Systems and Control
Papers
Fresh from arXiv, typeset properly. Highlighted words trade on the market.
All of arXiv
Wednesday 2 September
The Balearic Islands are striving to achieve 100% renewable energy, which poses new challenges in operating the power system securely and reliably. With a higher integration of inverter-based resources (IBRs) and a reduced presence of conventional synchronous generators, the system strength, particularly short circuit levels, becomes weaker and more susceptible to disturbances; and the total inertia of the system becomes lower. Technology enablers are planned to achieve energy transition in the Balearic power system: a new 2x200 MW VSC-HVDC link (bipole with metallic return), Synchronous Compensators (SC) and Battery Energy Storage Systems (BESS), as fully-integrated network components. Detailed ElectroMagnetic Transient (EMT) simulation studies may be needed to analyse stability of the system, due to the high amounts of power-electronics devices in the system. This paper describes the implementation of an EMT model of the Balearic power system.
Optimization and Control
Inverse Optimal Control with Convex Features and Box Constraints
We study an inverse optimal-control problem for identifying nonnegative weights in a convex-feature lower-level objective. The resulting model is an optimistic bilevel optimal-control problem whose lower level is a strongly convex, linearly constrained control problem with box constraints in \(L^2\). For the reduced response map \(x\mapsto (y_x,u_x)\), we prove global single-valuedness and an explicit Lipschitz estimate, without assuming differentiability of active sets. Exploiting the affine dependence of the lower-level functional on \(x\), we show that the value function is concave and Lipschitz and derive an asymptotically exact optimal-value relaxation. We further establish local identifiability and \(O(\delta)\) noise stability by a gradient-free argument based on second-order growth. Finally, a KKT reformulation yields a function-space MPCC; Scholtes relaxation gives C-stationarity under uniform multiplier boundedness. A decoupled scalar example provides an explicit strict-complementarity certificate via finiteness of threshold contacts.
Read the full abstract
Probability
Optimal Sub-Gamma Scales for Spectrally Positive Infinitely Divisible Laws
Let X be a centered spectrally positive infinitely divisible random variable with finite nonzero variance V. We determine the smallest scale c in a right-sided sub-gamma bound whose quadratic proxy is fixed at V. A probability transform of the normalized Kolmogorov canonical measure produces a nonnegative variable R_X for which log E e^{tX} = V t^2 M_{R_X}(t)/2. The scale problem therefore reduces to comparing the MGF of R_X with that of an exponential law, leading to an exact one-dimensional variational formula. If X is non-Gaussian and has a positive exponential moment, the optimal scale is at least kappa_3(X)/(3V), with equality precisely when R_X belongs to the M-class of Klar and Mueller. Equality in the standard sub-gamma envelope occurs only for the Gaussian law or, in the non-Gaussian case, for centered compound Poisson laws with exponential jumps. For a general compound Poisson law, R_X is the second-order equilibrium distribution of the jump size. Exact calculations for gamma and bounded two-point jumps show that the optimum may occur at the origin, at the positive MGF boundary, or at an interior point.
Read the full abstract
Computer Vision
Seeing the City or Recognizing the Place? What Street-View Imagery Adds Beyond Existing Urban Data in VLM-Based Urban Sensing
Street-view imagery is increasingly analysed with vision-language models (VLMs) to infer urban attributes, but predictive accuracy alone does not show how much a photograph contributes beyond data already available for the same place. Using three VLMs, we compare image-based predictions with existing urban data for seven attributes drawn from five public resources. Each urban unit is evaluated with imagery, with location or text context, and against non-image predictions from nearby observations or public records. We also replace images and add conflicting records to test which source the predictions follow. Nearby observations or public records matched or exceeded image-only predictions for road damage, curb ramps, population, and house price. Images were more informative for building type, building function, and the floor count of low-rise buildings. For floor count, the advantage of images over nearby OpenStreetMap labels increased by 5.7 percentage points per doubling of distance to the nearest labelled building and declined for tall buildings whose rooflines often fell outside the frame. When images and records disagreed, predictions usually moved towards the supplied record. Released OpenFACADES floor annotations, generated with OpenStreetMap floor values as input, showed the same dependence: their agreement with the reference increased with building height, whereas that of image-only reruns decreased. The value of street-view imagery therefore depends on whether an attribute is visible and how well the place is already covered by existing data. Because machine-derived labels are often reused as references, these comparisons also bear on how urban datasets are documented and evaluated.
