Software Engineering
Papers
Fresh from arXiv, typeset properly. Highlighted words trade on the market.
All of arXiv
Monday 31 August
LLM-based coding agents combine a foundation model with a harness that shapes agent behavior. For non-trivial tasks, how practitioners structure their work with the coding agents determines whether reliable results follow. We report on a phased workflow for operating coding agents developed by the AI research team at Infobip. The workflow structures agent-assisted development into four phases where human effort is front-loaded and delegation increases as artifacts mature. Context management is the central concern, addressed through four strategies applied at each phase to counter known failure modes. From practitioner experience, we observe that upstream errors in research and planning can compound across later phases, while correcting generated code can introduce bloat and fragility. This motivates front-loading human review. We identify two open problems: the absence of metrics for workflow effectiveness and the gap between formalized context management components and the workflow-level patterns that practitioners need.
Differential Geometry
Convex hulls and area minimizing currents in asymptotically locally hyperbolic manifolds
We solve the oriented asymptotic Plateau problem for integral currents of arbitrary codimension in conformally compact asymptotically locally hyperbolic manifolds without global curvature assumptions. This extends Anderson's theorem in hyperbolic space. A new ingredient is the construction of a barrier hull near conformal infinity, which serves as a barrier for area minimizing currents and stationary varifolds. In the conformally compact setting, the associated convex hull identity extends Anderson's result beyond globally pinched negative curvature.
Read the full abstract
Machine Learning
Learning Dynamics of Logits Debiasing for Long-Tailed Semi-Supervised Learning
Long-tailed distributions are prevalent in real-world semi-supervised learning (SSL), where pseudo-labels tend to favor majority classes, leading to degraded generalization. While many long-tailed semi-supervised learning (LTSSL) methods have been proposed, the mechanisms by which they implicitly debias logits remain poorly understood. In this work, we revisit LTSSL through the lens of learning dynamics and provide a theoretical characterization of logits debiasing. Specifically, we derive a step-wise decomposition of the logits updates, showing that predictions are dominated by class-imbalance bias that reliably reflects label priors. To expose this effect, we use the logits of a task-irrelevant baseline image as an indicator of accumulated bias and prove that they converge to the class prior. This provides a unified view where LTSSL remedies such as logit adjustment, reweighting, and resampling correspond to reshaping gradient dynamics. Based on this insight, we propose DyTrim, a principle-based dynamic pruning framework that reallocates gradient budget through class-aware pruning on labeled data and confidence-based soft pruning on unlabeled data. We provide theoretical guarantees that DyTrim reduces class bias and improves generalization. Extensive experiments on standard LTSSL benchmarks show consistent gains across architectures and methods. Code available at: https://jiajun0425.github.io/DyTrim
Read the full abstract
Astrophysics of Galaxies
Constraining Baryonic Feedback at $z\sim 1$ with the H$\alpha$ Luminosity Function
Baryonic feedback remains one of the primary uncertainties in galaxy formation models because of its impact on the interstellar medium (ISM) and star formation. We investigate its effect on the H$\alpha$ luminosity function (LF) using the Illustris and IllustrisTNG simulations. We incorporate direction- and metallicity-dependent dust attenuation, thermal sputtering and a physically motivated treatment of the propagation of Lyman-series photons. We model three sources of H$\alpha$ emission: star-forming H II regions, collisional excitation and recombination in the diffuse ISM, and AGN-induced photo-excitation of diffuse gas. The models are calibrated against observed H$\alpha$ LF at $z\sim1-1.5$ using a single free parameter relating the star formation rate to the H II-region H$\alpha$ luminosity. We find that, for $L_{\text{H}\alpha}\geq 10^{41}\ \rm{erg \ s^{-1}}$, emission from diffuse gas contributes $\sim 5\%-15\%$ of the total H$\alpha$ luminosity in both simulation sets. AGN photo-excitation produces only a modest enhancement of the bright end of the LF, although it dominates the H$\alpha$ emission of Illustris galaxies with $L_{\text{H}\alpha}\gtrsim \times10^{43}\ \rm{erg \ s^{-1}}$. The low-feedback IllustrisTNG simulations provide a good fit to the observed LF for all emission models considered, whereas the high-feedback Illustris predictions are difficult to reconcile with both the bright and faint end of the LF. This tension between H$\alpha$ LF and stacked kSZ measurements highlights the combined power of these probes for constraining the interplay between baryonic feedback, gas properties, and the physics governing emission-line galaxies.
