arXiv:2605.15569v1 Announce Type: new Abstract: Microservices are widely adopted in modern cloud systems due to their scalability and fault tolerance. However, microservice architectures introduce significant complexity in privilege and permission control, creating risks of privilege escalation where attackers can gain unauthorized access to resources or operations. Detecting such vulnerabilities is challenging due to complex cross-service interactions, polyglot codebases, and diverse privileged operations and permission checks. We present Neo, an agentic program analysis framework that combines large language models (LLMs) with classic program analysis to address these challenges. Neo leverages an LLM-based agent that dynamically generates analysis plans, adapts code search strategies, and validates semantics. We develop code search primitives that enable Neo to perform scalable and flexible code exploration across services and languages. We evaluated Neo on 25 open-source microservice applications spanning 7 programming languages and 6.2 million lines of code. Neo uncovered 24 zero-day privilege escalation vulnerabilities and achieved 81.0% precision and 85.0% recall on a ground-truth dataset. Compared to existing program analysis and agentic solutions, Neo demonstrated significant improvements in both detection accuracy and scalability. We further showcased Neo's extensibility by applying it to other application domains and vulnerability types, uncovering 18 additional zero-day vulnerabilities.
Science Journals
arXiv:2605.15485v1 Announce Type: new Abstract: We report the first observation on the EXL-50U spherical torus that energetic particles injected by neutral beam injection (NBI) can be stably accelerated to significantly higher energies - reaching up to 2.5 times the injection energy, occurring without significant large-scale magnetohydrodynamic (MHD) bursts. Simulations based on EXL-50U parameters indicate that small-scale magnetic reconnection, mediated by multiple magnetic islands, fails to accelerate bulk thermal ions but efficiently energizes seed fast ions. Unlike global MHD events, such small-scale reconnection is ubiquitous in magnetic confinement devices and does not degrade core confinement. This mechanism offers a novel and potentially universal channel for auxiliary ion heating in future fusion reactors.
arXiv:2605.15608v1 Announce Type: new Abstract: Decoder-only transformers compute the conditional probability of the next token from a sequence of past observations. This paper derives, from first principles, inference architectures that solve the same prediction problem - and in doing so, recovers transformer-like layer operations as a consequence of optimal control theory. The framework is developed for two model classes: a nonlinear model of discrete-valued processes, directly motivated by the transformer, and a linear Gaussian model as a tractable baseline. For both model classes, the prediction objective is reformulated as an optimal control problem whose solution yields an explicit inference algorithm, the dual filter, with a layer structure that mirrors the layer structure of a decoder-only transformer. Numerical experiments provide a comparison of the optimal control to attention weights from a trained transformer. These experiments reveal that when the embedding dimension is insufficient, the transformer implicitly exploits non-Markovian structure.
arXiv:2605.15473v1 Announce Type: new Abstract: We propose using validated behavioral hypotheses as a lens for evaluating human-likeness in LLM-based agents. Our key idea is simple: If an agent is human-like, a population of such agents should reach the same inferential conclusion as the human population when run through the same experiment. Decades of social science have produced many such validated findings, each anchored to concrete experimental protocols and robustly established through independent replication. This yields an evaluation that is objective, decomposable, and scalable. We operationalize this lens through HumanStudy-Bench, an open platform that turns published human-subject studies into reusable simulation environments and administers the evaluation to configurable agents. It scores agent-human alignment on two metrics: the Probability Alignment Score (PAS) for inferential agreement and the Effect Consistency Score (ECS) for effect-size agreement. We curated an initial suite of 12 studies whose hypotheses are robustly established through independent replication, and evaluated 10 models under 4 agent designs. Results show that agent responses polarize between full replication and complete failure; agent design influences alignment more than model scale, but its effect is non-monotonic.