Read the full abstract
Computer Vision
Domain generalization and synthetic data in object detection: the enabler, the probe, and the gap
Object detection models often experience performance degradation when deployed under distribution shifts, caused by for example changes in weather type, operational environment, or object appearance. Domain Generalization (DG) aims to develop models that remain robust under such shifts and generalize well to unseen domains. DG research specifically focused on object detection models is scarce, although these models face additional challenges around localization and multi-scale representations. Synthetic data is a promising tool to support in DG, by enabling large-scale generation of diverse new samples. In this paper, we present an object detection-centric review of DG and examine the role of synthetic data from three complementary perspectives. First, synthetic data acts as an enabler of DG through diversification and alignment strategies that aim to improve robustness to distribution shifts. Second, it serves as a probe that enables controlled experimentation to identify and understand failure modes. Third, we discuss the synthetic-to-real gap, a particularly challenging form of domain shift that arises when models trained on synthetic imagery are deployed on real-world data. Through reviewing these perspectives, we identify limitations of current DG approaches for object detection and argue that future research requires representation-aware methods that explicitly address both localization and classification under domain shift.
Read the full abstract
Algebraic Geometry
Direct Group Laws on Classical Genus-One Models
This exposition constructs group laws by residual intersection on short Weierstrass cubics, the curves $(u^2+u)(v^2+v)=d$, Edwards curves, Hessian curves, Jacobi quartics curves, Jacobi intersections curves, Huff curves, and a symmetric QRT biquadratic. Lines give the operation on plane cubics, while planes give the operation on smooth intersections of two quadrics in $\PP^3$. Associativity is proved by comparing two products of three line or plane equations. After the common zeros are cancelled, equality of the remaining zeros follows from the elementary fact that a smooth projective curve of genus one admits no nonconstant rational function with one simple pole. The formulas for $(u^2+u)(v^2+v)=d$ are valid in every characteristic and are written explicitly in characteristics $2$ and $3$; the other models are treated in characteristic different from $2$ and $3$.
Read the full abstract
Number Theory
Two-Line Elimination and the Weierstrass Group Law in Arbitrary Characteristic
Let $K$ be a field and let $E/K$ be a nonsingular projective curve in general Weierstrass form. We prove associativity of its chord–tangent operation by combining polynomial factorization, formal differentiation, and quadratic interpolation. The main identity eliminates the two nonvertical lines $Y=\ell(X)$ and $Y=m(X)$ occurring in the successive additions $S=P\oplus Q$ and $T=S\oplus R$. There are $c\in K$ and a monic quadratic polynomial $h\in K[X]$ such that \[ \ell+m+\Apol=c(X-x_S), \qquad \Gpol+\ell m=(X-x_S)h, \] and \[ (Y-\ell)(Y-m) =(X-x_S)(h-cY) +Y^2+\Apol(X)Y-\Gpol(X). \] When $c\ne0$, the graph of $q=h/c$ contains $P,Q,R$, and $-T$. The two bracketings determine the same quadratic by ordinary or Hermite interpolation and therefore yield the same final point. The proof applies in every characteristic, since tangency is expressed through the formal partial derivatives of the Weierstrass equation. For the short Weierstrass form, a separate concise proof identifies $q$ with the interpolating parabola in Zwegers's construction.
Read the full abstract
Software Engineering
Software Project Management with LLM-Based Automation: Coordination, Validation, and Governance in Practice
As software engineering has evolved, development environments have increasingly integrated automated support for tasks such as testing, analysis, and code generation, requiring project management to coordinate both human work and automated workflows. In this paper, we investigate how software project managers perceive changes in their work and learning demands associated with LLM-based automation. Motivated by a gap in software engineering research, which has largely focused on task-level and developer-centered uses of LLMs, we adopted an exploratory case study approach to capture managerial perspectives on automation in practice. Based on the experience of software project managers working in a large, multi-project software organization, our analysis indicates that LLM-based automation influences planning, estimation, coordination, monitoring, and governance activities rather than introducing new formal management practices. Participants described LLMs as becoming embedded in everyday project work, producing uneven effects on productivity, increasing the need for review and validation, and reducing visibility into task execution. These effects contribute to greater reliance on managerial judgment and coordination. Learning demands were perceived as experiential and incremental, centered on understanding LLM capabilities and limitations, critically assessing generated artifacts, and guiding responsible use within teams. The findings provide empirical evidence on how software project management work is adapted in contexts where LLM-based automation is integrated into ongoing software development practice.
Read the full abstract
Sound
High-Value Synthetic Supervision for Parameter-Efficient Adaptation of a Compact Japanese Speech Model
Private domain speech is difficult to collect and redistribute, while compact models need task-specific supervision. We study an auditable synthetic pipeline that maps Japanese care handoffs directly to six-field structured notes. Using 182 synthetic training and development clips, we adapt a 1.47B audio model by full fine-tuning and rank-16 LoRA. On a 39-clip scenario-seed-disjoint synthetic test, an unadapted model obtains a model-judged factuality-recall score of 0.0500, full tuning 0.8664, and LoRA 0.8461. LoRA reaches 97.7% of the full-tuning aggregate as a descriptive ratio while project telemetry reports 12.4M trainable parameters, about 0.85% of the backbone. Both adaptations show large paired gains over the same base; the full-versus-LoRA interval crosses zero, and differing optimization settings preclude an equivalence claim. This is a parameter-efficient capability-acquisition result, not a device-performance result: latency, memory, energy, and real-time factor were not measured. All evaluation speech and targets are synthetic, references are model-proposed, and the judge is uncalibrated. The evidence shows that a compact model can acquire a narrow audio-to-structure transformation from a few hundred provenance-linked synthetic examples; it does not establish clinical validity, real-speech transfer, or superiority to a clean cloud system.
