Papers

Speech representation learning requires a discrete modeling interface while preserving high waveform fidelity. Existing approaches rarely achieve both: discrete tokens ease modeling but lose fine-grained acoustic information, while continuous latents preserve fidelity but require specialized generative objectives. We introduce SemiCodec, a novel Semi-discrete neural speech Codec with synchronized discrete tokens and continuous latents. Direct joint optimization may create a reconstruction shortcut through the continuous pathway, leaving the discrete tokens insufficiently grounded. To address this issue, we propose a Preserve-then-Ground training paradigm that first establishes a high-fidelity synthesis space and then grounds SSL-guided discrete tokens within it. Experiments show that SemiCodec outperforms fully discrete codecs in reconstruction quality and intelligibility while remaining competitive with continuous representations in acoustic fidelity and speaker preservation. Audio samples are available at https://pennyw1-se.github.io/SemiCodec-Demo-Page/

Robust manipulation under uncertain object dynamics and geometry remains a core challenge in real-world robotics. Policies trained under fixed assumptions often degrade when object mass, friction, scale, or shape vary. Embedding-based adaptation methods such as Rapid Motor Adaptation (RMA) address this by conditioning actions on latent representations of environment dynamics inferred from interaction history. We investigate whether this paradigm scales from single arm settings to coordinated bimanual manipulation, where higher dimensional control, coupled contacts, and shared object forces introduce additional complexity. We evaluate bimanual RMA on three contact-rich ManiSkill tasks with randomized YCB object properties. Bimanual RMA, using observation-action history and dual wrist depth images, retains privileged policy performance on bimanual picking and lifting, while an otherwise matched stationary PPO policy without latent conditioning achieves low success; all methods remain limited on long-horizon stacking. We further demonstrate direct sim-to-real transfer on a dual Franka lifting task without real-world finetuning. Together, these results suggest that latent conditioning provides useful structure for scaling robust control to coordinated bimanual systems, while longer horizon tasks require additional temporal structure.

This paper presents a reproducibility study of “Boosting the Visual Interpretability of CLIP via Adversarial Fine-Tuning”, published at ICLR 2025, which proposes an unsupervised adversarial fine-tuning (AFT) method with norm regularisation to enhance the visual interpretability of CLIP's image encoder. We attempt to reproduce the key claims regarding improved saliency map quality, increased concept alignment, transferability to out-of-distribution datasets, and the trade-off with zero-shot accuracy. Beyond reproduction, we propose a saliency-guided regularisation extension that introduces an Energy Pointing Game loss, directly supervising the spatial alignment of Simple Gradient saliency maps with target objects. We evaluate our extension across a range of saliency-loss weights and show that explicit saliency supervision \rev{substantially} improves localisation metrics at a modest cost: the reduction in zero-shot accuracy and adversarial robustness is small at the lowest saliency-loss weight ($w_s=0.2$) and moderate at the strongest setting ($w_s=0.6$). We restructure the codebase for configurable, systematic experimentation.

The $(d+1)$-type $\Lambda$-Wright–Fisher process with polynomial drift is a Lévy-type process on the $d$-dimensional unit simplex whose generator entails a particular jump and diffusive component, as well as a polynomial drift component. In this work, the process is connected to a (Bernstein) dual process, which admits an interpretation as an urn model with coloured balls. Exploiting this duality, we identify various parameter regimes: one in which the $\Lambda$-Wright–Fisher process is absorbed almost surely at a vertex of the simplex, and others in which it admits a unique nontrivial ergodic probability measure. The results include verifiable criteria for absorption versus persistence, as well as representations of absorption and stationary measures. Moreover, the duality relation is exploited to devise and implement a simulation-based approach for estimating moments of the $\Lambda$-Wright–Fisher process.

The rapid expansion of generative AI is driving computing infrastructures toward interconnected AI data centers. Multi-band optical networks have become the backbone of such connectivity. To accommodate increasingly dynamic traffic demands, these networks require continual resource re-optimization to improve resource utilization and reduce service blocking. However, multi-band transmission amplifies inter-channel stimulated Raman scattering(ISRS), inducing nonlinear coupling among services. This coupling complicates re-optimization, as local adjustments perturb the global power profile, creating interdependent optimization decisions across services. Moreover, the nonlinear interactions vary with the subset of services selected for re-optimization, and inappropriate selection may reduce re-optimization effectiveness or diminish achievable benefits. To address these challenges, we propose a Stackelberg Game with Multi-Agent Transformer (SG-MAT) approach for ISRS-aware resource re-optimization. SG-MAT employs a leader-follower structure, in which the MAT-based follower models nonlinear coupling and generates coordinated channel reallocations, feeding the utilization improvements to the leader. Using this feedback, the leader identifies critical service subsets, approximating a Stackelberg equilibrium to guide global decisions and enabling consistent and efficient re-optimization. Simulation results show that SG-MAT achieves the highest utilization improvement and reduces the blocking ratio by approximately 30% under moderate traffic loads, with computational complexity comparable to benchmarks.

