent and costly external verifiers to evaluate step-level correctness. To address this, we propose AutoWebWorld, a novel framework for synthesizing controllable and verifiable web environments by modeling them as Finite State Machines (FSMs) and use coding agents to translate FSMs into interactive websites. Unlike real websites, where state transitions are implicit, AutoWebWorld explicitly defines all states, actions, and transition rules. This enables programmatic verification: action correctness is checked against predefined rules, and task success is confirmed by reaching a goal state in the FSM graph. AutoWebWorld enables a fully automated search-and-verify pipeline, generating over 11,663 verified trajectories from 29 diverse web environments at only $0.04 per trajectory. Training on this synthetic data significantly boosts real-world performance. Our 7B Web GUI agent outperforms all baselines within 15 steps on WebVoyager. Furthermore, we observe a clear scaling law: as the synthetic data volume increases, performance on WebVoyager and Online-Mind2Web consistently improves.

Human-Aligned MLLM Judges for Fine-Grained Image Editing Evaluation: A Benchmark, Framework, and Analysis

Evaluating image editing models remains challenging due to the coarse granularity and limited interpretability of traditional metrics, which often fail to capture aspects important to human perception and intent. Such metrics frequently reward visually plausible outputs while overlooking controllability, edit localization, and faithfulness to user instructions. In this work, we introduce a fine-grained Multimodal Large Language Model (MLLM)-as-a-Judge framework for image editing that decomposes common evaluation notions into twelve fine-grained interpretable factors spanning image preservation, edit quality, and instruction fidelity. Building on this formulation, we present a new human-validated benchmark that integrates human judgments, MLLM-based evaluations, model outputs, and traditional metrics across diverse image editing tasks. Through extensive human studies, we show that the proposed MLLM judges align closely with human evaluations at a fine granularity, supporting their use as reliable and scalable evaluators. We further demonstrate that traditional image editing metrics are often poor proxies for these factors, failing to distinguish over-edited or semantically imprecise outputs, whereas our judges provide more intuitive and informative assessments in both offline and online settings. Together, this work introduces a benchmark, a principled factorization, and empirical evidence positioning fine-grained MLLM judges as a practical foundation for studying, comparing, and improving image editing approaches.

Agent Banana: High-Fidelity Image Editing with Agentic Thinking and Tooling

We study instruction-based image editing under professional workflows and identify three persistent challenges: (i) editors often over-edit, modifying content beyond the user's intent; (ii) existing models are largely single-turn, while multi-turn edits can alter object faithfulness; and (iii) evaluation at around 1K resolution is misaligned with real workflows that often operate on ultra high-definition images (e.g., 4K). We propose Agent Banana, a hierarchical agentic planner-executor framework for high-fidelity, object-aware, deliberative editing. Agent Banana introduces two key mechanisms: (1) Context Folding, which compresses long interaction histories into structured memory for stable long-horizon control; and (2) Image Layer Decomposition, which performs localized layer-based edits to preserve non-target regions while enabling native-resolution outputs. To support rigorous evaluation, we build HDD-Bench, a high-definition, dialogue-based benchmark featuring verifiable stepwise targets and native 4K images (11.8M pixels) for diagnosing long-horizon failures. On HDD-Bench, Agent Banana achieves the best multi-turn consistency and background fidelity (e.g., IC 0.871, SSIM-OM 0.84, LPIPS-OM 0.12) while remaining competitive on instruction following, and also attains strong performance on standard single-turn editing benchmarks. We hope this work advances reliable, professional-grade agentic image editing and its integration into real workflows.

AOrchestra: Automating Sub-Agent Creation for Agentic Orchestration

Language agents have shown strong promise for task automation. Realizing this promise for increasingly complex, long-horizon tasks has driven the rise of a sub-agent-as-tools paradigm for multi-turn task solving. However, existing designs still lack a dynamic abstraction view of sub-agents, thereby hurting adaptability. We address this challenge with a unified, framework-agnostic agent abstraction that models any agent as a tuple Instruction, Context, Tools, Model. This tuple acts as a compositional recipe for capabilities, enabling the system to spawn specialized executors for each task on demand. Building on this abstraction, we introduce an agentic system AOrchestra, where the central orchestrator concretizes the tuple at each step: it curates task-relevant context, selects tools and models, and delegates execution via on-the-fly automatic agent creation. Such designs enable reducing human engineering efforts, and remain framework-agnostic with plug-and-play support for diverse agents as task executors. It also enables a controllable performance-cost trade-off, allowing the system to approach Pareto-efficient. Across three challenging benchmarks (GAIA, SWE-Bench, Terminal-Bench), AOrchestra achieves 16.28% relative improvement against the strongest baseline when paired with Gemini-3-Flash. The code is available at: https://github.com/FoundationAgents/AOrchestra

Low-Complexity Distributed Combining Design for Near-Field Cell-Free XL-MIMO Systems

In this paper, we investigate the low-complexity distributed combining scheme design for near-field cell-free extremely large-scale multiple-input-multiple-output (CF XL-MIMO) systems. Firstly, we construct the uplink spectral efficiency (SE) performance analysis framework for CF XL-MIMO systems over centralized and distributed processing schemes. Notably, we derive the centralized minimum mean-square error (CMMSE) and local minimum mean-square error (LMMSE) combining schemes over arbitrary channel estimators. Then, focusing on the CMMSE and LMMSE combining schemes, we propose five low-complexity distributed combining schemes based on the matrix approximation methodology or the symmetric successive over relaxation (SSOR) algorithm. More specifically, we propose two matrix approximation methodology-aided combining schemes: Global Statistics \& Local Instantaneous information-based MMSE (GSLI-MMSE) and Statistics matrix Inversion-based LMMSE (SI-LMMSE). These two schemes are derived by approximating the global instantaneous information in the CMMSE combining and the local instantaneous information in the LMMSE combining with the global and local statistics information by asymptotic analysis and matrix expectation approximation, respectively. Moreover, by applying the low-complexity SSOR algorithm to iteratively solve the matrix inversion in the LMMSE combining, we derive three distributed SSOR-based LMMSE combining schemes, distinguished from the applied information and initial values.