Read the full abstract
Machine Learning
PALM: Point-in-Time Adaptation for Financial Language Models
Language models used in financial backtests suffer from look-ahead bias, as a model trained on text published after the study period has already observed the outcomes it is asked to predict. To handle this issue, point-in-time (PIT) language models are pretrained on chronologically filtered corpora and released as one checkpoint per calendar year, each with a documented cutoff. However, each additional year costs another pretraining run, and whether that run is necessary has never been tested. In this paper, we show that a newer checkpoint is not detectably better on the decision it informs, as its information coefficient on a fixed evaluation window shows no decay with age that we can detect. Motivated by this observation, we propose PALM (Point-in-time Adaptation for financial Language Models), a simple yet effective alternative to annual pretraining that fits a low-rank adapter on financial text the checkpoint is allowed to read, without modifying any pretrained weight. We further find that the gain comes from the financial domain rather than from recency, which leaves the point-in-time constraint without a detectable cost. We validate PALM on a decade of financial news and on three families of PIT models, whose cutoffs span two decades. The pooled gain keeps its sign across experimental settings and comes at a small fraction of the cost of pretraining. Code is available at: https://github.com/seunghan96/palm.
Read the full abstract
Number Theory
Arithmetic progressions in dense subsets of primitive elements of finite fields
Let $p$ be a prime and let $\mathcal{P}_p$ be the set of primitive elements of $\mathbb{F}_p$. Inspired by the work of Cohen, Oliveira e Silva and Trudgian on consecutive primitive elements, and by the work of Chang on arithmetic progressions in multiplicative subgroups of finite fields, in this paper, using estimates for multiplicative character sums over systems of linear forms together with the relative Szemerédi's theorem of Conlon, Fox and Zhao, we show that, for every fixed integer $k\ge3$ and any fixed number $0<\delta\le 1$, each subset $A\subseteq\mathcal{P}_p$ with $\#A\ge\delta\#\mathcal{P}_p$ contains a nontrivial $k$-term arithmetic progression, provided that $p$ is sufficiently large. Also, by applying Behrend's theorem, we construct a large subset $A'$ of $\mathcal{P}_p$ with $\#A'=\lfloor(\#\mathcal{P}_p)^{1-\varepsilon}\rfloor$ such that $A'$ contains no nontrivial $k$-term arithmetic progressions.
Read the full abstract
Artificial Intelligence
Cross-Modal Emotion Understanding: A Transformer-GAT Approach for Dialogue Emotion Recognition
Multimodal emotion recognition is a key research area in affective computing, with applications in sentiment analysis, intelligent customer service, and human-computer interaction. However, existing methods often rely on single-modal features or simple multimodal fusion, failing to capture the synergy between global and local contexts, which limits model performance and emotion understanding. To address this challenge, we propose Transformer-GAT, a hybrid framework that combines Transformer and the Graph Attention Network to enable cross-modal emotion understanding. The Transformer is used to capture global semantic information, while the Graph Attention Network is employed to model fine-grained relationships between modalities, thereby enhancing the representation of emotional features. Experiments on the IEMOCAP and MELD datasets show that our model achieves weighted F1 scores of 72.45% and 77.37%, outperforming state-of-the-art methods. These results demonstrate that Transformer-GAT effectively integrates multimodal features, balances global and local contexts, and provides deeper emotional insights, offering new directions for multimodal emotion computing.
Read the full abstract
Artificial Intelligence
ERRAND: Budgeted Maintenance of Agent Memory
Deployed agents run on handed-over knowledge: a frozen policy consults a briefing of consolidated items written before the stream begins. The world then moves while the store stands still: paths close, flags change, price bands move; every item was true at handover, and the failure is staleness, not ignorance. We introduce ERRAND, which treats revalidation as a priced errand: a recheck competes with the task it protects for the same scarce actions, funded only when the value per action of resolving a doubt clears a running wage. The errand index is single-peaked, vanishing at both ends of belief, so certainty in either direction costs nothing; free en-route receipts maintain on-path knowledge, and repair writes a version, never a deletion. Under equal action budgets in two drifting tool-use worlds, ERRAND clears every non-oracle policy on the preregistered calibers, primary in every setting and conditional at every binding budget, leading eager revalidation by 10.0pp at the base cap. Restraint wins: given no cap, ERRAND stops on its own, spending 11.0% of steps, while uncapped eager revalidation spends 70.7% and still finishes 4.5pp behind capped ERRAND. The margin sits where the briefing's coverage is thinnest, the shadow price of long-tail knowledge: a small budget, well priced, beats a bigger store that never rechecks.
