arXiv:2507.05257v4 Announce Type: replace Abstract: Recent benchmarks for Large Language Model (LLM) agents primarily focus on evaluating reasoning, planning, and execution capabilities, while another critical component-memory, encompassing how agents memorize, update, and retrieve long-term information-is under-evaluated due to the lack of benchmarks. We term agents with memory mechanisms as memory agents. In this paper, based on classic theories from memory science and cognitive science, we identify four core competencies essential for memory agents: accurate retrieval, test-time learning, long-range understanding, and selective forgetting. Existing benchmarks either rely on limited context lengths or are tailored for static, long-context settings like book-based QA, which do not reflect the interactive, multi-turn nature of memory agents that incrementally accumulate information. Moreover, no existing benchmarks cover all four competencies. We introduce MemoryAgentBench, a new benchmark specifically designed for memory agents. Our benchmark transforms existing long-context datasets and incorporates newly constructed datasets into a multi-turn format, effectively simulating the incremental information processing characteristic of memory agents. By carefully selecting and curating datasets, our benchmark provides comprehensive coverage of the four core memory competencies outlined above, thereby offering a systematic and challenging testbed for assessing memory quality. We evaluate a diverse set of memory agents, ranging from simple context-based and retrieval-augmented generation (RAG) systems to advanced agents with external memory modules and tool integration. Empirical results reveal that current methods fall short of mastering all four competencies, underscoring the need for further research into comprehensive memory mechanisms for LLM agents.
Science Journals
arXiv:2511.01472v2 Announce Type: replace Abstract: The rapid progress of vision--language models (VLMs) has sparked growing interest in robotic control, where natural language can express the operation goals while visual feedback links perception to action. However, directly deploying VLM-driven policies on aerial manipulators remains unsafe and unreliable since the generated actions are often inconsistent, hallucination-prone, and dynamically infeasible for flight. In this work, we present AERMANI-VLM, the first framework to adapt pretrained VLMs for aerial manipulation by separating high-level reasoning from low-level control, without any task-specific fine-tuning. Our framework encodes natural language instructions, task context, and safety constraints into a structured prompt that guides the model to generate a step-by-step reasoning trace in natural language. This reasoning output is used to select from a predefined library of discrete, flight-safe skills, ensuring interpretable and temporally consistent execution. By decoupling symbolic reasoning from physical action, AERMANI-VLM mitigates hallucinated commands and prevents unsafe behavior, enabling robust task completion. We validate the framework in both simulation and hardware on diverse multi-step pick-and-place tasks, demonstrating strong generalization to previously unseen commands, objects, and environments.
arXiv:2511.02253v3 Announce Type: replace Abstract: We present a predictive master spectrum describing turbulence-like flows in microfluidic systems. Extending Pao's viscous-range closure, the model introduces (i) an adaptive inertial-range slope dependent on measurable dimensionless numbers and (ii) a physics-specific cutoff that captures entropy-producing sinks such as electrokinetic forcing, compliant walls, active stresses, and interfacial tension. This formulation unifies turbulence regimes -- electrokinetic, active, interfacial, and compressible -- within one compact expression. Comparison with reported data reproduces both spectral slopes and dissipation cutoffs while requiring only global observables (velocity, viscosity, Taylor microscale, and forcing strength). The framework provides a design-level predictive tool for turbulent microflows prior to computationally heavy DNS or CFD.
arXiv:2501.17559v2 Announce Type: replace Abstract: After the achievement of solving two-player zero-sum games, more AI researchers focus on solving multiplayer games. Urban Network Security Games (\textbf{UNSGs}) represent a class of such games, modeling real-world scenarios where law enforcement must strategically allocate limited resources to intercept criminals escaping within urban networks, and have gained considerable research attention. However, progress in this field has been limited by the absence of a standardized experimental platform and realistic benchmarks with heterogeneous travel costs. To address this limitation, we introduce \textbf{GraphChase}, an open-source platform designed to support the development and evaluation of algorithms for UNSGs. GraphChase offers a unified environment for modeling diverse UNSG variants on unweighted and weighted road networks across urban topologies. It also incorporates learning-based algorithms as baseline references for researchers. Furthermore, our experiments with GraphChase reveal that existing approaches to UNSGs still face challenges in terms of robustness and scalability, and suffer performance degradation when deployed under weighted edge costs, highlighting a sim-to-real generalization gap. GraphChase thus provides a realistic testbed for developing and validating UNSGs solvers under realistic travel-time heterogeneity.