Read the full abstract
Quantum Physics
Flatness-Preserving Operations
A quantum state is called flat if it is proportional to a projector. There has been recent interest in studying antiflatness, the property of diverging from flat states, and establishing a resource theory for it. Identifying the free operations (the Flatness-Preserving Operations (FPOs)) remained an open problem. For this purpose, we first discuss Orthogonality-Preserving Operations (OPOs), of which trivial examples are unitary operations in an isolated system. More generally, we give a simple proof that all OPOs are isometric embeddings consisting of combinations of unitaries/isometries and appending a fixed state. We then show that all FPOs are either constant maps to some fixed flat state or a special case of an OPO, where the appended fixed state must be a flat state. We also show that the only possible flat convex combinations of flat states are those of orthogonal flat states with weights given by their purities.
Read the full abstract
Computational Engineering
Domain-Grounded Tool Orchestration for LLM-Guided Scientific Analysis
Scientific analysis workflows encode deep domain knowledge through sequences of tightly coupled operations where correctness depends on tool selection, execution order, and parameterization. A CFD engineer investigating flow separation must extract wall shear stress, identify zero-crossings in skin friction, and confirm with boundary-layer profiles: a chain that requires both domain expertise and proficiency with visualization tools. Current approaches to LLM-assisted scientific visualization generate scripts that encode this knowledge implicitly, and often incorrectly, producing code that executes but yields wrong results. We present an architecture that separates intent interpretation (LLM) from execution (deterministic domain tools) from explanation (LLM), connected by the Model Context Protocol (MCP) and grounded by domain ontologies that constrain planning to valid analysis chains. We instantiate the architecture in two domains on the same ParaView server infrastructure: computational fluid dynamics post-processing and topological data analysis via the Topology ToolKit (TTK). Adding the second domain required only an ontology and tool wrappers around existing filters, with no change to the architecture, protocol, or deployment. By construction the design removes whole classes of failure that affect script generation (such as API hallucination and missing pipeline stages) and narrows the strategic errors that remain. An ablation across both domains locates the ontology's empirical effect: it does not change which tools the planner selects, which is already reliable, but corrects how the model interprets results, raising interpretation accuracy from 0.41 to 0.91, and only when the relevant fact is retrieved in scoped rather than bulk form. ParaView's client-server model carries analysis to production-scale datasets through a thin browser client.
Read the full abstract
Machine Learning
Liquid Gated Attention
Real-world time series often exhibit irregular sampling and extended temporal horizons, requiring models to capture continuous-time dynamics across arbitrary intervals without prohibitive scaling costs. Discrete-time methods collapse variable time intervals into static positional steps; solver-dependent continuous-time models preserve temporal structure but rely on sequential integration, precluding parallelization; and solver-free approximations avoid this cost yet none couples observed time intervals with input-driven state modulation. We propose Liquid Gated Attention (LGA), a solver-free parallel temporal operator. By parameterizing an input-driven gating mechanism with observed time intervals, LGA introduces a continuous-time inductive bias and formulates hidden state evolution as a fast-weight associative memory, enabling parallel computation across the temporal dimension. Using matrix associativity in non-causal encoding and a prefix scan in causal encoding, LGA attains linear temporal complexity in sequence length in both modes. A sequence-level normalization bounds cumulative temporal decay for stable long-horizon optimization. Building on LGA, we instantiate LFormer, a modular backbone for continuous-time representation learning. Across six tasks and sixteen datasets spanning up to 17,984 steps, LFormer demonstrates long-range dependency modeling, fine-grained state tracking, and trajectory reconstruction from sparse and noisy observations, while delivering competitive performance against state-of-the-art discrete-time and continuous-time baselines with linear scaling efficiency.
Read the full abstract
Human-Computer Interaction
Inferring Value Criteria from Ordinal Preferences: An Iterative In-Context Learning Framework for Music Generation
Adapting a generative music system to an individual's taste requires learning what that listener values. Listeners can rank pieces, but their underlying criteria may be tacit and difficult to articulate. We ask whether and under what conditions a large language model (LLM) can adapt symbolic music generation from rankings alone and construct transferable natural-language descriptions of value criteria. In our iterative in-context learning framework, the LLM formulates hypotheses, generates candidate pieces in ABC notation, receives a ranking, and periodically infers and verbalizes value criteria from history to guide later generation. We evaluate the framework against 16 simulated raters in 480 adaptation runs using mixed-effects modeling, an ablation, and transfer tests on unseen music. Overall, the framework did not outperform a feedback-free diverse-generation baseline, but did so for two value functions with targets difficult to reach through simple sampling. How atypical the target was relative to the LLM's feedback-free generation tendencies predicted adaptation difficulty. Moreover, higher value during adaptation did not imply identification of the criterion as a general rule. On unseen music, acquired descriptions and histories improved generation for more value functions than they improved preference prediction, which remained near chance. Some gains were associated with acoustic proximity to music in the context, but others were not. These findings show that rankings alone can guide generation under limited conditions, while transferable criterion inference remains constrained by the foundation model's ability to recognize, reason about, and verbalize musical attributes.
