Papers

We propose a data-driven framework for estimating the Bayesian Cramér-Rao bound (CRB) in high-dimensional imaging systems with complex, analytically intractable priors. Direct CRB computation is challenging in this setting due to the need to model the prior score and to form and invert the Bayesian Fisher information matrix in very high dimensions. To address these issues, we first reformulate the inverse problem in the latent space of a pre-trained variational autoencoder, thereby dramatically reducing the dimensionality of the bound estimation problem while preserving the spatial structure of the images. The Bayesian CRB is formed in this latent space and mapped back to the native parameter space using a change-of-variables formula. Second, to learn the latent prior score, we introduce a new sliced score matching objective defined in a Bochner space endowed with an $H^1(\Omega)$ Sobolev norm in space. This "Bochner-space sliced score matching" objective is consistent with standard sliced score matching, but penalizes errors in the spatial gradients of the score, suppressing high-frequency artifacts that otherwise contaminate the resulting CRB estimates. We validate the approach on a stylized quantitative photoacoustic computed tomography (qPACT) breast imaging problem with over one million unknown parameters, using a foundation-model autoencoder derived from Stable Diffusion. The proposed method yields stable, artifact-free Bayesian CRB estimates that reflect the highly non-Gaussian structure of the learned prior and reveal the substantial impact of the prior on the relative performance of competing qPACT design schemes.

Large language model (LLM) watermarking is important for ownership verification and intellectual property protection. However, existing approaches focus on algorithmic design while treating LLM inference engines as separate components. This separation often introduces auxiliary models or external tools, increasing latency and memory overhead while limiting the use of modern inference optimizations. As a result, a deployment gap remains: practical watermarking must preserve utility, detectability, and robustness while minimizing serving overhead. To bridge this gap, we propose SWIFT, a framework that co-designs LLM watermarking with modern inference infrastructure. SWIFT (i) integrates text generation and watermark construction within a shared LLM backend to reduce re-computation, improve cache reuse, and simplify system complexity; (ii) uses instruction-guided candidate generation with semantic protection to preserve factual consistency and meaning; (iii) executes watermarking concurrently with generation through asynchronous co-serving and adaptive scheduling to reduce latency and improve resource utilization; and (iv) leverages vLLM optimizations to enhance serving efficiency. Extensive experiments across domains and tasks show that SWIFT achieves strong utility, detectability, robustness, and downstream performance while substantially improving serving efficiency. Specifically, it achieves the highest utility score (4.87), 99.65% detection accuracy compared with 76.65% for the low-latency baseline SynthID, and up to 97.7% detection under watermark removal attacks, with 5.9x lower latency than the highest-utility baseline, SafeSeal. These results demonstrate the benefits of co-designing watermarking with inference infrastructure for practical LLM serving.

Video games hold a prominent role in the entertainment industry, which constitutes a multibillion-dollar global market. Within this ecosystem, independent studios, small teams, and solo developers operate with constrained budgets while making their games available through platforms such as itch.io, where many projects are also available as open-source. Where even if the project is fully released, other developers can still have access to the development process through the source code, helping other developers to develop their own games. These open-source games also constitute a valuable corpus for empirical software engineering research, offering transparent access to development artifacts that are typically inaccessible in commercial projects. This study characterizes open-source video games from a Software Engineering standpoint, focusing on their genres, popularity, repository activity, implementation characteristics, and testing practices. To this end, we extracted metadata from 491 video games from itch.io, published as open-source projects in GitHub. The results indicate that open-source games predominantly fall within limited genres (Puzzle, Action, and Platformer), that player ratings, when available, tend to be high, bringing a perception that the community receives well those kind of games, that repositories are largely dormant or minimally maintained, that the majority of these projects adopt either Godot or Unity as their game engine, and that documentation of testing-related practices is seldom included. The results offer insights into the development of open-source video games and may contribute to advancing software engineering practices and tools designed for small teams and independent game developers.