arXiv:2605.15227v1 Announce Type: new Abstract: Self-driving laboratories (SDLs) have attracted increasing attention as a means of accelerating scientific discovery; however, developing SDL software remains technically demanding. To improve accessibility, orchestration software frameworks have been proposed to coordinate SDL components. Nevertheless, existing frameworks are primarily designed for human interaction and do not provide standardized interfaces suitable for AI agents. In this work, we propose an SDL software architecture based on the Model Context Protocol (MCP), in which all SDL functionalities are exposed through MCP servers. Following this design principle, we introduce an MCP-based SDL orchestrator, named NIMO Controller. It provides a visual programming interface automatically generated through MCP-based tool discovery, allowing human users to design experimental workflows without writing code. The same MCP backend can also be accessed by AI agents, providing a unified interface for both human users and AI agents. We demonstrate the proposed system through a case study on a color-matching SDL. The results validate the usability of the proposed MCP-based SDL architecture.
arXiv:2605.15228v1 Announce Type: new Abstract: Modern cloud and enterprise systems rely on identity-centric authorization, assuming that callers possessing valid credentials are safe to execute commands. The emergence of autonomous AI agents invalidates this assumption: agents can generate syntactically valid but semantically unsafe actions, making standing privileges a significant operational risk. This risk becomes especially acute in sovereign AI systems, where autonomous agents may interact with cloud infrastructure, regulated data, financial workflows, and national-scale digital services. Governed mutation substrates reduce this risk by interposing on agent actions: agents submit intents, infrastructure evaluates context and policy, and execution is mediated. However, this shifts the trust boundary: how can the decision to authorize an intent be made verifiable, distributed, and replayable? We introduce a Distributed Trust Framework (DTF), a verification framework for governed mutation systems that computes execution authority from structured, verifiable artifacts. DTF introduces a Justification Proof to encode the admissibility basis of an action, a consensus model for independent evaluation, an ephemeral Execution Identity derived from the approved proof, and an append-only Evidence Chain that preserves the authorization lifecycle. Under stated substrate assumptions, this architecture enforces a compact authorization invariant: no high-stakes execution without a proof object, no derived authority without consensus, and no valid mutation detached from evidence. We define the model, instantiate it over an OpenKedge-based governed mutation substrate, and show how it maps onto cloud-native environments. By shifting authorization from standing identity to proof-derived authority, DTF provides an infrastructure foundation for making agentic execution governable, auditable, and bounded in sovereign AI deployments.
arXiv:2506.14829v2 Announce Type: replace Abstract: AI for Social Impact (AI4SI) is an emergent field harnessing interdisciplinarities between the fields of artificial intelligence (AI), machine learning (ML), and the social sciences to address societal issues aligned with the United Nations Sustainable Development Goals (UN SDGs), such as universal healthcare, climate action, etc. Despite AI4SI's rising popularity, achieving tangible, on-the-ground impact remains a significant challenge. In particular, identifying collaborators open to co-designing and deploying AI4SI-based solutions in real-world settings is often difficult. Thus, many projects stall at the proof-of-concept stage, unable to scale to production-level deployment. Drawing on twenty-six AI4SI researchers' interviews, primarily from academic institutions though also including some industry researchers and practitioners, and the authors' own lived experiences, this paper employs thematic analysis to highlight structural, organizational, communication, collaboration, and operational challenges hindering socially impactful AI4SI deployments. While there are no easy fixes, the authors synthesize best practices and actionable strategies from interviews and personal experiences, positioning this paper as a practical guide for AI4SI researchers and organizations pursuing socially impactful collaborations$^1$. $^1$We note that our findings are most directly applicable to academic research groups in the global north, as governmental, startup, and global south researchers' perspectives are underrepresented in our sample.