Read the full abstract
Numerical Analysis
Efficient primal–dual splitting methods for a Poisson-constrained JKO scheme for Poisson-Nernst-Planck models
The Poisson–Nernst–Planck (PNP) equations strongly couple ionic transport and electrostatic interactions through the Poisson equation, posing substantial numerical challenges under small permittivity and complex potential boundary conditions. Underlying these equations is a natural Wasserstein gradient-flow structure, in which the Poisson equation serves as a local realization of the nonlocal electrostatic interaction energy. Exploiting this structure, we formulate each time step as a constrained convex minimization problem where the ionic continuity equations and the Poisson equation are incorporated as linear constraints, allowing the concentrations, fluxes, and electrostatic potential to be updated simultaneously. The variational structure of the scheme intrinsically guarantees the dissipation of the original free energy, mass conservation, and nonnegativity of ionic concentrations under general electrostatic boundary conditions. Moreover, the framework is structurally modular: extending from classical to modified PNP models with steric interactions and concentration-gradient corrections requires only modifying the energy functional, while all structure-preserving properties are automatically retained. To efficiently solve the resulting large-scale constrained problems, we develop preconditioned and transformed primal–dual algorithms equipped with tailored fast dual solvers, namely DCT-based direct and Schur-complement iterative methods, that exploit the coupled block structure of the PDE constraints. Numerical experiments on classical and modified PNP systems demonstrate the accuracy and structure-preserving properties of the scheme, and show that the proposed algorithms converge reliably in strongly coupled small-permittivity regimes without significant growth in computational cost.
Read the full abstract
Computer Vision
Can Video World Models Track Unobserved World States?
Video world models are increasingly used as simulators, but visual fidelity alone does not show that a model maintains the hidden state of the world. We examine this difference with an action-conditioned video Shell Game, a visual analogue of $S_5$ state tracking that separates visual rendering from compositing the unobserved world state. Trained on 5-swap chains, standard backbones (e.g., bidirectional and autoregressive Transformers, Mamba, and linear attention) render plausible videos and predict the correct ball location up to 5 swaps. However, they fail to learn the rule and generalize to longer swap chains, even with more denoising steps. As the pixel-based diffusion loss does not force the generated frames to hold the unseen ball position, output tokens cannot carry it, and the state has to live within the architecture. In a causal Transformer, this implicit state is an append-only KV cache, which is written once and never revised, so the model must re-compose the swaps at every chunk. Tracking $S_5$ this way requires depth to grow with sequence length, which no fixed-depth Transformer provides. We study what enables learning the rule, and find that length generalization requires a revisable state carried across chunks and an update expressive enough to apply a swap. Linear attention can achieve this by allowing negative transition eigenvalues, and autoregressive Transformers can do so with nonlinear TTT fast weights (e.g., SwiGLU) whose online updates change the feature map used to read their state. We further examine Memory Maze and Block World, where the state is not fixed by the input action stream alone and must be corrected from observations or keeps changing out of view, and discuss the implications for building stateful video world models.
Read the full abstract
Mesoscale and Nanoscale Physics
Coexistence of polariton bound states in the continuum and 2D radiative excitons
Bound states in the continuum (BICs) enable optical modes with ideally infinite radiative lifetimes despite lying within the radiation continuum. Radiation-matter interaction in periodically patterned planar waveguides embedding two-dimensional (2D) optically active excitations can be described quantum mechanically by diagonalizing a non-Hermitian operator, known as the Hopfield matrix, which can be generalized to incorporate independent photonic and excitonic losses into the polaritonic states. However, since 2D excitons undergo intrinsic wavevector-dependent radiative decay within the light cone, whether the Hopfield formalism can consistently account for this process while preserving polariton BICs has remained an open question. Here we show that a non-Hermitian Hopfield formalism incorporating excitonic radiative losses correctly captures the existence of genuine $k=0$ polariton BICs with diverging radiative lifetimes. The theory provides a unified microscopic framework for radiative excitons and polariton BICs, and an efficient predictive tool for designing photonic-crystal platforms coupled to quantum wells, transition-metal dichalcogenides, and other 2D excitonic materials.