A one-step curvature probe evaluates an initial direction, whereas continual learners are judged after reaching comparable new-task fit. In an overparameterized linearization, projected gradient descent converges to $\Delta_P=PJ^\top(JPJ^\top)^{-1}r$, and its squared-displacement inflation is exactly the reciprocal of the retained fitting capacity $c_P(r)$. More generally, the terminal old-task quadratic ratio factorizes as $1/[c_P(r)G_{\rm end}(P,r)]$, where $G_{\rm end}$ compares curvature along endpoint directions. In the rank-one case, $G_{\rm end}$ equals the one-step probe gain; for multiple outputs, the two gains can differ. The terminal quadratic also separates into a curvature-optimal fitting floor and an algorithm-dependent null-space excess, motivating terminal null-space correction, which preserves linearized new-task outputs on its Jacobian batch. Controlled checks validate the local quadratic and show that projection can greatly reduce matched-norm curvature while barely changing terminal forgetting. Across common-threshold configurations, projection yields the larger signed old-task loss change in 73/99 matched pairs, retained fitting capacity falls with rank, and fixed-rank comparisons separate terminal forgetting even when probe gain is approximately matched. On Permuted MNIST and Split CIFAR-100, terminal correction decreases signed old-task loss in 166/180 method–dataset–seed pairs under the all-seed intention-to-correct analysis; because acceptance and outcome reporting use the same held-out split, this result is descriptive and test-conditioned. Overall, terminal cost depends jointly on retained fitting capacity, endpoint-direction curvature, and the null-space component selected by optimization.

Large language models (LLMs) are increasingly used as classifiers, yet they operate as opaque systems whose decisions are difficult to interpret, which complicates their use in regulated domains such as credit scoring or medical diagnosis. We propose an evolutionary framework that iteratively discovers natural language feature definitions (rubrics) for interpretable classification. An LLM generates candidate binary features, evaluates each sample against them, and the resulting vectors can be used to train a transparent classifier such as logistic regression. The feature set evolves over multiple iterations guided by classification errors, per-class activation rates, and feature ablation scores. We evaluate across three benchmarks, including a credit risk dataset representative of regulated domains, comparing single-shot LLM rubrics, evolved rubrics, and direct zero-shot LLM classification. Evolved features improve over single-shot rubrics by +2.9 pp on average and outperform zero-shot LLM classification on two of three tasks, while providing fully auditable decision logic. On the credit risk task, the zero-shot LLM performs at chance (50.7%) with a strong bias toward a single class, whereas evolved features achieve balanced, interpretable predictions. Crucially, this failure is invisible in aggregate accuracy and surfaces only under per-class auditing. Analysis reveals that evolution is most effective when label boundaries cannot be inferred from category names alone or when the LLM lacks reliable domain-specific reasoning.

Autonomous driving decision systems must balance safety, efficiency, and social norms in complex traffic interactions. Philosophical and ethical considerations have received limited attention in existing autonomous driving decision-making approaches based on numerical optimization, sequence prediction, and large language models (LLMs). We propose Chinese Philosophical Wisdom-Guided Driving (CPW-Drive), a closed-loop retrieval-augmented generation (RAG) framework that incorporates value guidance derived from Chinese philosophy into autonomous driving decision-making. Using Chinese Confucian thought as its knowledge source, CPW-Drive consolidates LLM-extracted keywords from relevant classical texts into driving-relevant value principles through manual screening and validation. It then contextualizes these principles through scenario-specific cases to form retrievable and reusable value guidance. We further propose Physics-aware Spatial Similarity Retrieval (PSSR), which compares vehicle layouts and velocity-extrapolated states to retrieve physically relevant historical cases. On Highway-env's multilane highway-driving task, CPW-Drive achieves success rates of 93.0%, 86.0%, and 72.0% across three traffic configurations. These results outperform the strongest baseline by 8.0, 22.5, and 25.0 percentage points, respectively. Across all configurations, CPW-Drive achieves the highest collision-free step count and maintains a low lane-change frequency. The results suggest that structured value guidance can improve simulated closed-loop safety and stability while introducing efficiency and latency trade-offs.

Finding a compact formula that fits a set of input-output pairs and predicts outputs on unseen inputs is a fundamental problem in science. Symbolic regression automates the search for such formulae: search-based methods explore the space of possible formulae directly, while transformers pre-trained on synthetic data produce formulae of comparable quality substantially faster. Existing transformers, however, are prone to overfitting — they find formulae that fit the training data well but do not extrapolate to input ranges unseen during training. We address this by shaping the set of formulae used to train a transformer, and show that the resulting formulae extrapolate substantially better. Fine-tuning the transformer on data with noise-corrupted target values further makes the synthesized formulae robust to noise in the observations. On SRBench and LLM-SRBench our transformer synthesizes a formula in about ten seconds and extrapolates better than all evaluated methods at a comparable budget. Search-based methods surpass our accuracy only when given one to three orders of magnitude more time.