We study a conjecture of Conlon, Fox, Gunby, He, Mubayi, Suk, Verstraëte, and Yu, which characterizes the $3$-graphs $H$ for which the off-diagonal Ramsey number $r(H,n)$ is polynomial. We prove the conjecture for $3$-graphs with at most five tight components. Additionally, we show that the full conjecture is equivalent to its specialization to linear $3$-graphs. We prove quasipolynomial upper bounds for a class of `forward-doubleable' $3$-graphs, which includes the Fano plane. Finally, we give an exponential lower bound for `non-forward-laminar' $3$-graphs and use this to completely classify the Ramsey growth of a natural family of 3-graphs called chains.

Office workflows are a demanding testbed for self-improving agents because their deliverables must be correct, natively editable, and mutually consistent. A central question is whether gains on feedback tasks reflect reusable procedures rather than task-specific fixes. We introduce EvoCowork, a 240-task benchmark for evaluating harness-level recursive self-improvement across four office modalities, with model weights held fixed. The benchmark covers presentations, documents, spreadsheets, and cross-office workflows, and scores submitted artifacts for content, native structure, formulas, cross-file consistency, and visual quality where needed. Its protocol separates feedback, revision selection, and final testing: methods learn from single-modality tasks, then are evaluated on unseen single-modality tasks and cross-office workflows under a shared initial harness and explicit budgets. In static evaluations of two models and four harnesses, the best configuration scores 42.25 overall and 30.62 on cross-office tasks. Starting from the same MSA harness with GLM 5.3, ACE and AHE improve in-domain scores by 4.90 and 2.64 points, respectively, but change cross-office scores by only +0.03 and -1.72. Thus, gains on the file types that provide feedback do not necessarily transfer to coordinated workflows. EvoCowork makes this distinction explicit while recording the capability, transfer, and cost of harness evolution. Code and task packages: https://github.com/DeepThinkingZhouLiu/cowork-evolution-benchmark.

The PyLMT transient detection pipeline has been operational since November 2023 and has detected $\sim$10,000 transients and variable objects in ILMT images in near real time. The survey has enabled the discovery and characterisation of a range of bright and faint optical transients. The detected population spans a broad diversity of astrophysical phenomena, including supernovae, variable stars, active galactic nuclei (AGN) variability, and Solar System objects such as asteroids and comets. We provide a concise overview of the detection of a variety of transient and variable classes identified with the ILMT, emphasising the survey's potential contributions to a broad class of astrophysical sources. We highlight representative science cases, including supernovae, AGNs, variable stars, and asteroids, to demonstrate the scientific potential of the ILMT survey.

Recent alignment-free Text-to-Speech (TTS) models achieve high synthesis quality by removing explicit alignment modules and simplifying training. However, they often struggle in cross-lingual scenarios due to differences in character durations across languages, leading to duration mismatch and degraded speech quality. Existing approaches typically address this issue using explicit duration predictors, increasing model complexity. Systematic evaluation of TTS performance for Indian languages in cross-lingual settings remains under-explored due to the lack of standardized evaluation benchmarks. In this work, we construct a cross-lingual evaluation benchmark covering English and nine Indian languages and propose a language-aware inference-time duration estimation strategy for alignment-free TTS that incorporates language-specific duration-per character statistics without retraining or architectural modifications. We further develop heuristic formulations that refine duration estimation based on language-dependent duration variations. Experiments on bilingual and multilingual A-DMA models demonstrate improved speech naturalness, intelligibility, and speaker similarity while reducing duration mismatch.