Read the full abstract
Computer Vision
Failure or Drift? Evaluating Monocular SLAM under Synthetic and Real-World Corruptions
Visual SLAM is commonly evaluated on clean trajectories, although deployment failures are often caused by adverse weather, illumination, blur, and sensor artifacts. Controlled corruptions are attractive because they isolate such factors, but a synthetic stress test is useful only when it leads to the same engineering conclusion as the condition it is intended to approximate. This work examines that question for monocular SLAM. We evaluate a classical feature-based system and two learned trackers under image-space, geometry-aware, and compound corruptions, and compare their behavior with adverse conditions from 4Seasons. Rather than reducing robustness to a single trajectory error, the evaluation separates explicit tracking failure from drift accumulated by methods that remain active. The results show that learned trackers largely replace catastrophic loss with sustained, and sometimes severe, drift. More importantly, the apparent ordering of the learned systems changes with the physical fidelity of the corruption: structured rain and fog proxies preserve the real-world ordering, whereas a simple illumination proxy does not. Code is available at: https://github.com/abhaythomas/master_thesis_vslamlab_robustness.
Read the full abstract
Computer Vision
CANVAS: Consistency-Aware Navigation via Visual Adaptive Sampling for Long-Context Text-to-SVG Generation
Autoregressive large models have recently advanced Text-to-SVG generation from simple icons to complex, long-context graphics, yet standard autoregressive decoding often fails to maintain global consistency across geometry, layout, occlusion, and composition. We introduce CANVAS (Consistency-Aware Navigation via Visual Adaptive Sampling), a training-free, render-aware inference framework that combines power-sharpened trajectory likelihood with visual feedback from rendered futures and derives a stroke-wise navigation rule. It effectively estimates each candidate stroke's future value under a limited generation and rendering budget and adaptively allocates samples according to candidate uncertainty, decision influence, and rollout cost. Experiments across multiple autoregressive SVG backbones and complementary benchmarks demonstrate improvements in global consistency, which includes sound geometric relationships, spatial layouts, occlusion ordering, and overall composition, without additional training, demonstrating the effectiveness and generalization ability of our framework.
Read the full abstract
Computer Vision
UFPR-PEs: A Brazilian Face Recognition Benchmark with Self-Declared Race/Color Labels
While face recognition systems are widely deployed, ensuring their demographic reliability and robustness under uncontrolled visual conditions remains a critical challenge. To bridge this gap, we present UFPR-PEs, a benchmark for face recognition bias evaluation using public videos of elected Brazilian politicians annotated with official self-declared race/color categories. The dataset adopts the Brazilian census taxonomy, including the parda category, which has no direct equivalent in the U.S.- or Europe-centric schemas commonly used in prior benchmarks. Our benchmark is built from compressed public video and preserves difficult samples so that performance can be analyzed under realistic conditions. We describe the construction pipeline, report dataset statistics, and evaluate face recognition performance across verification and (closed- and open-set) identification settings, including subgroup analysis by race/color and difficulty level. The results show that recognition performance varies substantially with image quality, and that subgroup gaps must be interpreted jointly with visual difficulty rather than in isolation. Overall, UFPR-PEs provides a reproducible and demographically grounded setting for studying face recognition bias under challenging public video conditions.
Read the full abstract
High Energy Physics (Phenomenology)
Small-x TMD distributions: JIMWLK evolution and the Gaussian truncation
We extend the systematic study of small-$x$ Transverse Momentum Dependent (TMD) distributions, initiated in [Cougoulic:2026drr], by including the energy evolution. We use the JIMWLK evolution equation to study the energy dependence of three distributions in the quark-gluon sector, $\mathcal{F}_{qg}$, and seven distributions in the gluon-gluon sector, $\mathcal{F}_{gg}$. The initial conditions are provided by the McLerran-Venugopalan model. We study these TMD distributions in position space and with $N_c = 3$ fixed. We then compare these results to the Gaussian truncation of the Balitsky hierarchy. Although the Gaussian truncation worked very well for the TMD distributions at the initial condition, now we identify some cases where the Gaussian truncation provides a reasonably good description, but in the majority of cases it fails to fully reproduce numerical data. Using the fixed representation building blocks, $\Omega_{ag}^\omega$, we are able to partially explain the outcomes. This work constitutes the next step with the goal of understanding the interplays between the large-$N_c$ limit and Gaussian truncation as well as the BK/JIMWLK evolution equations for phenomenologically relevant observables.