arXiv:2603.01290v3 Announce Type: replace Abstract: The 2026 Formula 1 technical regulations introduce a fundamental change to energy strategy: under a 50/50 internal combustion engine / battery power split with unlimited regeneration and a driver-controlled Override Mode, the optimal energy deployment policy depends not only on a driver's own state but on the hidden state of rival cars. This creates a Partially Observable Stochastic Game that cannot be solved by single-agent optimisation methods. We present a tractable two-layer inference and decision framework. The first layer is a 40-state Hidden Markov Model (HMM) that infers a probability distribution over each rival's ERS charge level (four modes: H, M, L_harvest, L_derate), Override Mode status, and tyre degradation state from six publicly observable telemetry signals. The second layer is a Deep Q-Network (DQN) policy that takes the HMM belief state as input and selects between energy deployment strategies. We formally characterise the counter-harvest trap, a deceptive strategy in which a car deliberately suppresses observable deployment signals to induce a rival into a failed attack, and show that detecting it requires belief-state inference over both ERS level and the harvest/derate sub-mode. On synthetic races, the HMM achieves 96.8% ERS-level accuracy (random baseline 25%), classifies L_harvest vs. L_derate with 89.4% accuracy, and detects counter-harvest trap conditions with 96.3% recall. Pre-season analysis indicates circuit-dependent recharge availability (1.0x to 2.2x per lap) as the primary confound; Melbourne is the hardest-case validation environment. Baum-Welch calibration on 2026 race telemetry begins with the Australian Grand Prix (8 March 2026).
arXiv:2605.15406v1 Announce Type: new Abstract: This work presents a new approach for implementing polymorphism for bottom-up relational languages, without monomorphization. We begin by introducing semiringKanren, a bottom-up weighted relational programming language. We extend this base language to support polymorphism. We describe a new method to compile polymorphic semiringKanren programs into non-polymorphic ones, based on equality patterns and large-enough instances of polymorphic relations. We prove the correctness of this method. Finally, we consider existing work and suggest directions for future research.
arXiv:2506.15255v3 Announce Type: replace Abstract: Scalar mixing fronts develop at the interface of agitated fluids of different solute concentrations. In such fronts, scalar fluctuations form at both microscopic and macroscopic scales, due to stretching-enhanced molecular diffusion and hydrodynamic dispersion respectively. While these two elementary processes are well understood separately, predicting how their coupling governs the evolution of concentration statistics within dispersing fronts remains a challenge. Here, we propose a theoretical framework to describe scalar fluctuations in fronts mixed by smooth chaotic flows. We find that the transfer of energy between the macroscopic and microscopic scalar fluctuation scales operates at a characteristic length scale $s_i$, for which dispersion and stretching-enhanced diffusion are of equal strength. This leads to a closed expression for the concentration variance, which captures the results of direct numerical simulations with no fitting parameters, for a broad range of P\'eclet numbers. These findings open a new avenue for predicting both conservative and reactive mixing in smooth chaotic flows such as porous media or microfluidic flows.
arXiv:2605.15617v1 Announce Type: new Abstract: Large language model (LLM) training today runs on clusters spanning thousands of GPUs. While this scale enables rapid model advances, developing, debugging, and performance-tuning the training framework inevitably becomes complex and costly. This is because engineers often need to reproduce production behaviors to diagnose failures or evaluate optimizations, thereby demanding frequent and even exclusive access to production-scale clusters -- which becomes increasingly hard given that the majority of GPUs are already committed to production workloads. Simulation relies on complex performance models that are difficult to maintain, and downscaled experiments often fail to capture scale-dependent behaviors. We present PrismLLM to decouple large-scale execution from the need to access large clusters, enabling engineers to run and observe ranks of interest under faithful large-scale behavior using only a few GPUs. PrismLLM constructs a high-fidelity execution graph via a slicing-based approach that captures computation, communication, and dependencies of the target scale. Then, PrismLLM performs hybrid emulation where selected ranks execute the original program while the remaining ranks are replayed as virtual participants. Experiments on large-scale LLM training workloads show that PrismLLM accurately reproduces performance and memory behavior, achieving only 0.58\% average error in iteration time and less than 0.01\% error in peak GPU memory usage. PrismLLM can emulate clusters of up to 8192 GPUs using fewer than 1\% of the physical GPUs required by the original deployment.