Read the full abstract
Machine Learning
The Impossible Trinity of Time-Series Validation: A Conservation Law among Training Sufficiency, Test Coverage, and Temporal Causality
Validating a model on a time series asks for three things at once: each training run should use most of the sample (sufficiency), the test sets should together cover most of the sample (coverage), and training data should come before test data (causality). We prove that the three cannot be had together and price each one. Let $\alpha$ be the smallest training fraction over folds, $\beta$ the fraction of the sample covered by tests, $\Lambda$ the fraction of the sample used as training data from the future of a test point, and $\delta$ the distance from a test point to the nearest training point in its future. Every scheme on a sample of length $T$ satisfies $\alpha+\beta \le 1+\Lambda$ and $\alpha+\min\{\beta,\delta/T\} \le 1$, and under $\beta$-mixing the leakage bias at a test point is at most $2M\beta_{\mathrm{mix}}(\delta)$. In words: going beyond the causal frontier $\alpha+\beta=1$ requires training on the future; that future data must sit within $(1-\alpha)T$ of a test point; and its harm depends on its distance, not its amount. Hence expanding walk-forward is exactly the Pareto frontier of causal validation, $k$-fold cross-validation buys the most future data, and purged $k$-fold with an embargo pays in distance instead, which is cheap when the process forgets quickly but cannot repair the part of causality demanded by non-stationarity. On pure noise, shuffled 5-fold reports an information coefficient of $+0.32$, while contiguous 5-fold, using the same amount of future data, reports $+0.004$.
Read the full abstract
Classical Physics
Compensated Free and Bound Sources and Nonradiating Conditions in Classical Electrodynamics
Classical macroscopic electrodynamics allows free and bound sources to cancel as distributions while the auxiliary equations retain nonzero free-source terms. A neutralized uniformly polarized sphere has $\mathbf{E}=\mathbf{0}$ and $\mathbf{D}=\mathbf{P}$, and a compensated uniformly magnetized sphere has $\mathbf{B}=\mathbf{0}$ and $\mathbf{H}=-\mathbf{M}$. These auxiliary fields are fixed by $\mathbf{P}$ and $\mathbf{M}$ rather than forming independent degrees of freedom. Time dependence separates charge cancellation from complete four-current cancellation. Canceling the charge alone leaves a divergence-free total current whose on-shell transverse transform controls radiation. For a neutralized polarized sphere of radius $R$ maintained by a tangential free-current sheet, the exterior field equals that of a point electric dipole whose effective moment is proportional to the spherical Bessel function $j_2(kR)$, where $k=\omega/c$. Every charge multipole and the ordinary magnetic dipole vanish, yet the source radiates at generic frequencies, and the complete exterior field vanishes at nonzero roots of $j_2$. Complete four-current cancellation removes all retarded source-generated fields. Among the compactly supported separable currents that maintain the same charge, it is the only choice that never radiates. In vacuum the auxiliary fields reduce to $\mathbf{D}=\varepsilon_0\mathbf{E}$ and $\mathbf{H}=\mathbf{B}/\mu_0$, so no radiation can be carried by $\mathbf{D}$ and $\mathbf{H}$ alone.
Read the full abstract
Quantum Physics
Quantization and Mirror Reduction Do Not Commute in Hamiltonian Embeddings of Nonreciprocal Dynamics
Hamiltonian embeddings can represent dissipative nonreciprocal classical dynamics exactly on invariant manifolds of enlarged reciprocal systems. We show that canonical quantization of such an embedding need not commute with reduction to the target dynamics. For the mirror construction recently introduced for pairwise nonreciprocal interactions, the invariant manifold is Lagrangian. Exact quantum enforcement of the mirror condition therefore removes the dynamical sector rather than producing a quantum analogue of the reduced flow. If the constraint is imposed only semiclassically, contraction of the target dynamics generates inverse-transpose expansion in the conjugate mirror directions. Quantum uncertainty then gives $\ln D\ge\Xi(t)+n\ln[\pi/(\sigma\epsilon)]$, where $\Xi=-\ln|\det M|$ is the accumulated contraction and $D$ the torus Hilbert-space dimension. Exact finite-dimensional Weyl evolution of one- and two-degree-of-freedom nonreciprocal models confirms the resulting logarithmic correspondence time, including its predicted change when the initial localization scales with $\hbar$. Thus the classical embedding is exact, but its direct canonical quantization is a singular semiclassical construction.