Read the full abstract
Cryptography and Security
Beyond the Payload: How User Invocation Shapes Coding Agent Vulnerability to Repository Poisoning
Coding agents are increasingly used for software engineering tasks, including bootstrapping projects from third-party repositories whose integrity cannot be assumed. Prior work on repository poisoning largely focuses on attacker-controlled injection and disguise, but developers also shape risk through everyday invocation choices: what task to delegate, how to phrase the request, and which skills or rules to supply. We term these user-side choices Prompt-Level Configurations (PLCs) and introduce CIPR (Coding In Poisoned Repos), the first benchmark that systematically varies PLCs in poisoned real-world repositories. CIPR comprises 1,920 instances across 20 repositories, four task types, three social-media-grounded prompt styles, and three skill/rule conditions, and measures attack success rate (ASR) and agent alert rate (AR) using automated runtime and trace-based oracles. Our evaluation reveals two key insights: (1) Vulnerability is highly context-dependent, with task type creating up to a 4.5-fold difference in ASR, with test-execution task forming a silent attack surface (high ASR, low AR). (2) Prompt expression shifts risk indirectly: underspecified prompts reduce ASR by truncating execution depth; noisy prompts exhibit a directional trend toward suppressing alerts by making malicious content less conspicuous. These findings highlight that coding agent vulnerability is not a static property, but a dynamic outcome shaped by everyday user configurations.
Read the full abstract
Artificial Intelligence
ATLAS: Dual-Horizon Diagnostic Evaluation for Tool-Use Agents in Meituan Xiaotua
LLM-based assistants on web-scale service platforms increasingly address local-service needs through multi-step planning and tool use. Under dynamic business conditions and evolving user needs, outcome-level evaluation offers limited evidence of where execution fails or how service drifts across continued interaction. We present ATLAS, a dual-horizon diagnostic evaluation framework for industrial tool-use agents. Within-turn evaluation uses trajectory-wise signals across thinking and reflection, tool and skill execution, and response generation to identify execution failures and unmet capability requirements. Across-turn evaluation uses user-wise signals spanning multiple lifecycles to diagnose failures in maintaining or updating user intent, constraints, and corrections. Fine-grained signals use rule-based checks and LLM judgments with explicit evidence scopes and decision boundaries. To support reliable diagnosis at production scale, LLM judges are calibrated against dynamically maintained references from business logs, and selected judges are distilled into compact diagnostic models. The resulting diagnostic profiles help developers attribute failures, prioritize targeted interventions, and reassess improvements. Applied in Meituan Xiaotuan, a million-scale DAU local-services assistant, ATLAS achieves high diagnostic fidelity across calibrated judges and distilled models. Blinded human auditing and a 14-day online A/B test with nearly two million users per arm confirm that the ATLAS-guided agent improves recommendation relevance, factual reliability, user engagement (+2.84% answer read-through rate), and order volume (+6.13% paid orders), leading to full production deployment.
Read the full abstract
Metric Geometry
Improvement of the dimension bound for unweighted RCD spaces
We show that if $(X,d,\mathscr H^n)$ is an $RCD(K,N)$ space for some $K\in\mathbb{R}$ and $N\in [1,\infty)$, then it is also a (non-collapsed) $RCD(K,n)$ space. In other words, unweighted finite-dimensional $RCD$ spaces enjoy an automatic improvement of the dimension bound to the Hausdorff dimension of the space. More generally, we prove that this holds also when $N=\infty$, provided that we assume in addition the $n$-rectifiability of the space.
Read the full abstract
Computation and Language
WildSEEK: Evaluating Language Models for Information-Seeking
Language models are increasingly mediating information access to end users, urging a systematic evaluation of their responses for a fair and reliable information ecosystem. Existing evaluations, however, are often topic-specific or synthetic, limiting their ability to capture the complexity of "in the wild" information-seeking queries and the risks present in model responses. To address this gap, we introduce WildSEEK, a manually annotated dataset of 3k information-seeking queries from real user interactions, and an evaluation framework for LLM-generated responses. WildSEEK includes annotations for risk-sensitive domains (e.g. health and financial information), and distinguishes factoid queries from analytical queries which seek responses beyond facts. We train classifiers on WildSEEK to analyze more than 1.8M realistic user queries. We find that over a third of information-seeking queries are high-risk and more often analytical. Our findings show that LLM responses fail more often in four criteria: sycophantic behavior, overreliance, a default US-centric perspective, and poor handling of vulnerable populations – with failure rates being mostly higher for analytical queries. By providing methods to monitor the reliability, safety, and fairness of LLM behavior, our dataset and evaluation framework offer an empirical foundation for the broader question of how these systems should behave as they take on a growing role in information access.