arXiv:2602.14092v2 Announce Type: replace Abstract: Operating complex real-world systems, such as soft robots, can benefit from precise predictive control schemes that require accurate state and model knowledge. This knowledge is typically not available in practical settings and must be inferred from noisy measurements. In particular, it is challenging to simultaneously estimate unknown states and learn a model online from sequentially arriving measurements. In this paper, we show how a recently proposed gray-box system identification tool enables the estimation of a soft robot's current pose while at the same time learning a bending stiffness model. For estimation and learning, we only need a nominal constant-curvature robot model and measurements of the robot's base reactions (e.g., base forces). The estimation scheme -- relying on a marginalized particle filter -- allows us to conveniently interface nominal constant-curvature equations with a Gaussian Process (GP) bending stiffness model to be learned. This, in contrast to estimation via a random walk over stiffness values, enables prediction of bending stiffness and improves overall model quality. We demonstrate, using a real-world soft robot, that the method learns a bending-stiffness model online while accurately estimating the robot's pose. Notably, reduced error in multi-step forward predictions indicates that the learned bending-stiffness GP improves overall model quality.
arXiv:2605.16227v1 Announce Type: new Abstract: Deep Neural Networks (DNNs) are high-value intellectual property (IP), yet deploying them to edge environments exposes them to \textbf{unrestricted oracle access}, rendering them vulnerable to model extraction and inversion attacks. Existing defenses fail to address this practically: passive watermarking only offers post-hoc provenance, while active defenses impose prohibitive latency or require persistent access to sensitive training data. To bridge this gap, we propose \textit{LymphNode}, a novel post-hoc defense framework that acts as an intrinsic ``immune system" within the model. \textit{LymphNode} enforces a strict ``default-deny'' policy: it actively neutralizes model utility for unauthorized queries via \textbf{Generalized Sparse Universal Adversarial Perturbations (GSUAP)} injected into the feature space, effectively blocking gradient estimation and data inference. Utility is selectively restored only for authorized inputs carrying a stealthy feature-domain credential. Our framework is highly practical: it is \textbf{data-efficient}, establishing robust protection with fewer than 100 samples ($<1\%$ of training data), and \textbf{cross-dataset adaptable}, enabling protection using public surrogate datasets. \textit{LymphNode} thus provides a lightweight, immediately deployable defense for high-stakes scenarios where original training data is restricted or unavailable.
arXiv:2605.09391v2 Announce Type: replace Abstract: It is becoming increasingly necessary to have monitors check for harmful behaviors during language model interactions, but text-only monitoring has not been sufficient. This is because models sometimes exhibit strategic deception and sandbagging, changing their behavior during evaluation. This motivates the use of white-box monitors like linear probes, which can read the model internals directly. Currently, such probes can fail under distribution shift, limiting their usefulness in real settings. We study whether there exists a low-dimensional subspace of the model internals that captures harmful behaviors more robustly, while leaving out spuriously correlative features. Inspired by the Assistant Axis and Persona Selection Model, we construct persona axes for deception and sycophancy using contrastive persona prompts. The first principal components, obtained by unsupervised PCA of the persona-specific vectors, cleanly separate harmful and harmless personas. Across 10 evaluation datasets, we show that persona-derived directions transfer non-trivially and probes trained on persona-PC projections generalize better than probes trained on raw activations. We also find that a unified axis consisting of multiple harmful and harmless behaviors improves generalization across behaviors and datasets. Overall, persona vectors provide a useful inductive bias for building more transferable behavior probes.