A constrained optimization problem may involve a parameter in its objective and active constraints, yet the final decision may remain insensitive to small changes in that parameter. This raises a fundamental question: which inputs does a decision making system truly depend on? Building on this question, we introduce DePICT, a procedure for constructing decision preserving interfaces by ranking context directions according to the optimizer's solution sensitivity and aggregating them across an operating regime. We study this problem in a high dimensional setting where primitive context parameterizes a constrained task and the downstream agent observes only a selected subset of context directions. For locally regular constrained programs, we derive a Karush Kuhn Tucker (KKT) based characterization of when a context direction is optimizer relevant. Our analysis shows that appearing in the active optimization problem does not necessarily imply that a variable affects the final decision. Some context directions can alter the KKT conditions while leaving the optimal solution unchanged because their effect is absorbed by the dual variables. DePICT is designed to remove exactly these directions. In a controlled diagnosis, it recovers the decision relevant interface exactly and reduces linear predictor regret to 0.009, compared with 0.475 for the strongest competing baseline.

When using diffusion models to target image restoration problems, diffusion inversion is typically employed to retain relevant image information from the degraded images. Instead of inverting back to the initial time step (i.e., T), many methods invert to a pre-determined intermediate time step, in order to better preserve information from degraded source images. However, a pre-determined time step for inversion is not ideal for reconstruction, as a severely degraded image requires an earlier starting time step than a mildly degraded one. In addition, DDIM-based models corrupt the original signal by adding Gaussian noise, which can be mismatched to the nature of blur-like degradations, such as blur, haze, and low-light. To address these problems, we propose two solutions: (1) we adopt an alternative diffusion process, called the Inverse Heat Dissipation Model, that diffuses the input image by gradually blurring a data point (2) we propose to implement a time predictor to estimate the starting time step for the inversion, with the model learning to adapt to the degradation severity. Extensive experiments on standard benchmarks show that our method achieves state-of-the-art performance in both quantitative and qualitative evaluations, with excellent generalization to many restoration tasks.

During fine-tuning, a language model can assign less probability to previously learned answers even when the current gradient acts to preserve that probability. With momentum, each update also carries gradients computed at earlier model states, and these stored contributions can push the model in the opposite direction. We investigate how this optimiser memory contributes to forgetting by separating old-task loss into confusion among its answers and leakage of probability outside the answer set. Across three language-model families, answer mass consistently declines while discrimination among old answers usually improves: the model becomes less likely to produce answers that it can still distinguish correctly. Decomposing Adam updates reveals opposing contributions to this loss of answer mass. Over training, accumulated history favours leakage, while the current gradient opposes it. Resolving history by age shows that the harmful contributions come mainly from older gradients of the new task, whereas recent gradients tend to protect the old answers. Changes in history's effect are dominated by its orientation relative to the old-task gradient. Interventions that reset momentum while matching the initial update norm establish that stored history affects retention, with state-dependent immediate effects and lower final old-task loss over longer Adam continuations, mainly through recovered answer mass. Finally, integration along finite updates shows that most sampled large loss increases are captured by local projections, while curvature along history amplifies some events. Together, these findings reveal how an optimiser's memory can erode learned behaviour even as its current gradient acts to preserve it.

Many prediction tasks evaluate a trained tree ensemble on groups of rows that share feature values: discrete-time survival models expand each patient into $G$ time steps, click-through-rate models score every item in a search session, and scenario analyses vary a few inputs while holding the rest fixed. Standard inference treats each row independently and ignores this input-side correlation. We present TreeWalker, which instantiates partial evaluation for grouped tree-ensemble inference: constant features are static, varying features dynamic. TreeWalker walks each tree once per group, partitions a row bitmask at varying splits, and skips empty subtrees via an unsplit shortcut. It reads standard LightGBM and XGBoost models; training is unchanged. We prove a structural work decomposition: per-tree work splits into the constant-projected subtree size $|T_c|$, $G$ leaf writes, and a predicate-mask provisioning cost $Q$. For the trace evaluator, per-row work approaches $(d_v+1)/(d+1)$ as $G \to \infty$, giving asymptotic speedup $1/(1-\bar\rho_G)$. On Intel, TreeWalker is ${\sim}3\times$ faster than a row-independent traversal at the reference model size ($T{=}500$, $L{=}8$; $G{=}16$ for SUPPORT/FLCHAIN and empirical groups for Expedia) and $7.7$–$9.1\times$ faster at $G{=}128$ on the survival datasets; on Arm, $2.5$–$2.9\times$ and $5.5$–$6.0\times$. On a scenario-analysis benchmark it is faster in all $16$ configurations on both machines. For f64 models, outputs match treelite GTIL up to tree-ordering roundoff; for f32 models, TreeWalker's f64 leaf accumulation is closer than native f32 to an exactly rounded reference on every audited row.