arXiv:2502.10687v3 Announce Type: replace Abstract: Integrated sensing and communication (ISAC) has garnered substantial research interest owing to its pivotal role in advancing the development of next-generation (6G) wireless networks. However, achieving a performance balance between communication and sensing in the dual-function radar communication (DFRC)-based ISAC system remains a significant challenge. In this paper, a low-altitude intelligent reflecting surface (IRS)-assisted ISAC system is explored, where a base station (BS) supports dual-functional operations, enabling both data transmission for multiple users and sensing for a blocked target, with the channel quality enhanced by an IRS mounted on the unmanned aerial vehicle (UAV). Moreover, we formulate an integrated communication, sensing, and energy efficiency multi-objective optimization problem (CSEMOP), which aims to maximize the communication rate of the users and the sensing rate of the target, while minimizing UAV propulsion energy consumption by jointly optimizing the BS beamforming matrix, IRS phase shifts, the flight velocity and angle of the UAV. Considering the non-convexity, trade-off, and dynamic nature of the formulated CSEMOP, we propose a generative diffusion model-based deep deterministic policy gradient (GDMDDPG) algorithm to solve the problem. Specifically, the diffusion model is incorporated into the actor network of DDPG to improve the action quality, with noise perturbation mechanism for better exploration and recent prioritized experience replay (RPER) sampling mechanism for enhanced training efficiency. Simulation results indicate that the GDMDDPG algorithm delivers superior performance compared to the existing methods.
arXiv:2502.11491v3 Announce Type: replace Abstract: Large language models (LLMs) have shown remarkable capabilities in natural language processing. However, in knowledge graph question answering tasks (KGQA), there remains the issue of answering questions that require multi-hop reasoning. Existing methods rely on entity vector matching, but the purpose of the question is abstract and difficult to match with specific entities. As a result, it is difficult to establish reasoning paths to the purpose, which leads to information loss and redundancy. To address this issue, inspired by human reverse thinking, we propose Ontology-Guided Reverse Thinking (ORT), a novel framework that constructs reasoning paths from purposes back to conditions. ORT operates in three key phases: (1) using LLM to extract purpose labels and condition labels, (2) constructing label reasoning paths based on the KG ontology, and (3) using the label reasoning paths to guide knowledge retrieval. Experiments on the WebQSP and CWQ datasets show that ORT achieves state-of-the-art performance and significantly enhances the capability of LLMs for KGQA.
arXiv:2505.07124v3 Announce Type: replace Abstract: We study inverse problems where an unknown potential is observed only through samples from the measure it induces by a convex variational principle. Such problems arise in learning costs, energies, and dynamics from distributional data, but the associated forward solution map is typically nonlinear and implicit. We show that its optimality gap nevertheless yields convex empirical objectives for finite-dimensional potential classes, and we introduce sharpened Fenchel--Young losses that add a data-dependent discrepancy inside the forward problem. This keeps the estimator calibrated while improving the local geometry of the loss. Our main stability theorem separates the inverse error analysis into measurement error, forward perturbation, and empirical curvature. We instantiate this principle for inverse entropic unbalanced optimal transport and for inverse Jordan--Kinderlehrer--Otto (JKO) learning from independent snapshot samples, obtaining high-probability parameter recovery bounds. JKO schemes discretize Wasserstein gradient flows through a sequence of variational problems over measures, making them a natural language for population dynamics observed through snapshots. In this JKO case, the sharpened objective reduces to an unbalanced transport problem, which also clarifies the connection between variational gap losses and quadratic iJKO\(^\star\) surrogates. Numerical experiments illustrate the conditioning effect of sharpening and its benefits for sparse inverse-gradient-flow recovery.