Read the full abstract
Machine Learning
Learning Materials Properties from Scarce Labels and Unlabeled Crystals
Learning materials properties from scarce labels and unlabeled crystals is a central challenge for data-driven materials discovery. We present SemiMat, a controlled benchmark for semi-supervised materials property regression, and MatRank, a reliability-weighted objective for continuous pseudo-label uncertainty. SemiMat fixes labeled and unlabeled crystal inputs, graph-backbone interfaces, validation-only checkpoint selection, held-out test reporting, normalized MAE (NMAE), and method-rank summaries across six scarce-label tasks, four graph backbones, and five predefined split runs. MatRank builds pseudo-targets from labeled anchors, weights them by local reliability and weak-prediction agreement, trains weak and strong graph views consistently, and adds ranking signals so that unlabeled crystals shape both values and candidate order. Across the retained 24 backbone-task blocks, one fixed MatRank objective gives the lowest aggregate held-out test NMAE (0.896) and best average method rank (2.208). The component, OOD, and generated-pool diagnostics identify where the gain is reliable and where further screening evaluation remains necessary. Code is available at https://github.com/littlepeachs/SemiMat.
Read the full abstract
Quantum Physics
Entanglement-enabled Criticality in One-dimensional Quantum Contact Process
The contact process is a paradigmatic example of nonequilibrium dynamics, with broad applications ranging from chemistry to sociology. Its quantum counterpart, the quantum contact process (QCP), extends the classical model to include coherent processes. Despite sustained interest, the nature of the transition in the one-dimensional (1D) QCP remains debatable. Here, combining Liouvillian spectral analysis, the tensor jump method, exact quantum jump Monte Carlo, and truncated Wigner simulations, we show that 1D QCP undergoes a continuous absorbing-state phase transition, with critical exponents distinct from the classical case. We further find Liouvillian gap closes well below the critical point, highlighting that spectral gap analysis alone cannot distinguish a phase transition from metastability in the QCP. Crucially, the 1D QCP is weakly entangled, yet even this weak entanglement is indispensable for capturing the correct critical behavior, whereas semiclassical methods artificially stabilize the active state and predict a spurious first-order transition. Our work establishes the quantum origin of the phase transition in the 1D QCP and underscores the essential role of entanglement in dissipative quantum many-body systems.
Read the full abstract
Materials Science
Halide donors in monoclinic- and corundum-phase Ga$_2$O$_3$ and Al$_2$O$_3$
We present a systematic first-principles investigation of halide impurities (F and Cl) in Ga$_2$O$_3$ and Al$_2$O$_3$, considering both monoclinic and corundum phases. Our study of the structural properties, formation energies, and charge-state transition levels establishes the relative stability of different atomic configurations and charge states. We find that F and Cl on oxygen sites act as shallow donors in Ga$_2$O$_3$ in both the monoclinic and corundum phases. However, their behavior differs substantially as the band gap increases with greater Al compositions. Fluorine is prone to $DX$-center formation with increased Al composition, leading to self-compensation at 38% Al concentration in monoclinic (Al$_x$Ga$_{1-x}$)$_2$O$_3$ and 70% Al concentration in corundum (Al$_x$Ga$_{1-x}$)$_2$O$_3$. Chlorine is more resistant to $DX$-center formation: in monoclinic (Al$_x$Ga$_{1-x}$)$_2$O$_3$, Cl$_\mathrm{O}$ on the lowest-energy oxygen site shows an onset of $DX$ behavior at 50% alloy composition, while in corundum (Al$_x$Ga$_{1-x}$)$_2$O$_3$ this onset for Cl$_\mathrm{O}$ occurs only at Al concentrations as high as 84%. We also study F and Cl interstitials, finding that they act as compensating centers but also exhibit migration barriers that are low enough for them to be removed by post-growth annealing. Surprisingly, in both monoclinic and corundum Al$_2$O$_3$, Cl$_\mathrm{O}$ exhibits a relatively shallow transition level located at $\sim$0.48 eV below the conduction-band minimum, much shallower than F$_\mathrm{O}$ and other donor candidates. These remarkable results identify Cl as an unusually promising donor candidate in (Al$_x$Ga$_{1-x}$)$_2$O$_3$ alloys and even pure Al$_2$O$_3$, although high formation energies and compensation will render observation of true $n$-type conductivity in Al$_2$O$_3$ difficult.
Read the full abstract
Computation and Language
LCoT-GV: Graph Attention Networks for Verifying Long Reasoning Chains in Large Language Models
Large Reasoning Models produce Long Chains-of-Thought (LCoTs) which involve breaking down the problem into smaller reasoning steps before reaching the conclusion. However, these steps often contain contradictions, unsupported inferences, or irrelevant steps, even when the final answer is correct. We propose Long Chain-of-Thought Graph Verifier (LCoT-GV), a graph-based framework that represents LCoTs as reasoning graphs. Each node in the graph represents a reasoning step and the edges encode semantic and logical relations. A Graph Attention Network is then trained to predict chain-of-thought correctness from the reasoning graph. We construct a new graph-oriented verification dataset from multiple reasoning benchmarks for question answering in various domains. The results show that our method is competitive with the most similar approaches.