Agentic multimodal large language models (MLLMs) have recently pushed the frontier of visual reasoning by calling tools such as zooming and tagging. Despite the recent strong success of agentic MLLMs, this work uncovers a critical safety failure in the tool-use paradigm: agentic tool-using MLLMs become less capable of refusing harmful requests. Our experiments confirm that, across three popular safety benchmarks, all the top open- and closed-weight MLLMs we test exhibit significantly lower safety in tool-using settings than in non-tool settings, with a relative refusal failure rate increase of up to 68.7%. Based on analysis of 100,000+ responses, including extended experiments, we also propose two possible reasons for this safety degradation.

General decision models, such as Jev, have recently emerged as efficient alternatives to LLMs for structured judgment and selection. But what kinds of decisions can these models reliably make, and how does their behavior change when individual decisions are composed into larger systems? To study this, we introduce JEVal, a bilingual benchmark comprising 11,257 instances from 36 datasets across 10 application domains, and evaluate 25 model configurations spanning general decision models and generative LLMs. Our results show that (1) general decision models are most competitive when decisions can be resolved from available evidence, but weaken when they require specialist knowledge or faithful uncertainty estimation: they can often identify the most likely outcome while substantially overstating its probability. (2) In more dynamic and realistic systems involving long-horizon, multi-step interactions, the advantages of fast local decision making are offset by reliability failures at the system level. on $\tau$-bench, faster local decisions reduce median episode time but lower task success as decision errors accumulate over long trajectories. (3) In large-scale social simulation, decision models approach strong generative LLMs on individual response prediction at substantially lower inference cost, yet remain weaker in user profiling and exhibit larger aggregate estimation errors and systematic bias. Finally, we propose SII-Decision-4B and SII-Decision-27B, which internalize an open-weight LLM's own reasoning into a single-pass first-token decision through Reasoning-to-Readout Self-Distillation, with SII-Decision-27B performing on par with Jev on JEVal while answering a typical query in about 0.1 s.

Existing analog integrated circuit design benchmarks make two questions hard to answer: whether a model has learned transferable circuit skills rather than recalled familiar examples, and whether its output works under defined process and test conditions. We introduce a benchmark built from human-designed, open-source circuits associated with Tiny Tapeout manufacturing shuttles. The collection contains 273 topologically distinct top-level designs. Every source is retrieved at the revision recorded for its shuttle submission and processed in a fixed containerized environment. The pipeline exports each eligible schematic image and its SPICE netlist from the same source file, giving transcription an exact structural reference. Commit dates support model-specific training-cutoff analysis, while author testbenches provide the simulation context for sizing. The benchmark evaluates schematic-to-netlist transcription and device sizing. Across seven models on a fixed set of 66 transcription tasks, the strongest model reaches 56.1% exact graph isomorphism, and six of seven models drop sharply from the small to the medium tier. For one frontier model, removing author-chosen labels reduces exact matches while preserving aggregate structural F1, suggesting that labels can aid connectivity tracing. On the 17 sizing tasks, the leading model converges on all 17 proposals and reaches 91.2 out of 100 against the human reference, while the two newest Claude models refuse 4 and 11 of the same prompts they transcribe without objection; a proposal without sizes scores zero. The two tasks produce different model rankings, exposing distinct visual and design capabilities and, in one family, a policy rather than capability limit.

Public sector agencies have adopted participatory design as a prominent procedural safeguard proposed against algorithmic harm. Across methods, institutional contexts, and well-intentioned implementations, the participation gap persists. Communities engage, input is documented, and systems are deployed largely unchanged. This paper locates the persistence of that gap in the mechanism institutions use to document participation itself. We develop the concept of the legibility trap, in which institutional mechanisms for documenting participation systematically select for forms of community input that fit institutional categories and filter out the forms that would most constrain state power, including foundational dissent, demands for non-deployment, and what Kelly Oliver calls witnessing to structural conditions of harm. Drawing on James Scott's theory of state simplification, we demonstrate the legibility trap through critical discourse analysis of twenty-two official, community-produced, and public-record documents across three public sector AI deployments: the Allegheny Family Screening Tool, Detroit Project Green Light, and the Community Control Over Police Surveillance (CCOPS) ordinances in Oakland. We identify the structural conditions under which participation partially escapes the trap, and we theorize that repeated non-response constitutes a harm to community subjectivity that AI ethics frameworks have not yet named.