arXiv:2511.02734v3 Announce Type: replace Abstract: Current evaluations of Large Language Model (LLM) agents primarily emphasize task completion, often overlooking resource efficiency and adaptability. This neglects a crucial capability: agents' ability to devise and adjust cost-optimal plans in response to changing environments. To bridge this gap, we introduce CostBench, a scalable, cost-centric benchmark designed to evaluate agents' economic reasoning and replanning abilities. Situated in the travel-planning domain, CostBench comprises tasks solvable via multiple sequences of atomic and composite tools with diverse, customizable costs. It also supports four types of dynamic blocking events, such as tool failures and cost changes, to simulate real-world unpredictability and necessitate agents to adapt in real time. Evaluating leading open-sourced and proprietary models on CostBench reveals a substantial gap in cost-aware planning: agents frequently fail to identify cost-optimal solutions in static settings, with even GPT-5 achieving less than 75% exact match rate on the hardest tasks, and performance further dropping by around 40% under dynamic conditions. By diagnosing these weaknesses, CostBench lays the groundwork for developing future agents that are both economically rational and robust.
arXiv:2503.00539v2 Announce Type: replace Abstract: Reinforcement learning from human feedback (RLHF) has evolved to be one of the main methods for fine-tuning large language models (LLMs). However, existing RLHF methods are non-robust, and their performance deteriorates if the downstream task differs significantly from the preference dataset used in fine-tuning. In order to mitigate this problem, we introduce a distributionally robust RLHF for fine-tuning LLMs. In particular, our goal is to ensure that a fine-tuned model retains its performance even when the distribution of prompts significantly differs from the distribution encountered during fine-tuning. We formulate distributionally robust optimization (DRO) version of two popular fine-tuning methods -- (1) reward-based RLHF and (2) reward-free DPO (direct preference optimization). We propose a minibatch gradient descent based algorithms for both of them, and theoretically prove convergence guarantees for the algorithms. Subsequently, we evaluate our algorithms on an out-of-distribution (OOD) task by first training the model on the Unified-Feedback dataset and evaluating its performance on two different datasets. The experimental results show that our robust training improves the accuracy of the learned reward models on average, and markedly on some tasks, such as reasoning. Furthermore, we show that the robust versions of policy optimization methods, similarly improve performance on OOD tasks.
arXiv:2503.19797v3 Announce Type: replace Abstract: Property-based testing (PBT) relies on generators for random test cases, often constructed using embedded domain specific languages, which provide expressive combinators for building and composing generators. The effectiveness of PBT depends critically on the speed of these generators. However, careful measurements show that the generator performance of widely used PBT libraries falls well short of what is possible, due principally to (1) the abstraction overhead of their combinator-heavy style and (2) suboptimal sources of randomness. We characterize, quantify, and address these bottlenecks. To eliminate abstraction overheads, we propose a technique based on multi-stage programming, dubbed Allegro. We apply this technique to leading generator libraries in OCaml and Scala 3, significantly improving performance. To quantify the performance impact of the randomness source, we carry out a controlled experiment, replacing the randomness in the OCaml PBT library with an optimized version. Both interventions exactly preserve the semantics of generators, enabling precise, pointwise comparisons. Together, these improvements find bugs up to $13\times$ faster.
arXiv:2511.04080v4 Announce Type: replace Abstract: Generative Search Engine (GSE) leverages the Retrieval-Augmented Generation (RAG) technique and the Large Language Model (LLM) to integrate multi-source information and provide users with accurate and comprehensive responses. Unlike traditional search engines that present results in ranked lists, GSE shifts users' attention from sequential browsing to content-driven subjective perception, not only driving a paradigm shift in information retrieval but also highlighting the importance of enhancing the subjective visibility of content in generative search. In this context, Generative Search Engine Optimization (G-SEO) methods have emerged as a new research focus. With the rapid advancement of Multimodal Retrieval-Augmented Generation (MRAG) techniques, GSE can now efficiently integrate text, images, audio, and video, producing richer responses that better satisfy complex information needs. Existing G-SEO methods, however, remain limited to text-based optimization and fail to fully exploit multimodal data. To address this gap, we propose Caption Injection, the first multimodal G-SEO approach, which extracts captions from images and injects them into textual content, integrating visual semantics to enhance the subjective visibility in generative search. We systematically evaluate Caption Injection on MRAMG, a benchmark for MRAG, under both unimodal and multimodal settings. Experimental results show that Caption Injection significantly outperforms text-only G-SEO baselines under the G-EVAL metric, effectively improving the subjective visibility of content perceived by users, and demonstrating the practical benefits of multimodal information in G-SEO. The source code for this work is openly available at https://github.com/GrayChan04/Caption-Injection.