Statistics Approximation-Enabled Distributed Beamforming for Cell-Free Massive MIMO

We study a distributed beamforming approach for cell-free massive multiple-input multiple-output networks, referred to as Global Statistics \& Local Instantaneous information-based minimum mean-square error (GSLI-MMSE). The scenario with multi-antenna access points (APs) is considered over three different channel models: correlated Rician fading with fixed or random line-of-sight (LoS) phase-shifts, and correlated Rayleigh fading. With the aid of matrix inversion derivations, we can construct the conventional MMSE combining from the perspective of each AP, where global instantaneous information is involved. Then, for an arbitrary AP, we apply the statistics approximation methodology to approximate instantaneous terms related to other APs by channel statistics to construct the distributed combining scheme at each AP with local instantaneous information and global statistics. With the aid of uplink-downlink duality, we derive the respective GSLI-MMSE precoding schemes. Numerical results showcase that the proposed GSLI-MMSE scheme demonstrates performance comparable to the optimal centralized MMSE scheme, under the stable LoS conditions, e.g., with static users having Rician fading with a fixed LoS path.

Low-Complexity Multi-Agent Continual Learning for Stacked Intelligent Metasurface-Assisted Secure Communications

Stacked intelligent metasurfaces (SIMs), composed of multiple layers of reconfigurable transmissive metasurfaces, are gaining prominence as a transformative technology for future wireless communication security. This paper investigates the integration of SIM into multi-user multiple-input multiple-output (MIMO) systems to enhance physical layer security. A novel system architecture is proposed, wherein each base station (BS) antenna transmits a dedicated single-user stream, while a multi-layer SIM executes wave-based beamforming in the electromagnetic domain, thereby avoiding the need for complex baseband digital precoding and significantly reducing hardware overhead. To maximize the weighted sum secrecy rate (WSSR), we formulate a joint precoding optimization problem over BS power allocation and SIM phase shifts, which is high-dimensional and non-convex due to the complexity of the objective function and the coupling among optimization variables. To address this, we propose a manifold-enhanced heterogeneous multi-agent continual learning (MHACL) framework that incorporates gradient representation and dual-scale policy optimization to achieve robust performance in dynamic environments with high demands for secure communication. Furthermore, we develop SIM-MHACL (SIMHACL), a low-complexity learning template that embeds phase coordination into a product manifold structure, reducing the exponential search space to linear complexity while maintaining physical feasibility. Simulation results validate that the proposed framework achieves millisecond-level per-iteratio ntraining in SIM-assisted systems, significantly outperforming various baseline schemes, with SIMHACL achieving comparable WSSR to MHACL while reducing computation time by 30\%.

XGuardian: Towards Explainable and Generalized AI Anti-Cheat on FPS Games

Aim-assist cheats are the most prevalent and infamous form of cheating in First-Person Shooter (FPS) games, which help cheaters illegally reveal the opponent's location and auto-aim and shoot, and thereby pose significant threats to the game industry. Although a considerable research effort has been made to automatically detect aim-assist cheats, existing works suffer from unreliable frameworks, limited generalizability, high overhead, low detection performance, and a lack of explainability of detection results. In this paper, we propose XGuardian, a server-side generalized and explainable system for detecting aim-assist cheats to overcome these limitations. It requires only two raw data inputs, pitch and yaw, which are all FPS games' must-haves, to construct novel temporal features and describe aim trajectories, which are essential for distinguishing cheaters and normal players. XGuardian is evaluated with the latest mainstream FPS game CS2, and validates its generalizability with another two different games. It achieves high detection performance and low overhead compared to prior works across different games with real-world and large-scale datasets, demonstrating wide generalizability and high effectiveness. It is able to justify its predictions and thereby shorten the ban cycle. We make XGuardian as well as our datasets publicly available.

Using Large Language Models to Detect Socially Shared Regulation of Collaborative Learning

The field of learning analytics has made notable strides in automating the detection of complex learning processes in multimodal data. However, most advancements have focused on individualized problem-solving instead of collaborative, open-ended problem-solving, which may offer both affordances (richer data) and challenges (low cohesion) to behavioral prediction. Here, we extend predictive models to automatically detect socially shared regulation of learning (SSRL) behaviors in collaborative computational modeling environments using embedding-based approaches. We leverage large language models (LLMs) as summarization tools to generate task-aware representations of student dialogue aligned with system logs. These summaries, combined with text-only embeddings, context-enriched embeddings, and log-derived features, were used to train predictive models. Results show that text-only embeddings often achieve stronger performance in detecting SSRL behaviors related to enactment or group dynamics (e.g., off-task behavior or requesting assistance). In contrast, contextual and multimodal features provide complementary benefits for constructs such as planning and reflection. Overall, our findings highlight the promise of embedding-based models for extending learning analytics by enabling scalable detection of SSRL behaviors, ultimately supporting real-time feedback and adaptive scaffolding in collaborative learning environments that teachers value.