This paper presents sparse calibration-based personalization for kernel-based gait phase and walking speed co-estimation from wearable inertial measurement units (IMUs). A 28-speed reference library was constructed from bilateral thigh and shank trajectories in a 20-subject locomotion dataset. Rather than directly applying population-average kernels, the method combines a user-specific baseline estimated from three calibration speeds with a principal-component (PC) model of baseline-centered kinematic deviations. Offline validation showed that personalized kernels reduced phase error relative to the population kernel. In a single-participant online pilot evaluation, the personalized kernels produced numerically lower mean phase and speed errors and ran in real time on embedded hardware. However, the speed effect was not significant, and pairwise phase differences did not remain significant after multiple-comparison correction. The pilot evaluation establishes wearable implementation feasibility while indicating reference-to-online dataset mismatch and the need for larger-cohort validation.

Cooperative multi-agent networks deployed for target localization and tracking remain critically vulnerable to malicious cyberattacks, since a single compromised agent can corrupt the centralized target belief and mislead the estimation process across the entire network. This paper presents a resilient multi-agent target localization and tracking framework under a limited agent sensing range and false-data injection (FDI) attacks. The proposed method combines information-based navigation for target search and posterior target state uncertainty reduction, Bayesian attack graph (BAG) based attack detection for probabilistic identification of the compromised agents, and reachable set-guided recovery that guides the uncompromised agents to relocalize the target when the sole localizing agent is attacked. The framework preserves the network localization capability after an attack by trust-gating compromised sensing information and guiding attack-immune agents toward a dynamically constructed recovery set representing the new target belief using a Linear Quadratic Regulator (LQR). Simulation validates that the proposed framework maintains a robust tracking performance under FDI attacks, demonstrating the potential for resilient autonomy in safety-critical applications.

Screen localization and pointing estimation are key to low-cost interactive devices. Yet developing and evaluating these algorithms requires realistic data: synthetic captures cannot fully reproduce the optical distortion, rolling shutter, motion blur, and display processing of a physical acquisition, and most existing datasets provide static images rather than the video needed to assess continuous pointing. In this paper, we introduce the DABACO Dataset. Developed within the DABACO (Dispositivo Apuntador de BAjo COste, or Low-Cost Pointing Device) project, this dataset supports the development and evaluation of screen detection and camera-based pointing algorithms for embedded systems. It comprises video sequences captured with multiple low-cost camera sensors, including monochrome global-shutter and color rolling-shutter modules, on embedded platforms such as the Raspberry Pi 4B and ESP32-S3. We present a systematic annotation pipeline combining temporary visual watermarking, optical-flow tracking, manual verification, and marker removal. Both the original marked captures and the marker-free images, together with explicit corner annotations and modification masks, are released to support auditing and the study of potential reconstruction bias. In addition, we release an open-source evaluation toolkit with two reference baselines: a classical screen-detection pipeline based on edge and contour geometry, and a closed-vocabulary, off-the-shelf YOLOv8 segmentation model evaluated without dataset-specific training. Both are evaluated on a general-purpose computer using Intersection over Union, corner error, and pointing error, establishing initial reference results for future embedded implementations. Overall, DABACO addresses a gap in existing resources and is intended to support the development and evaluation of new low-cost screen-localization and pointing systems.

This paper addresses autonomous landing of unmanned aerial vehicle (UAV) rotorcrafts on a heaving ship deck using a recurrent policy trained via asymmetric actor-critic reinforcement learning: the critic sees 6 s of future deck height during training, while the actor sees only what the onboard sensors provide in flight, a 17-dimensional state made of the vehicle's position, velocity and attitude relative to the deck and outputs world-frame velocity and yaw-rate commands. Training is performed across 4096 parallel simulated environments, followed by fine-tuning in 16 environments with an onboard vision pipeline in the loop. A control-barrier-function (CBF) stopping margin on the deck-relative vertical state is incorporated at two stages: as a reward term during training, where it halves the median simulated contact speed relative to a policy trained without it, and as a runtime safety filter at deployment, evaluated at every control step to abort and retry the descent when the margin is violated. Because the autopilot's disarm logic cannot detect the vehicle resting on a moving deck, proximity-based thrust cutoff at touchdown is commanded directly in the landing pipeline. The proposed approach is validated using a parallel-manipulator-platform-based deck emulator that reproduces the heaving motion of the ship deck (scaled to 0.70 m peak-to-peak, 7.5 s mean period) and a quadcopter UAV with an onboard camera. Across 31 motion-capture and 20 vision-based trials, the UAV landed every time, with median times to contact of 6.5 and 7.3 s; 77% and 50% landed on the first attempt, with mean deck-relative speeds of 0.29 and 0.27 m/s, respectively, at the instant of thrust cutoff, which is the last speed under the policy's control. Supplementary video: https://youtu.be/S5hDkrSZJt4