Read the full abstract
Analysis of PDEs
Semi-differentiability of generalised $L^\infty$ envelopes with applications to supremal functionals
J. Danskin proved in 1966 that the envelope of a family of continuous functions $\mathrm F : \mathbb R^n \times \mathrm K \longrightarrow \mathbb R$, parameterised by the points of a compact $\mathrm K \subseteq \mathbb R^m$, given by \[ f(u) := \max_{k \in \mathrm K} \mathrm F(u,k), \ \ \ \ f \, : \, \mathbb R^n \longrightarrow \mathbb R, \] is semi-differentiable on $\mathbb R^n$, if $\mathrm F$ is sufficiently regular. This result is of utmost importance in numerous applications. Extensions have been proved towards almost every direction, but none permits to replace “max over $\mathrm K$" with “essential sup over $\mathrm K$". We establish the semi-differentiability of generalised $\mathrm L^\infty$ envelopes when $ \mathrm F$ is defined on the product of a Banach space with a measure space. As an application, we obtain a new foundational regularity result in the Calculus of Variations in $\mathrm L^\infty$, asserting that supremal functionals are semi-differentiable everywhere, with an explicit formula for their semi-derivative, despite generally being non-differentiable. Moreover, we establish a variational characterisation of (absolute) minimisers of general level-convex $\mathrm L^\infty$ functionals via their semi-differentials, revealing the genuine $\mathrm L^\infty$ counterpart of the Euler-Lagrange equations. Conventional Aronsson equations and related divergence PDE involving measures, deducible from $\mathrm L^p$ approximations as $p\to \infty$, in general are not sufficient for minimality in $\mathrm L^\infty$. Our results dislodge the trademark necessity for extrinsic $\mathrm L^p$-approximations as $p\to \infty$ to derive and study PDEs, reinterpreting the existing $\mathrm L^\infty$ theory. Most crucially, they provide a new powerful intrinsic $\mathrm L^\infty$ framework, evocative of the straightforward variational methods for integral functionals.
Read the full abstract
Robotics
Silent Sabotage: Internal State Triggered Backdoor Attacks on LLM-Powered Robotic Systems
The integration of Large Language Models (LLMs) into robotic control systems is enabling a new generation of autonomous agents capable of complex reasoning and planning. While this paradigm shift accelerates progress, it also introduces novel security risks that remain largely unexplored. Current research into LLM backdoors has focused on attacks triggered by external stimuli, such as specific words, visual objects, or environmental states. These attacks, while potent, overlook a more insidious class of vulnerability where the trigger is internal to the agent's own operational logic. This paper presents the first comprehensive study of history-based backdoor attacks on LLM-powered robotic systems. We demonstrate that an attacker can embed a stealthy backdoor into an LLM-based robot controller by manipulating its instructions. This backdoor is triggered not by an external cue, but by a specific, rare sequence of the robot's own past actions. It remains dormant during normal operation, preserving the robot's utility, but can be activated to induce a malicious behavior, such as a complete stop or a collision. Our experiments, conducted in a simulated environment with a variety of robots and LLMs, show that this history-based attack is highly effective, achieving a near-perfect attack success rate while remaining exceptionally difficult to detect. These findings reveal a critical and previously unaddressed vulnerability in autonomous systems and underscore the urgent need for security measures that account for an agent's internal state.
Read the full abstract
Computation and Language
Peerify: Benchmarking Peer-Review Claim Verification
Peer review plays a central role in scholarly publishing, yet verifying whether reviewer claims are supported by manuscript evidence remains a largely manual and time-consuming process. We present Peerify, a pipeline for manuscript-grounded verification of peer-review claims. Given a manuscript and a review comment, the Peerify pipeline decomposes reviews into atomic claims, retrieves relevant manuscript evidence, and determines whether each claim is supported by the paper. To support the development and evaluation of the pipeline, we construct a benchmark of 800 claims derived from authentic peer-review interactions collected from NeurIPS 2024 and ICLR 2024, including a 300-claim hand-labeled subset used to audit the automated supervision. We evaluate state-of-the-art language models and retrieval strategies within the Peerify pipeline, together with entailment baselines. Our results demonstrate the importance of retrieval-centered verification and claim decomposition, while highlighting the challenges posed by ambiguous and interpretive reviewer claims. Automated labels agree with human consensus on 90.3% of audited claims ($\kappa = 0.87$), while off-the-shelf entailment models stay below 0.24 macro-F1.
Read the full abstract
Computation and Language
Do LLMs Choose Like Humans? Using Cognitive Theory to Evaluate LLM Decision-Making
Large language models (LLMs) exhibit a range of human-like decision-making behaviors, but whether these reflect similar underlying mechanisms or surface-level mimicry remains unclear. We evaluate whether LLM context sensitivity aligns with a cognitive economic theory that explains human behavior through problem categorization and attention allocation. Across 12 open-source and commercial LLMs on a novel 140,000-trial product choice benchmark, context induces human-like shifts in choice and problem categorization, but does not reliably reweight attention between features like price and quality. Neither scale nor chain-of-thought reasoning reliably attenuates context sensitivity or generates human-like behavior. These results suggest that LLM decision mechanisms are distinct from human ones.