arXiv:2605.15817v1 Announce Type: new Abstract: Understanding how travelers form overall evaluations of public transport journeys is critical for improving travel satisfaction and encouraging sustainable mode choice. While travel satisfaction is discussed to influence attitudes and future behavior, the cognitive rules by which moment-to-moment experiences are aggregated into retrospective evaluations remain poorly understood in transport research. Drawing on psychological theories of experienced and remembered utility, this study investigates which temporal aggregation heuristics best predict post-trip travel satisfaction. Using a smartphone-based experience sampling approach, we collected high-frequency on-trip experience ratings and post-trip evaluations for 2576 real-world public transport trips across three German cities. Travel experience was assessed every five minutes during trips using a multi-item scale, allowing direct comparison of competing aggregation rules, including mean experience, peak-end, minimum-end, final moment, and trip duration. Multilevel regression models were estimated to evaluate the explanatory power of each heuristic. Results show that retrospective travel satisfaction is best predicted by a Minimum-End heuristic, combining the most negative moment of the journey and the final experience. Models based on mean experience, peak-end rules, final moment alone, or trip duration performed substantially worse. This pattern indicates that both negative extremes and the final phase of a journey independently contribute to remembered evaluations, rather than overall satisfaction reflecting an average of momentary experiences. The results have important implications for theory and practice, suggesting that targeted interventions at critical negative moments and at trip endings may yield substantial improvements in remembered satisfaction and, ultimately, support shifts toward sustainable mobility.
arXiv:2605.10893v2 Announce Type: replace Abstract: Large vision-language models suffer from visual ungroundedness: they can produce a fluent, confident, and even correct response driven entirely by language priors, with the image contributing nothing to the prediction. Existing confidence estimation methods cannot detect this, as they observe model behavior under normal inference with no mechanism to determine whether a prediction was shaped by the image or by text alone. We introduce BICR (Blind-Image Contrastive Ranking), a model-agnostic confidence estimation framework that makes this contrast explicit during training by extracting hidden states from a frozen LVLM twice: once with the real image-question pair, and once with the image blacked out while the question is held fixed. A lightweight probe is trained on the real-image hidden state and regularized by a ranking loss that penalizes higher confidence on the blacked-out view, teaching it to treat visual grounding as a signal of reliability at zero additional inference cost. Evaluated across five modern LVLMs and seven baselines on a benchmark covering visual question answering, object hallucination detection, medical imaging, and financial document understanding, BICR achieves the best cross-LVLM average on both calibration and discrimination simultaneously, with statistically significant discrimination gains robust to cluster-aware analysis at 4-18x fewer parameters than the strongest probing baseline.
arXiv:2605.15232v1 Announce Type: new Abstract: Test Suite Minimization (TSM) reduces the size of test suites while preserving their fault detection capability. In black-box TSM, reduction is performed without analyzing production code. While several black-box TSM approaches have explored metrics like test logs or test similarity, those often suffer from scalability and efficiency issues. On the other hand, change-proneness (CP), recently emerged as an efficient and scalable alternative metric, has only been applied at class level. To accurately identify fault-revealing test cases, we propose CP at finer-grained method-level and implement Method-level Change-proneness based Test-suite Minimization (MCTM). MCTM first calculates CP for each method from version control metadata, then determines the dependency between test cases and methods by analyzing the test-code call-graph. Next, it scores the association between test cases and their invoked methods using statistical measures such as Average, Geometric Mean etc. Finally, test cases with the highest scores are selected to form the reduced suite. Evaluation on 15 open-source Java projects with 635 buggy versions shows MCTM achieves 0.93 accuracy and 0.94 fault detection rate on average, significantly outperforming class-level CP and similarity-based approaches while maintaining superior efficiency.
arXiv:2603.15273v2 Announce Type: replace Abstract: Noble liquids, notably argon and xenon, are utilised as both detector media and as the detector target for dark matter and neutrino physics experiments. When the noble liquid is excited by particles, it scintillates vacuum ultraviolet light, which sensors then detect. A major focus of the detector development community is on producing precision light sensors for noble liquid detectors. We introduce a flash lamp to test VUV-sensitive light sensors with light at wavelengths observed at noble liquid detectors. This paper discusses the design and presents results from a flash lamp prototype operated at room temperature.