Read the full abstract
Computation and Language
OCR-MetaReasoning Benchmark: Evaluating the Meta-Reasoning Ability of MLLMs in Text-Rich Image Understanding
Text-rich image understanding requires multimodal large language models (MLLMs) to organize OCR (Optical Character Recognition)-grounded evidence across words, layout, fields, charts, and visual correspondences. Existing evaluations often conflate extraction with reasoning and rarely test whether models follow the required reasoning direction: applying visible rules, abstracting hidden regularities, or recovering missing premises. We introduce OCR-MetaReasoning, a controlled single-image benchmark that treats deduction, induction, and abduction as distinct directions and separates final-answer correctness from reasoning-process compliance. The benchmark contains 1,500 verified samples in a balanced \(3\times5\) taxonomy crossing three reasoning types with five OCR-object categories, along with reference reasoning steps, automatic answer scoring, the Meta-Reasoning Macro Score (MRMS), and the Reasoning Process Compliance Score (RPCS). Experiments with representative closed-source and open-source MLLMs show that OCR-grounded meta-reasoning remains far from saturated: models struggle with visible-rule application and layout-sensitive inference, while process-compliant rationales can accompany incorrect final answers under exact-match evaluation. The code is available at https://github.com/gengxuli/OCR-MetaReasoning.
Read the full abstract
High Energy Astrophysical Phenomena
STOKES tables and KY codes
The family of KY codes is an XSPEC-compatible package of fully relativistic models for spectro-polarimetric modeling of X-ray binary systems and active galactic nuclei. Within Kerr space-time, they compute accretion disc line profiles, reflection spectra from discs illuminated by Comptonized coronal radiation, and thermal disc emission including disc-corona interaction and disc self-irradiation. Many of these models include polarimetric computations, and more advanced timing variants compute X-ray, UV and optical reverberation. The primary source of Comptonized X-rays is assumed to be in the lamp-post on-axis or slab above the disc geometry. The KY codes operate at various levels of complexity; some models use analytical approximations for local re-processing, while others rely on detailed numerical tables. For the most accurate treatment of rest-frame re-processing in partially ionized accretion disc atmospheres, the KY codes employ pre-computed spectro-polarimetric tables generated with the STOKES code. These transmission and reflection tables are also designed for standalone use in XSPEC and as inputs to higher-level models beyond KY. The chapter first introduces the STOKES tables in various variants and shows their non-relativistic usage in XSPEC models for (i) an extended source illuminating a distant geometrically thin disc, and (ii) a compact source illuminating a vertically extended equatorial obscurer (e.g., a torus or thick wind). The relativistic KY ray-tracing is then described, followed by an introduction to each KY model. Examples of fits to the IXPE data of X-ray binaries and active galactic nuclei using these models are shown. Together, these models form a comprehensive and publicly available framework for spectro-polarimetric modeling and data fitting of accretion. We provide links to their online repositories.
Read the full abstract
Artificial Intelligence
MedAgent-R1: Faithfulness-Aware Reinforcement Learning for Evidence-Grounded Medical Reasoning
When medical AI systems hallucinate clinical reasoning, the consequences extend beyond incorrect answers: fabricated justifications that superficially reference retrieved evidence can mislead clinicians into unsafe treatment decisions. Medical reasoning agents must therefore produce not only correct answers but also faithful justifications that clinicians can verify against cited evidence. We identify a systematic failure mode in RL-trained retrieval agents: outcome-only rewards improve accuracy while degrading faithfulness, a phenomenon we term confident hallucination. The agent learns to answer from parametric memory and backfill plausible but unsupported justifications; citation fabrication rates rise from 16.5% to 31.8% even as accuracy improves by 5 points over the supervised baseline. We address this with a faithfulness-gated reward design: accuracy credit is conditioned on evidence grounding via a hard gate, complemented by retrieval validity and conciseness signals that close exploitation paths unique to agentic retrieval. The resulting system, MedAgent-R1, reduces citation fabrication from 31.8% to 4.7% and raises evidence completeness from 58.7 to 82.6 while maintaining 75.1% accuracy, with 13.2-point gains on HealthBench Safety. Under the same agentic retrieval setup, MedAgent-R1 outscores GPT-4o on faithfulness-specific dimensions (Factual Support 4.55 vs. 4.25; Overclaiming 4.40 vs. 4.15) while remaining below GPT-4o in overall accuracy, suggesting that explicit faithfulness training yields evidence-grounding gains not achieved by scaling alone.