Read the full abstract
Computation and Language
From Trait Vectors to Circuits: Tracing Refusal and Sycophancy Through Language Models
A direction in activation space that changes safety-relevant behavior when steered is not necessarily one the model uses to produce that behavior on its own. We therefore ask whether steering acts through the computation of the unmodified model or through a different set of components, studying two traits whose directions have been extracted and validated in prior work: refusal and sycophancy in Qwen2.5-7B-Instruct. For each, we use the trait vector to split the computation into a reconstruction circuit before the vector and a transmission circuit after it. We then test whether restoring the coordinate returns behavior removed by ablation. For refusal, both circuits are compact and faithful, restoring the coordinate alone recovers almost all of the refusal signal lost to ablation, and a circuit built around the vector matches the faithfulness of a direct input-to-output circuit at roughly half the edges. For sycophancy, transmission is compact, but reconstruction is broader and only partially faithful. Circuits that transmit a steering intervention are therefore not automatically the circuits that produce the behavior, and we report the counterfactual, target, and behavioral test for every circuit accordingly.
Read the full abstract
Computation and Language
EAVer: Long-Form Factuality Verification as an End-to-End Agentic Policy
Long-form factuality verification is commonly implemented as a static decompose-search-verify pipeline, with separately prompted modules processing claims and invoking external search. Treating claims independently makes LLM and search calls scale with claim count and causes repeated searches for overlapping evidence about related claims. We introduce EAVer, an End-to-end Agentic Verifier that learns to control the complete response-level verification workflow as a unified policy. EAVer groups semantically related claims, routes each group to direct verification or targeted search based on confidence, and keeps evidence returned by search in compact in-context memos for cross-claim reuse. To train this policy, we develop a privileged-teacher synthesis pipeline that converts gold claim annotations into executable multi-turn tool-interaction trajectories with live search rather than post-hoc rationales. Structural, label-alignment, tool-use, search-budget, and leakage checks yield 1,447 quality-controlled trajectories. We further construct 794 bidirectional same-trajectory preference pairs that keep claim grouping, search, and evidence fixed, enabling decision-focused Direct Preference Optimization (DPO) over factuality-decision tokens. The results with Qwen3-8B show that EAVer outperforms the strongest search-based baseline on each benchmark by 2.88 Macro-F1 points on VeriFastScore and 4.73 points on the out-of-distribution FaStFact-Bench, while using about 80% fewer searches than the most search-efficient baseline. Moreover, EAVer consistently improves performance across models ranging from 4B to 32B parameters, demonstrating its strong generalizability.
Read the full abstract
Machine Learning
Can Coding Agents Reproduce Official Statistics? Metadata, Retry Budget and the Limits of Execution Feedback in a Controlled Eurostat Benchmark
Large language models can generate executable data-analysis code, but successful execution is not equivalent to a valid official-statistics result. This study asks whether authoritative metadata and execution feedback improve the reproducibility of Eurostat answers produced by a coding agent, and isolates what execution feedback actually contributes. A benchmark of 30 natural-language tasks covering seven domains, seven Eurostat datasets and four difficulty tiers was run under four conditions: task only (A), task plus a frozen dataset metadata card (B), metadata plus a repair loop driven by sanitized execution feedback (C), and metadata plus the same attempt budget with no diagnostics of any kind (D). Claude Sonnet 5 generated Python through the Anthropic Messages API in three independent replicates, yielding 360 task-runs. Exact correctness required successful execution, the correct dataset, filters, output shape, values and unit. A companion experiment run under an under-specified output contract, in which the required ranking key and unit representation were never stated to the model, understated condition C by 23.4 points, showing that evaluator and contract design can dominate measured agent error. Reliable statistical coding agents need semantic validation against frozen specifications, a fully specified output contract, and a retry budget - not execution diagnostics.
Read the full abstract
Computation and Language
Team DArgk at the 2026 ELOQUENT lab for evaluating generative language model quality: Residuals of Humanity: AI Detection Evasion via GRPO Fine-Tuning
Large language models (LLMs) can generate fluent and coherent text that is increasingly difficult to distinguish from human writing, motivating the development of automatic AI-generated text detectors. However, the robustness of such detectors under adversarial generation remains uncertain. This paper presents SHADE (Stochastic Human-like generation via Adversarial Detector Evasion), a reinforcement learning framework that formulates detector evasion as a policy optimization problem. Instead of applying post-hoc perturbations or prompting-based rewriting, SHADE fine-tunes an instruction-tuned LLaMA model with Group Relative Policy Optimization (GRPO), using feedback from a surrogate detector based on the PAN 2025 mdok system. Our experiments show that full fine-tuning with a small KL regularization penalty achieves $98.5\%$ surrogate evasion, compared to $1.5\%$ for the base model, while LoRA-based adaptation is substantially less effective under regularization. Linguistic analysis reveals that successful evasion is associated with shorter, simpler, and less lexically diverse outputs, suggesting that high detector evasion does not necessarily correspond to more human-like writing. In the official Voight-Kampff competition setting, our submissions ranked sixth and seventh, indicating that optimization against a single surrogate detector only partially transfers to unseen evaluation classifiers. These results highlight both the potential and limitations of reinforcement learning for adversarial AI-text generation and motivate more robust, multi-detector evaluation protocols for AI-generated text detection.