arXiv:2511.05567v2 Announce Type: replace Abstract: An adaptive structural learning method of Restricted Boltzmann Machine (RBM) and Deep Belief Network (DBN) has been developed as one of prominent deep learning models. The neuron generation-annihilation algorithm in RBM and layer generation algorithm in DBN make an optimal network structure for given input during the learning. In this paper, our model is applied to an automatic recognition method of road network system, called RoadTracer. RoadTracer can generate a road map on the ground surface from aerial photograph data. A novel method of RoadTracer using the Teacher-Student based ensemble learning model of Adaptive DBN is proposed, since the road maps contain many complicated features so that a model with high representation power to detect should be required. The experimental results showed the detection accuracy of the proposed model was improved from 40.0\% to 89.0\% on average in the seven major cities among the test dataset. In addition, we challenged to apply our method to the detection of available roads when landslide by natural disaster is occurred, in order to rapidly obtain a way of transportation. For fast inference, a small size of the trained model was implemented on a small embedded edge device as lightweight deep learning. We reported the detection results for the satellite image before and after the rainfall disaster in Japan. This version of the article was improved the search algorithm at the border around image.
arXiv:2511.05715v3 Announce Type: replace Abstract: We study autonomous pickup-and-delivery routing problems in which internal adversarial agents spoof their locations to attract request assignments and then intentionally leave those requests unserviced. Such attacks disrupt the centralized scheduler, causing delays, cancellations, and routing instability. A routing policy is stable if its cost remains uniformly bounded over time. Existing policy-cost formulations typically characterize cost through the work required to service outstanding requests. Such a formulation requires analyzing agent-specific route execution and is therefore not well suited to adversarial settings, where non-cooperative agents may arbitrarily deviate from assigned routes or fail to service requests altogether. We introduce a new policy-cost formulation based only on observable system signals, namely the numbers of outstanding and canceled requests. Under bounded arrivals and finite request time windows, we show that stability under this formulation is equivalent to keeping the expected cumulative number of canceled requests uniformly bounded over time, an important operational metric in both cooperative and adversarial settings. We also extend cooperative fleet-sizing guarantees to finite time-window settings and highlight that request time windows are not merely a modeling detail, but are essential for ruling out \emph{degenerate stability}, a regime in which policies are certified as stable despite undesirable large request backlogs.
arXiv:2511.05934v2 Announce Type: replace Abstract: Generative modeling frameworks have emerged as an effective approach to capture high-dimensional image distributions from large datasets without requiring domain-specific knowledge, a capability essential for disease progression modeling. Recent generative approaches have attempted to capture progression by mapping images to a latent space and guiding representations to generate follow-up images from previous time points. However, these methods impose constraints on distribution learning, resulting in latent spaces with limited controllability for generating follow-up images without paired subject-specific longitudinal guidance. In order to enable controlled movements in the latent representational space and generate progression images from a previous time-point image without subject-specific guidance, we introduce a conditionable Diffusion Auto-encoder framework that forms a compact latent space capturing high-level semantics and providing means to control generation. Our approach leverages this latent space to condition and apply controlled shifts to the representations of previous time-point images by isolating progression and subject identity information for generating follow-up images. The shifts are implicitly guided by correlating with progression attributes and constraining to Alzheimer's disease specific regions, without paired longitudinal guidance. We validate the generations through image quality metrics, volumetric progression analysis, and downstream tasks in Alzheimer's disease datasets from different sources. This demonstrates the effectiveness of our approach for Alzheimer's progression modeling and longitudinal image generation.