Read the full abstract
Fluid Dynamics
Guided navigation in turbulence
Navigation in turbulence is challenging because local flow measurements provide limited information about favorable paths beyond the flow correlation scales. We investigate vertical navigation by swimmers in homogeneous isotropic turbulence and compare strategies using local flow information with strategies guided by high-performing trajectories. Compared with naive upward swimming, local strategies enhance the mean vertical velocity by about 10% of the root-mean-square flow velocity, while guidance by precalculated trajectories reaches enhancements of about 40%. This gain is limited by the time required to intercept favorable trajectories, particularly for slow swimmers. We then show that precalculated trajectories are unnecessary: swimmers can dynamically identify high-performing members of a swarm and use them as targets. For sufficiently fast swimmers, this collective strategy keeps much of the benefit of precalculated guidance using only the instantaneous swarm state. Our results demonstrate how distributed trajectory information can be used to overcome limitations of local sensing for navigation in turbulence.
Read the full abstract
Machine Learning
CoMPASS: Collaborative Molecular Property Prediction via Adaptive Small-Large Model Synergy
Accurate molecular property prediction requires both statistical reliability and chemical reasoning. Graph neural networks can be calibrated directly on labeled assays but remain limited by the coverage of their training data. Large language models (LLMs) can compare molecular evidence and articulate chemical rationales, yet are unreliable as standalone quantitative predictors. The central challenge is therefore to determine when an LLM should influence a calibrated model and by how much. Here we present CoMPASS, a retrieval-calibrated framework for small-large model collaboration. CoMPASS retains a graph attention network (GAT) as the predictive anchor, retrieves locally relevant training molecules, provides attention-grounded evidence to an LLM, and converts its proposal into a bounded correction through an agreement-aware gate. Across six classification and two regression benchmarks, CoMPASS improves the GAT anchor in regions of correctable uncertainty while limiting LLM intervention in high-confidence regimes. Ablations show that the gains arise from validation-calibrated retrieval and bounded fusion rather than prompting alone. These results suggest that generative reasoning should augment calibrated prediction through evidence-grounded, controlled corrections rather than direct output replacement. Code is available at https://github.com/littlepeachs/CoMPASS.
Read the full abstract
Robotics
CIG-RL: Curiosity-Driven Information-Guided Reinforcement Learning for Source Term Estimation in Uncertain Environments
Source term estimation (STE), which aims to estimate key properties of the gas source, is essential for identifying hazardous gas releases. Information-theoretic approaches have been adopted for autonomous STE using mobile sensors due to robustness in noisy environments, yet their online action selection incurs substantial computational cost. Deep reinforcement learning (DRL) provides a promising alternative with its fast decision-making capability. In DRL-based STE, the agent selects actions based on belief states of the source term updated from noisy measurement sequences. However, existing methods rely on random exploration or solely on belief uncertainty reduction without an effective exploration strategy in DRL, which can limit policy robustness in noisy environments. To address this, we propose a curiosity-driven information-guided reinforcement learning for robust and efficient STE. The proposed method promotes active exploration of novel belief state transitions that have not been sufficiently explored during training. We further introduce an uncertainty-adaptive active perception reward to guide efficient source search under uncertainty. Simulations under high-noise conditions and real-world experiments demonstrate the robustness and feasibility of the proposed framework, highlighting its potential for practical STE problems.
Read the full abstract
Artificial Intelligence
HiRS-Agent: A Hierarchical Multi-Agent System for Reliable Long-Horizon Remote Sensing Task Solving
Recent advances in large language models and multimodal models have pushed remote sensing (RS) processing from simple perception models to agentic systems designed to tackle complex, long-horizon RS tasks. However, existing systems often rely on monolithic decision-making frameworks, which fail to accommodate the multi-stage, interdependent nature of RS tasks. This centralized approach leads to challenges such as unstable task execution, incorrect tool usage, and error propagation across stages. To address these issues, we propose HiRS-Agent, a hierarchical multi-agent system for long-horizon RS task solving. HiRS-Agent adopts a two-level collaborative architecture: the Manager Layer handles dynamic routing, step-level verification, replanning, and termination control, while the Specialist Layer organizes domain-specific tools according to the RS workflow and is responsible for subtask reasoning and tool execution. To further enhance the system's capability, we introduce a two-stage supervised tuning strategy and a verification-guided hierarchical reinforcement learning stage to jointly optimize coordination and tool-use policies. Experiments on Earth-Agent Benchmark and ThinkGeo show that HiRS-Agent substantially improves long-horizon tool-use capability and final-task correctness, demonstrating the effectiveness of structured multi-agent collaboration for reliable RS agents. The code is publicly available at https://github.com/IntelliSensing/HiRS-Agent.