Read the full abstract
Machine Learning
Measuring the Checker: Mutation Analysis for GPU-Kernel Benchmark Oracles
Benchmarks for LLM-generated GPU kernels decide correctness with a few random inputs and a loose floating-point tolerance, and their verdicts now feed leaderboards and reinforcement-learning rewards. Recent work agrees these checkers are weak and patches them by hand—extra input distributions, fuzzing recipes, tighter tolerances—with no way to measure whether any patch suffices. We introduce mutation analysis as an adequacy metric for kernel-benchmark oracles: deterministic rules inject 10{,}303 compilable faults into verified CUDA implementations of 188 KernelBench problems, 7{,}384 of them with an independent kill witness; any test protocol is scored by the fraction it detects. The official check misses one in six witnessed faults (16.9%), deterministically, and the misses are skewed by family: 8.7% of arithmetic faults escape, but 78.6% of precision faults do. The metric explains why (a tolerance blind band growing with reduction size; a measured ceiling on input aggressiveness set by legitimate floating-point variance), audits the strongest existing patch (KernelBench-Verified's gain splits into $+4.0$ points from hidden inputs and $+4.5$ from tighter tolerance, a split its authors could not compute), and exposes a published fuzzing recipe that rejects correct kernels 107 times. Optimizing suites over the kill matrix reaches 98.0% detection with two inputs per problem (94.8% held-out), and the measurement's fault taxonomy teaches a test generator more than the raw faults themselves. Across 48 whole architectures, the blindness grows with scale, concentrating in deep homogeneous pipelines, and two problems prove unrefereeable: their official references violate the benchmark's own tolerance against fp64. We release everything as \href{https://huggingface.co/datasets/Elfsong/KernelBench-M}{KernelBench-M}.
Read the full abstract
Computation and Language
Knowing, and Saying It Only When Asked: LLM Endognostics and the Schizognosis of Minerva-7B
Evaluating an aligned language model by reading its answers assumes the answers carry the distinction the evaluator cares about. We introduce LLM endognostics, a white-box internal auditing framework designed to extract and causally manipulate latent knowledge within the residual stream. Applied to Minerva-7B-Instruct-v1.0 on 124 minimal prompt pairs over 12 categories of professional risk, behavioral evaluation fails on most of the set: the model acts identically on 63.7% of the pairs (95% CI [55.0%, 71.6%]), complying with or refusing both members. Yet, projecting the residual stream onto the vocabulary by a Jacobian lens reveals a statistically significant Contrastive Endognostic Margin, proving the model maintains robust risk differentiation internally. In a second protocol crossing 25 facts with five linguistic framings, we show that the model conforms to presupposed falsehoods in 72% of cases, despite representing the true entity in its latent layers. Surgically ablating the direction of the planted falsehood restores the correct answer in 11 of 25 suppressed cases (McNemar p = 0.0010), validated by blind human annotation (binary agreement kappa = 0.68). In contrast, an out-of-sample linear probe achieves 77% accuracy at layer 10, but its orthogonal ablation yields a 0% recovery rate. This establishes a fundamental theoretical dissociation: abstract linear representation does not imply causal control over verbalization. Our main contribution is the formalization of endognostics to prove that behavioral evaluation and internal reading systematically disagree in the common case, and that linear decodability is decoupled from causal control.
Read the full abstract
Machine Learning
Toollery: Scaling LLM Agents to Thousands of Skills and Tools
As LLM agents are exposed to hundreds to tens of thousands of skills, tools, and API functions, full-library prompting becomes costly, slow, and less reliable: each added candidate increases prompt tokens and latency, while longer candidate lists introduce more distractors for LLM selection. We present Toollery, a training-free candidate-compression framework for scalable LLM skill/tool selection. Following established document-side query expansion, Toollery generates user-intent queries from each skill/tool specification and builds a retrieval index that maps real user requests to compact candidate sets before final LLM decision-making. By treating high-level skills and atomic tools as selectable capabilities, Toollery can be applied to both skill libraries and tool registries. We evaluate Toollery on the roughly 79K-capability SkillRouter benchmark, BFCL-V4 with over 440 atomic tools, and 3,396 proprietary smart-cockpit requests over 220 tools. Across these settings, Toollery keeps online selection bounded to a compact top-K candidate set and improves recall over ordinary specification retrieval. At a fixed top-10 budget, Toollery improves end-to-end selection on the cockpit dataset, and maintains comparable AST Accuracy on BFCL-V4. These results support Toollery as a practical candidate-compression framework for large and evolving agent capability libraries, while showing that quality and cost gains depend on workload coverage and provider caching.