LLM-driven sequential recommendation formulates next-item prediction as autoregressive generation conditioned on natural-language prompts. However, minor wording changes in semantically equivalent prompts can cause substantial performance fluctuations, undermining robustness and requiring costly manual prompt engineering. Continuous prompt learning reduces template dependence but faces two interacting challenges: shared task-level instructions lack user-specific reasoning guidance, while gradient updates can push continuous prompts outside the LLM's effective semantic space. Injecting personalized signals can further amplify this semantic drift. To address these challenges, we propose LRPRec, a learnable prompting framework that initializes continuous instruction prompts from discrete templates and introduces two complementary mechanisms. Personalized prompt injection encodes user behavior into a preference embedding and additively injects it into shared prompts, enabling parameter-efficient user-level adaptation. A semantic drift constraint regularizes the shared prompts within a trust region around their initialization anchors to preserve semantic validity during optimization. By constraining the shared component while allowing additive personalization, LRPRec decouples stability from expressiveness. Extensive experiments on three benchmark datasets demonstrate consistent improvements over strong baselines while eliminating the need for manual tuning of background and task inference templates.

Flashblocks divide a block's priority gas auction into shorter sequential auctions that commit transaction order before the block is complete. We ask how this auction cadence affects bidding and competition among automated arbitrageurs, or searchers. In July 2025 Base replaced a single 2 s auction with ten 200 ms auctions. We use this change to estimate the searcher response from on-chain data alone. On an address-day panel of 3,032 searchers and 8,053 activity-matched controls, a difference-in-differences design estimates a 0.187 gwei fall in the effective priority fee (59% of the searcher pre-period mean). The share of searcher priority-fee value paid in the first tenth of block gas falls from 0.98 to 0.28. This fee compression is the identified effect. The revert-rate response is not (causally) identified and is reported descriptively. At the level of a single opportunity, the winning fee splits into an approximately invariant floor and a competitive premium. The premium scales about linearly with auction duration at high contention and less steeply at low contention. A first-price auction model with uncertain arrival and partial payments by losing searchers accounts for this pattern. We match attempts to the pool they contest. A higher-fee transaction reverts behind a lower-fee winner that landed earlier in 1.9 to 2.7% of contested opportunities by value, an upper bound. Latency therefore decides part of the ordering. A micro-auction model projects diminishing returns from shortening the window below 200 ms, as bids approach a nonzero floor. Rollups such as Arbitrum One now make auction cadence a governance tunable parameter, so these estimates inform how it is set.

Motility-induced phase separation enables swarm aggregation to be regulated without prescribing a target spatial distribution. This paper develops a partial differential equation-constrained model predictive control framework for a vibration-driven robotic swarm. Three continuum models retaining progressively less orientation information are considered to study the trade-offs among model fidelity, information requirements, and computational cost. The framework is evaluated under centralized and decentralized settings, with focus on how limited sensing and local-to-global phase-separation estimation affect the performance of predictive control.

Traditional sleep-staging methods apply the same model to every EEG epoch. Such uniform deployment expends computation on epochs that a smaller model could handle reliably, motivating cascades in which a primary classifier accepts its reliable predictions and defers the remainder to a more capable model. In this paper, we study the first stage of such a cascade: maximizing the coverage of fixed primary predictions subject to a prescribed accepted-risk target. We propose COVER (COVerage-oriented Error Ranking), which integrates two key innovations: (i) auxiliary-informed primary-error learning, which replaces maximum softmax probability (MSP) with a learned error score while preserving the primary labels, and (ii) fixed-scale scorer refinement, which builds on this score to directly maximize coverage under an empirical accepted-risk constraint rather than error-prediction accuracy over all epochs. We evaluate COVER on Sleep-EDF-20 at a 5% accepted-risk target, with subjects held out from all fitting and selection. Auxiliary-informed error learning raises mean subject coverage from 31.5% for MSP to 48.8% at similar subject-equal risk. At equal acceptance volume, with MSP accepting the same number of epochs as the learned scorer in each subject (20,639 in total), errors fall from 1,467 to 867. Fixed-scale refinement then adds 1.7 percentage points of coverage over its initialization in nested development, and COVER attains the highest mean coverage among eight evaluated scorers, 50.4% at 4.5% subject-equal risk, above the selective-ranking baseline SELE (49.3%) and the probability-fusion comparator DuoF (43.6%). To the best of our knowledge, this is the first work to combine auxiliary-informed primary-error learning with fixed-scale coverage refinement for selective sleep staging, offering a basis for reliability-aware allocation of computation in cascaded sleep staging.