arXiv:2503.21661v3 Announce Type: replace Abstract: Ontologies enable knowledge sharing and interdisciplinary collaboration by providing standardized, structured vocabularies for diverse communities. While logical axioms are a cornerstone of ontology design, natural language elements such as annotations are equally critical for conveying intended meaning and ensuring consistent term usage. This paper explores how meaning is represented in ontologies and how it can be effectively represented and communicated, addressing challenges such as indeterminacy of reference and meaning holism. To this end, instead of following the conventional approach of beginning with existing ontologies and working toward alignment or modularization, this article proposes a reversal of perspective: taking the ontological term as the starting point and introducing a new structure, named 'ontological unit', characterized by: a term-centered design; enhanced characterization of both formal and natural language statements; and an operationalizable definition of communicated meaning based on general assertions. By formalizing the meaning of ontological units, this work seeks to enhance the semantic robustness of terms, improving their clarity and accessibility across domains. Furthermore, it may offer a more effective foundation for ontology generation and significantly improves support for key maintenance tasks such as reuse and versioning. This article aims to establish the theoretical groundwork for the proposed approach and to lay the foundations for future applications in applied ontologies.
arXiv:2504.14969v2 Announce Type: replace Abstract: This paper proposes a framework for evaluating large language models (LLMs) on Chinese topic constructions, focusing on their sensitivity to island constraints. Drawing inspiration from Tian et al. (2024), we outline an experimental design for testing LLMs' grammatical knowledge of Mandarin syntax. While no experiments have been conducted yet, this proposal aims to provide a foundation for future studies and invites feedback on the methodology.
arXiv:2509.15942v3 Announce Type: replace Abstract: Internal variability is a dominant contributor to the uncertainty of predictions at the interannual to decadal timescale. A typical approach to separating the internal variability from forced climate responses is to generate large ensembles of simulations under different initial conditions. Due to the complexity of Earth System Models, generating these large ensembles is computationally expensive. In this work, we present ArchesClimate, a deep learning-based climate model emulator designed to reduce the cost of exploring internal variability at timescales ranging from monthly to decadal. ArchesClimate is trained on decadal hindcasts of the IPSL-CM6A-LR climate model. We train a flow matching model following ArchesWeatherGen, which we adapt to predict near-term climate. Once trained, the model generates states at a one-month lead time from the states of the two preceding months, and can be used to auto-regressively emulate climate model simulations. We show that for up to 10 years, these generations are stable and physically consistent. We also show that for several important climate variables, ArchesClimate generates simulations that are interchangeable with the IPSL model. This work suggests that climate model emulators could reduce the cost of generating large ensembles with climate models.
arXiv:2511.06090v3 Announce Type: replace Abstract: Optimizing the performance of large-scale software repositories demands expertise in code reasoning and software engineering (SWE) to reduce runtime while preserving program correctness. However, most benchmarks emphasize what to fix rather than how to fix code. We introduce SWE-fficiency, a benchmark for evaluating repository-level performance optimization on real workloads. Our suite contains 498 tasks across nine widely used data-science, machine-learning, and HPC repositories (e.g., numpy, pandas, scipy): given a complete codebase and a slow workload, an agent must investigate code semantics, localize bottlenecks and relevant tests, and produce a patch that matches or exceeds expert speedup while passing the same unit tests. To enable this how-to-fix evaluation, our automated pipeline scrapes GitHub pull requests for performance-improving edits, combining keyword filtering, static analysis, coverage tooling, and execution validation to both confirm expert speedup baselines and identify relevant repository unit tests. Empirical evaluation of state-of-the-art agents reveals significant underperformance. On average, agents achieve less than 0.23x the expert speedup: agents struggle in localizing optimization opportunities, reasoning about execution across functions, and maintaining correctness in proposed edits. We release the benchmark and accompanying data pipeline to facilitate research on automated performance engineering and long-horizon software reasoning.