arXiv:2605.15308v1 Announce Type: new Abstract: LLM-driven program evolution has emerged as a powerful tool for automated scientific discovery, yet existing frameworks offer no principled guide for designing their individual components and provide no guarantee that the search converges. We introduce SMCEvolve, which recasts program search as sampling from a reward-tilted target distribution and approximates it with a Sequential Monte Carlo (SMC) sampler. From this view, three core mechanisms emerge as principled components: adaptive parent resampling, mixture of mutation with acceptance, and automatic convergence control. We further provide a finite-sample complexity analysis that bounds the LLM-call budget required to reach a target approximation error. Across math, algorithm efficiency, symbolic regression, and end-to-end ML research benchmarks, SMCEvolve surpasses state-of-the-art evolving systems while using fewer LLM calls under self-determined termination. The code is available at https://github.com/kongwanbianjinyu/SMCEvolve.
arXiv:2605.15774v1 Announce Type: new Abstract: Traditional Fully Homomorphic Encryption (FHE) schemes often suffer from prohibitive computational overhead and complex noise management. In this paper, we propose a novel symmetric FHE through a mechanism of plaintext fragmentation and dynamic interposition. Our approach is built upon a modular encryption foundation, c = mk + rp, which is naturally additive but typically limited by exponential noise growth during multiplication. To resolve this, we introduce an interposition framework where the plaintext is partitioned into multiple fragments across distinct logical positions. We introduce a dual-regulator system to govern the multiplication process; exponent regulators (t_i) redirect the product of fragments to a new target position, preventing the accumulation of secret key exponents, while coefficient regulators (d_i) normalize the resulting scalars. Security is established through a binding mechanism where exponents and coefficients are mutually dependent, shielding the secret key k from algebraic manipulation and substitution attacks.
arXiv:2505.21698v3 Announce Type: replace Abstract: Vision-language foundation models achieve promising performance in natural image classification, yet their direct application to medical imaging is limited by severe domain shifts, resolution mismatches, and the multi-label nature of clinical diagnosis. Training dedicated medical foundation models from scratch, however, is costly and data-intensive. Here, we propose MedBridge, a lightweight adaptation framework that opens a new direction in domain-gap mitigation by jointly combining domain alignment, resolution preservation, and multi-label reasoning via complementary VLM experts for medical image diagnosis. Specifically, MedBridge transforms pretrained VLMs into multi-view query encoders that inject a compact set of learnable query tokens into intermediate layers, enabling non-destructive domain alignment while preserving fine-grained pathological cues via multi-view high-resolution sampling. These query tokens further act as routing signals for a mixture-of-experts, dynamically integrating heterogeneous foundation models for multi-label reasoning without requiring a shared representation space. We evaluated MedBridge on five chest radiograph benchmarks in three key adaptation tasks. MedBridge demonstrates superior performance in both cross-domain generalization (out-of-distribution transfer) and in-domain specialization (same-distribution tuning) settings, yielding a significant 6-15% AUC improvement over state-of-the-art adaptation methods for multi-label thoracic disease diagnosis. Furthermore, MedBridge is model-agnostic and demonstrates broad extensibility across eight diverse VLMs (e.g., CLIP, LLaVA, Qwen-VL, MedGemma), highlighting its ability to flexibly adapt arbitrary foundation models into a powerful medical diagnostic tool. Our code will be released upon acceptance.