We study the minimum Vertex Cover problem under two non-adaptive offline advice models, where predictions about a fixed optimum solution are provided once as part of the input. These forms of prediction were introduced independently by Cohen-Addad, d'Orsi, Gupta, Lee, and Panigrahi (2024) and by Ghoshal, Makarychev, and Makarychev (2025). In Partial Predictions, independently revealed vertices come with correct membership labels. For every sufficiently small fixed $\varepsilon>0$, we give a randomized polynomial-time algorithm that has expected approximation ratio at most $2-(2-o(1))\frac{\log\log(1/\varepsilon)}{\log(1/\varepsilon)}$. In Noisy Predictions, every vertex instead receives a mutually independent noisy membership label of bias $\varepsilon$. For every sufficiently small fixed $\varepsilon>0$, we give a randomized polynomial-time algorithm that has expected approximation ratio at most $2-(1-o(1))\frac{\log\log(1/\varepsilon)}{\log(1/\varepsilon)}$. Our Noisy Predictions bound matches the leading asymptotic improvement below $2$ obtained by Aamand, Chen, Gollapudi, Silwal, and Wu (2025), despite using only one noisy label per vertex rather than a separate independent label for each endpoint of each incident edge. Both our bounds hold as $\varepsilon$ goes to $0$ and are strictly below $2$, the optimal approximation threshold for minimum Vertex Cover without advice under the Unique Games Conjecture, as shown by Khot and Regev (2008).

When documents supporting an agent's derived facts are revoked or replaced, should it repair memory or re-read current evidence? We introduce an evidence-revision evaluation on medication- and problem-list tasks from public ICU records. Under revocation, replacement and control events, we compare full and source-filtered re-reading with caching, rebuilding and graph-local repair across two 7B models. Memory is supplied in full without retrieval, and costs include ingest, revision and every use. On short records, local repair uses 5-10$\times$ fewer revision tokens than rebuilding, yet every memory pipeline costs at least twice full re-reading in held-out conditions. In a small pre-specified development sweep, adding task-ineligible documents extended records to about 10,000 tokens; at that length, memory's mean cumulative cost fell below full re-reading's after 2-14 uses, partly through truncated extraction, while source-filtered re-reading remained cheapest. In the replacement study, none of the four primary confirmatory tests reached statistical significance. These results show why the cost of agent memory after evidence revision must be assessed against source-filtered re-reading over the full pipeline.

Model-based control of an aquatic robotic platform depends on a hydrodynamic model that is costly to identify and specific to the hull, payload, and conditions it was measured in. Here, we present MOSAIC-SV, a deployable real-time adaptive dynamics identification and control system that identifies a control-sufficient dynamics model from a spec-sheet engineering prior, without dedicated identification trials, and re-estimates it at every control step of a closed-loop mission while the controller plans on it. A physically admissible unscented Kalman filter re-estimates hydrodynamic, disturbance, and actuator parameters at every control step of the closed-loop mission; while a command-dependent consider projection withholds corrections the current command cannot attribute between actuator effectiveness and external force; and a model predictive path integral controller plans on the current estimate. In simulation on a CyberShip II plant, MOSAIC-SV recovers the transit performance of the calibrated model under static mismatch and transient changes, and stays within 20% of its own transit time at the unscaled prior when its inertia or damping prior is wrong by an order of magnitude. In on-water field trials on the Blue Robotics BlueBoat, a twin-thruster catamaran, MOSAIC-SV transits at least 25% faster and predicts its own motion with at least 56% less error than its frozen engineering prior, including under an unmodeled payload. The same MOSAIC-SV system concept was also feasibly deployed on a 6.3-tonne dual outboard monohull.

Graphoid automata evaluate graphs by composing relations between words of states. We examine the equations that make this evaluation independent of the chosen graph expression, and what changes when relations are replaced by matrices over a commutative semiring. We prove that the classical graphoid equations are equivalent to the special commutative Frobenius equations together with the required symmetry laws. The known cospan and hypergraph presentations then describe equality of structural and labelled expressions, with the bone law deleting precisely the isolated internal vertices. We use established partition representations to compare this equality with equality of evaluated matrices. For any nonempty finite state set $Q$ and commutative semiring $K$, we give a uniform criterion for two structural expressions to have the same matrix. The criterion determines which isolated-vertex counts are distinguished and when a vanishing scalar makes matrices from different boundary partitions equal. For the canonical structure, the bone law holds precisely when $|Q|\cdot1_K=1_K$. Preserving all relational compositions is a separate requirement, equivalent to additive idempotency for $|Q|\geq2$. We then examine the characteristic matrices of the known relational structures on $Q$ given by disjoint unions of Abelian groups. Their Frobenius equations hold over every commutative semiring, whereas specialness depends on the order of each group and the bone law depends on the number of groups. These conditions specify which equations remain valid when weights are assigned to a graphoid automaton.