arXiv:2511.08370v3 Announce Type: replace Abstract: This paper presents an $\mathcal{H}_\infty$ model matching control-based approach to the problem of power hardware-in-the-loop (PHIL) interfacing. The objective is to interconnect a grid simulation and a physical device via an interface in a way that is stable and accurate. Conventional approaches include the ideal transformer method (ITM) and its impedance-based variants, which trade accuracy for stability, as well as some $\mathcal{H}_\infty$ control-based approaches, which do not make use of all the available information in their optimization for accuracy. Designing for transparency, as opposed to accuracy as existing approaches do, would achieve both accuracy and stability, while making use of all the dynamical information present in the idealized interconnection of the grid and device. The approach proposed in this paper employs model matching to formulate the PHIL problem as an $\mathcal{H}_\infty$ control problem using transparency as the explicit frequency-domain control objective. The approach is experimentally validated in a real-time resistive-load PHIL setup, and is found to achieve accuracy levels that are comparable or superior to those of an ITM-based interface.
arXiv:2506.06597v2 Announce Type: replace Abstract: Side-channel attacks can recover neural network parameters from physical signals, even on commercial edge accelerators. Existing defenses require changes to hardware, instruction set, or compiler, and cannot be deployed on fixed-function platforms such as TPUs. We present the first training-time defense that protects models on off-the-shelf TPUs without modifying the hardware or firmware. Our approach trains multiple functionally equivalent parameter versions per layer and randomly composes them at inference. This reduces the correlation that side-channel attacks rely on while preserving model accuracy. We enforce diversity between parameter versions by adding a regularization term in the loss function during training. We show that this diversity increases leakage variance while leaving the mean signal unchanged, which provably reduces the signal-to-noise ratio exploited by attackers. We derive theoretical bounds that relate leakage to the number of parameter versions and their pairwise distance, and provide a simple calibration method to predict leakage for new configurations before deployment or side-channel measurements. We implement our method on a Google Edge TPU and evaluate it on representative and real-world models. Our defense, in a high-diversity configuration, can hide leakage by reducing the Test Vector Leakage Assessment t-score below the standard leakage detection threshold of 4.5 for the majority of a neural network, with less than 1% accuracy change and moderate overhead. Our results thus show, for the first time, that training-time defenses can provide practical side-channel protection for widely deployed AI hardware.
arXiv:2506.19045v5 Announce Type: replace Abstract: Fault localization (FL) is a critical step in debugging, which typically relies on repeated executions to pinpoint faulty code regions. However, repeated executions can be impractical in the presence of non-deterministic failures or high execution costs. While recent efforts have leveraged Large Language Models (LLMs) to aid execution-free FL, these have primarily focused on identifying faults in the system-under-test (SUT) rather than in the often complex system-level test code. However, the latter is also important, as in practice, many failures are triggered by faulty test code. To overcome these challenges, we introduce a fully static, LLM-driven approach for system-level test code fault localization (TCFL) that does not require executing the test case. Our method uses a single failure execution log to estimate the test's execution trace through three novel algorithms that identify only code statements likely involved in the failure. This pruned trace, combined with the error message, is used to prompt the LLM to rank potential faulty locations. Our black-box, system-level approach requires no access to the SUT source code and is applicable to complex test scripts that assess full system behavior. We evaluate our technique at the function, block, and line levels using an industrial dataset of faulty Python test cases that were not used in pre-training LLMs. Results show that our best-estimated traces closely match the actual traces, with an F1 score of around 90%. Additionally, pruning the complex system-level test code reduces the LLM's inference time by up to 34% without any loss in FL performance. Our method achieves equal or higher FL accuracy, requiring over 85% less average inference time per test case and 93% fewer tokens than the latest LLM-guided FL method.
arXiv:2506.15066v5 Announce Type: replace Abstract: As the complexity of integrated circuit designs continues to escalate, functional verification becomes increasingly challenging. Reference models, critical for accelerating the verification process, are themselves becoming more intricate and time-consuming to develop. Despite the promise shown by large language models (LLMs) in code programming, effectively generating complex reference models remains a significant hurdle. Therefore, we introduce ChatModel, an LLM-aided agile reference model generation and verification platform. ChatModel streamlines the transition from design specifications to fully functional reference models by integrating design standardization and hierarchical agile modeling. Employing a building-block generation strategy, it not only enhances the design capabilities of LLMs for reference models but also significantly boosts verification efficiency. We evaluated ChatModel on 300 designs of varying complexity, demonstrating substantial improvements in both efficiency and quality of reference model generation. ChatModel achieved a peak performance improvement of 58.99% compared to alternative methods, with notable enhancements in generation stability, and delivered a 9.18x increase in its capacity to produce reference model designs. Moreover, ChatModel accelerates the reference model design and validation cycles by an average of 7.11x over traditional manual approaches. These results highlight the potential of ChatModel to significantly advance the automation of reference model generation and validation.