arXiv:2605.15309v1 Announce Type: new Abstract: Despite remarkable progress, image generation is far from solved. The dominant metric, FID, conflates sample fidelity with mode coverage and is close to being saturated. Yet a model can still exhibit mode collapse while achieving a low FID, since a handful of sharp, near-duplicate images can outscore a model that faithfully covers the full data distribution. We argue that precision and recall are essential complements to FID, and that because FID is already saturated, the more meaningful goal is to improve diversity and coverage. Achieving high recall requires a model that explicitly prioritizes mode coverage, unlike most generative models, which optimize sample fidelity. We introduce RTM, which replaces the single-pass latent mapping in style-based generators with an iterative refinement process, and show that this consistently improves both quality and diversity. Integrated with Implicit Maximum Likelihood Estimation (IMLE), which optimizes mode coverage by design, RTM achieves the highest precision and recall among current state-of-the-art approaches while maintaining competitive FID, with improvements across CIFAR-10, CelebA-HQ at 256x256, and nine few-shot benchmarks. RTM also improves StyleGAN2 and StyleGAN2-ADA on CIFAR-10 and AFHQ-v1 at 512x512, demonstrating that the benefit is not specific to IMLE. Unlike flow-matching baselines that achieve competitive FID at the expense of coverage, recursive refinement improves both quality and diversity simultaneously.
arXiv:2605.15234v1 Announce Type: new Abstract: Many complex systems can be reduced to their key components through spectrally decomposing matrices that capture their dynamics. These matrices can in turn be constructed from data, often by least-squares fitting: examples of algorithms to do this include Dynamical Mode Decomposition and variants, subspace identification and eigenvalue realisation algorithms. Typical outputs of these algorithms include a range of isolated, peripheral eigenvalues capturing persistent emergent patterns in the system. However, there is no objective way to assess which of these discrete eigenvalues are artefacts of finite data error, and which are reflections of a fully sampled operator. n this paper, we present a sampling pseudospectrum $P(\lambda)$, that provides probabilistic information on the behaviour of finite-data eigenvalues in the complex plane, and an estimator $\hat P(\lambda)$, which can be obtained by reprocessing our finite data sample. The estimator, which is computationally efficient to implement, allows us to test statistically for the location of the true eigenvalues. This gives us a rigorous and very general way to assess whether the patterns we extract from finite data are likely to be signal or noise.
arXiv:2605.15357v1 Announce Type: new Abstract: A complete model of the motion of a quadcopter along a smooth spatial trajectory is presented. Based on the model, a robust algorithm is proposed for controlling a quadcopter using measurements of linear coordinates and yaw angle. By introducing additional integrators, a dynamic control algorithm with a simplified controller tuning methodology is obtained. The control law is synthesized within the geometric approach, and its stability is proven. A realizable output-feedback version using an extended observer is also given. The results enable coordinated trajectory following in three-dimensional space despite unmeasured disturbances and incomplete state information.
arXiv:2605.15206v1 Announce Type: new Abstract: Autonomous agents powered by large language models (LLMs) are increasingly used to automate complex, multi-step tasks such as coding or web-based question answering. While remote, cloud-based agents offer scalability and ease of deployment, they raise privacy concerns, depend on network connectivity, and incur recurring API costs. Deploying agents locally on user devices mitigates these issues by preserving data privacy and eliminating usage-based fees. However, agentic workflows are far more resource-intensive than typical LLM interactions. Iterative reasoning, tool use, and failure retries substantially increase token consumption, often expending significant compute without successfully completing tasks. In this work, we investigate the time, token, and energy overhead of locally deployed LLM-based agents on consumer hardware. Our measurements show that agentic execution increases GPU power draw, temperature, and battery drain compared to single-inference workloads. To address this inefficiency, we introduce AgentStop, a lightweight efficiency supervisor that predicts and preemptively terminates trajectories unlikely to succeed. Leveraging low-cost execution signals, such as token-level log probabilities, AgentStop can reduce wasted energy by 15-20% with minimal impact on task performance (<5% utility drop) for challenging web-based question answering and coding benchmarks. These findings position predictive early termination as a practical mechanism for enabling sustainable, privacy-preserving LLM agents on user devices. Our project code and data are available at https://github.com/brave-experiments/AgentStop.