Read the full abstract
Machine Learning
UniGIO: Unified Generative Global In-situ Weather Modeling from Spatiotemporal Incomplete Observations
Global In-situ Observation (GIO) provides fine-scale, direct records of the global weather system from sparse point stations, making it an indispensable source for capturing localized and transient dynamics beyond the reach of satellite gridded data, and playing a critical role in key fields such as numerical weather prediction, disaster prevention, and agriculture. However, GIO exhibits strong spatiotemporal incompleteness, severely impairing accurate and real-time in-situ weather modeling. Unlike existing methods waiting for completed AI-ready data with extra introduced errors, in this work, we explore UniGIO, a novel generative framework for directly modeling global in-situ weather dynamics from native incomplete GIO. By generating missing data from observed ones annotated by masks, it unifies the coexisting forecasting, imputation, and generation under arbitrary missing ratios. Between the missing and observed, UniGIO captures station and region level complementarity through the Observation Mixer and Event Aligner, which diffuse discrete observations into continuous spaces where weather processes naturally span multiple stations. We further establish temporal dependencies with pattern shifts using the Adaptive Temporal Mixer, and track extreme events in chaotic local weather systems through a Mixture-ofExperts structure. Steady and extreme events are adapted in decoder by a Local Refiner. Extensive experiments on the up-todate largest global station weather dataset Weather-5K validate its SOTA performance with 11%, 12%, and 5% advantages on accuracy, fidelity, and extreme event capture, delivering a novel holistic solution for weather modeling in GIO networks.
Read the full abstract
Machine Learning
The Effect of Quantization on Clinical Benchmarks: Accuracy and Safety Across Model Families
Quantization enables deployment of large language models on resource-constrained clinical edge devices, but its effect on clinical accuracy and safety remains understudied. We evaluate five 7-8B parameter models at FP16, GPTQ-INT8, and GPTQ-INT4 precision across five benchmarks: MedQA, MedMCQA, Med-HALT, a risk-stratified sample of HealthBench, and MedSafetyBench. The study jointly varies quantization bit width, model family, and clinical task type, with explicit risk stratification and safety measures. INT8 GPTQ is universally safe (max. degradation -1.9%-1.9%), while INT4 degradation is substantial and model-dependent: BioMistral-7B, clinically fine-tuned, loses 19.7% on MedMCQA, more than any general-purpose model, showing clinical fine-tuning does not confer compression robustness. MedMCQA degrades more than MedQA under INT4; Med-HALT is largely unaffected. On HealthBench's emergency-risk subgroup, Qwen2.5-7B degrades by 26.8% under INT4, suggesting high-risk scenarios are disproportionately vulnerable to compression. On MedSafetyBench, the model family dominates over precision (refusal rates range 10.2%-74.9% at FP16), though Qwen2.5-7B (-17.8%) and Meditron-7B (-28.3%) show substantial INT4 safety degradation; notably, Qwen2.5-7B is simultaneously the most accuracy-robust model, demonstrating that accuracy and safety robustness are independent properties. We additionally test two recovery methods, clinical calibration substitution and QLoRA fine-tuning, both producing the same trade-off: MedMCQA recovers while MedQA further degrades, indicating recovery strategies require task-specific validation rather than being assumed universally beneficial. These findings indicate INT8 is broadly safe for clinical deployment, while INT4 safety must be assessed per-model and per-task, and that safety alignment is determined primarily by instruction tuning rather than clinical domain adaptation.
Read the full abstract
Computation and Language
On Mitigation of Subliminal Learning in Large Language Models
Knowledge distillation can transmit unintended behavioral traits from a teacher model to a student through training data that appear semantically unrelated to those traits, a phenomenon known as subliminal learning. Although recent work has established this effect, its training dynamics and mitigation remain underexplored. We study subliminal learning in open-weight language models ranging from 1.5B to 8B parameters, covering the Qwen, Gemma, and Llama families in number-sequence and chain-of-thought settings. Rather than evaluating only final models, we track trait-related probabilities throughout fine-tuning and find that subliminal acquisition can be highly non-monotonic, with transient spikes, reversals, and trait-specific failures of transfer. We then introduce liminal training, an annealed KL-regularized fine-tuning method that constrains early drift from the base model. Across our experiments, liminal training substantially reduces subliminal trait acquisition while largely preserving task gains, outperforming paraphrasing and layer freezing as mitigation strategies. The effect also extends beyond animal preferences: in a French-language response-style experiment, liminal training suppresses language transfer while retaining much of the GSM8K improvement. Finally, we show that KL timing matters: early regularization is more effective than late regularization, and sweeping the regularization strength reveals an empirical trade-off between task learning and trait suppression.