Large Language Models (LLMs) are increasingly used for automated algorithm design. However the computational cost of evaluating the generated algorithms can be excessive. We consider the common LLM-driven automated algorithm design (LLM4AD) setting in which a candidate algorithm is evaluated by aggregating its performance over a shared set of training instances. This instance-wise structure raises a natural question: must every candidate be evaluated on the entire instance set before deciding whether it remains competitive? Taking inspiration from algorithm configuration, we introduce AdaEva, a drop-in adaptive partial-evaluation framework that progressively evaluates candidates on larger subsets of the same instance pool and eliminates unpromising candidates as evidence accumulates. Importantly, AdaEva leaves the underlying LLM4AD procedure and per-instance evaluator unchanged and requires no prior knowledge about instance difficulty. We instantiate this idea using successive halving (AdaEva-S) and statistical racing (AdaEva-R), and evaluate both mechanisms across three representative LLM4AD frameworks, multiple LLM backbones, and optimization domains spanning combinatorial and continuous black-box optimization. Under matched evaluation budgets, AdaEva more reliably balances evaluation effort across candidates than fixed partial-evaluation strategies, yielding strong search efficiency and anytime performance together with improved held-out generalization across the evaluated settings.

A deployed LLM agent emits tool calls, queries, and code that can be silently wrong – by the time the error surfaces, the action has run. Frontier chat APIs hide the model's token probabilities; the agent's stated confidence barely beats chance on the mistakes that matter; and resampling does not help, since frontier models are highly repetitive, reproducing the same call across samples. We recover the missing signal from a low-cost open-weight surrogate run in parallel. It reads the same context, schema, and proposed action as the agent, then scores the call from its own log-probabilities through a family of complementary readouts: teacher forcing and request-PMI weigh the likelihood of each argument value, a discriminative verdict judges the call as a whole, and tool-choice competition tests the function against its siblings. One principle says which to trust: a generative likelihood localizes wrong argument values, while the verdict catches holistically wrong calls. When the error type is unknown, an ensemble is the low-regret default. The readout is training-free, needs no access to the agent's internals, and costs one prefill pass alongside the tool call. On difficult coding tasks it reaches AUROC 0.825 where the actor's stated confidence is near chance (0.598), and the generative readouts beat it by +0.07 to +0.28 across three further actors. Against self-consistency it gains +0.14 to +0.19 on near-deterministic actors, at 1/K the cost. The signal drives two deployment modes: a real-time gate escalating the least-trustworthy calls for review (+0.05 to +0.30 accepted-action accuracy at 50% coverage), and confidence feedback, returning the tool result with the score so the agent adapts its next step – lifting task success on live-execution benchmarks (+0.119 and +0.137, p <= 1e-4) and beating a random-value control where step errors are silent (+0.078, p = 0.003).

Reconstructing global sea surface pH from sparse observations is critical for monitoring ocean acidification and understanding marine carbon cycling. Traditional assimilation and inverse models are physically grounded but costly for large-scale reconstruction. Recent black-box and physics-guided AI models improve efficiency, but are mainly designed for passive tracers, where the reconstructed variable is also the transported inventory. In contrast, pH is an active carbonate tracer: it is the prediction target, while dissolved inorganic carbon (DIC) is the conserved carbon inventory. This mismatch can produce low pH error while violating carbonate closure and source-free carbon conservation. To address this, we introduce REACT, a carbon-first reconstruction framework that decouples transport, active correction, and chemical decoding. REACT transports a latent carbonate state with a conservative advection–diffusion solver, captures non-conservative carbon-cycle variations with a source module, decodes the corrected state into pH, and constrains the output through carbonate equilibrium. This design keeps pH as the target while enforcing consistency on the underlying carbon state. On simulation data, REACT reduces pH NRMSE by (14.7%) and chemical consistency error by (24.0%) over the best baseline. Cross-temporal-scale evaluations show robustness against error accumulation from coarse to fine temporal scales, and ablation studies validate the effectiveness of each component.