arXiv:2511.10480v3 Announce Type: replace Abstract: Optimizing the performance of large language models (LLMs) on large-scale AI training and inference systems requires a scalable and expressive mechanism to model distributed workload execution. Such modeling is essential for pre-deployment system-level optimizations (e.g., parallelization strategies) and hardware design-space explorations. While recent efforts have proposed collecting execution traces from real systems, access to large-scale infrastructure remains limited to major cloud providers. Moreover, traces capturing execution on a specific platform cannot be easily adapted to study alternate software and/or hardware configurations, especially at scale. We introduce STAGE, a framework that synthesizes high-fidelity execution graphs to accurately model distributed AI workloads (including LLMs and MoEs). STAGE supports a comprehensive set of parallelization strategies, allowing users to systematically explore a wide spectrum of model architectures and system configurations. STAGE demonstrates its scalability by synthesizing high-fidelity LLM traces spanning over 128K GPUs, while preserving tensorlevel accuracy in compute, memory, and communication. STAGE is publicy available at https://github.com/astra-sim/stage
arXiv:2512.08505v2 Announce Type: replace Abstract: Conditional diffusion models frequently suffer from language-image misalignments. Due to the ambiguity of intermediate noise corrupted latents, assessing prompt adherence currently requires completing the entire sampling trajectory. This late-stage evaluation incurs even higher computational costs during test-time scaling strategies, such as Best-of-N (BoN) sampling, as all misaligned trajectories must finish generation before being discarded. To tackle this, we propose NoisyCLIP, a noise-aware twin-tower model that enables early language-to-latent alignment estimation. By learning a vision encoder on noise-corrupted latents, we allow the model to "see" through the ambiguity of intermediate diffusion steps. To facilitate this training, we investigate noise-data augmentation sampling strategies and introduce two new benchmark datasets: Noisy-Conceptual-Captions and Noisy-GenAI-Bench. When applied as an early-stopping criterion for BoN, NoisyCLIP at half cost matches or beats frozen CLIP at full cost. Ultimately, this transforms alignment assessment from an expensive final check into a continuous monitoring tool, drastically reducing compute costs without sacrificing semantic fidelity.
arXiv:2511.13216v2 Announce Type: replace Abstract: Deployment of legged robots for navigating challenging terrains (e.g., stairs, slopes, and unstructured environments) has gained increasing preference over wheel-based platforms. In such scenarios, accurate odometry estimation is a preliminary requirement for stable locomotion, localization, and mapping. Traditional proprioceptive approaches, which rely on leg kinematics sensor modalities and inertial sensing, suffer from irrepressible vertical drift caused by frequent contact impacts, foot slippage, and vibrations, particularly affected by inaccurate roll and pitch estimation. Existing methods incorporate exteroceptive sensors such as LiDAR or cameras. Further enhancement has been introduced by leveraging gravity vector estimation to add additional observations on roll and pitch, thereby increasing the accuracy of vertical pose estimation. However, these approaches tend to degrade in feature-sparse or repetitive scenes and are prone to errors from double-integrated IMU acceleration. To address these challenges, we propose GaRLILEO, a novel gravity-aligned continuous-time radar-leg-inertial odometry framework. GaRLILEO decouples velocity from the IMU by building a continuous-time ego-velocity spline from SoC radar Doppler and leg kinematics information, enabling seamless sensor fusion which mitigates odometry distortion. In addition, GaRLILEO can reliably capture accurate gravity vectors leveraging a novel soft S2-constrained gravity factor, improving vertical pose accuracy without relying on LiDAR or cameras. Evaluated on a self-collected real-world dataset with diverse indoor-outdoor trajectories, GaRLILEO demonstrates state-of-the-art accuracy, particularly in vertical odometry estimation on stairs and slopes. We open-source both our dataset and algorithm to foster further research in legged robot odometry and SLAM. https://garlileo.github.io/GaRLILEO