This work combines experimental investigations and first-principles calculations to study hexagonal boron nitride (hBN) growth on a curved Pt(331) substrate. Low-energy electron diffraction (LEED) and scanning tunnelling microscopy (STM) were employed to investigate the structural properties, observing that hBN growth led to faceting with several stable facets: Pt(111), Pt(221), Pt(441), Pt(881), and Pt(110). Additionally, the electronic properties were examined by X-ray photoemission spectroscopy (XPS) and near-edge X-ray absorption fine structure (NEXAFS), revealing an increase in the hBN-substrate interaction strength, from the weakest at the Pt(111) to the strongest at the Pt(110). Density functional theory (DFT) calculations confirm that the selective faceting mechanism is induced by hBN, which changes the order of relative stability of the uncovered facets. Although a complete understanding of the process is not achievable using only thermodynamical arguments, the comparison between hBN adhesion energies and clean surface free energies clearly indicates that some facets are stabilized with respect to others by stronger overlayer-substrate interaction, whereas for others the clean surface energetics plays the dominant role.

We optimize the energy-divergence trade-off of laser-driven protons from a C-shaped target using two-dimensional particle-in-cell simulations and constrained multi-objective Bayesian optimization. Five geometry and focusing parameters are varied to maximize proton cutoff energy and minimize the RMS divergence of forward protons with E_k >= 5 MeV under a proton-only conversion-efficiency constraint of eta_H >= 9%. Of 82 designs, 60 are feasible and 15 Pareto-optimal. The Max-Energy and Min-Divergence endpoints yield 28.64 MeV at 52.51 degrees and 11.47 MeV at 11.13 degrees, respectively; a representative compromise yields 21.44 MeV at 38.44 degrees. Matched-stage diagnostics link these regimes to optical-field buildup, electron energization, charge separation, and outer-sheath organization. Numerical tests support the principal trends. Reflection timing does not support a monotonic trapping-duration law, and the charge-sheath relation remains non-local. These results establish a design-resolved route to controlling proton energy and angular quality.

Artificial intelligence is increasingly used to support product management tasks. At the same time, AI-native research and development (R&D) may increase software realization capacity. Organizations may realize software faster than they can identify, qualify, and reconsider what should be realized. This paper asks how product management should be re-envisioned under these conditions. We synthesize research on software product management, product learning, human-AI collaboration, and the Product Management and R&D boundary. We propose a vision in which AI becomes a persistent participant in product work. The vision covers sensing, exploration, decision-making, realization, and learning. We propose six principles and make the vision tangible through an operating model grounded in human professionals, AI personas, agents, skills, and a shared product context. We further argue that AI-native Product Management and AI-native R&D can form a continuous product-learning process. Finally, we identify four AI-specific research directions that require empirical investigation.

Multipath and non-line-of-sight (NLOS) propagation degrade global navigation satellite system (GNSS) positioning accuracy in urban environments. Although fisheye sky segmentation identifies obstructions along satellite directions, binary visibility does not fully characterize observation quality. This study proposes a point-supervised visual framework that predicts GNSS observation faults from fisheye images using supervision derived from GNSS observations. During offline labeling, a vision-assisted differential GNSS procedure selects multiple reference satellites based on sky visibility and elevation constraints. Double-difference pseudorange residuals automatically generate fault labels and sample importance weights at satellite projection locations. A residual-weighted loss is computed only at valid labeled points, enabling learning from sparse supervision without manual pixelwise fault annotations. The trained network generates dense fault probability maps, which are sampled at satellite projection locations to weight observations in weighted least squares positioning, reducing the influence of potentially faulty observations. The framework is evaluated using U-Net, DeepLabv3+, DINOv3, and SAM3 on urban vehicle data collected on a different day from the training data. All four models reduce horizontal root mean square error (RMSE) by 29.51% to 36.07% compared with SAM binary sky segmentation, with U-Net achieving the lowest RMSE of 0.78 m. The framework also provides positioning solutions during some intervals when the segmentation baseline fails. These results demonstrate the framework's generalization ability and show that GNSS-derived point supervision enables fisheye images to provide observation quality information beyond binary sky visibility, supporting GNSS fault detection.

Portfolio decisions are often made under uncertainty about expected returns, and investors revise their beliefs as new information arrives. Although richer information generally cannot reduce optimized value ex ante, the trading strategy it induces need not improve on the strategy induced by a less informative information structure in every fixed-drift scenario. We introduce scenario-conditional expected utility and value of information to evaluate strategies selected under the investor's current belief. In a finite-horizon Gaussian portfolio choice model, we compare a fixed-belief investor, a price observer, and an attentive investor who also processes an auxiliary signal under an attention capacity constraint. Ex ante, price learning and auxiliary attention are valuable, and the total value of attention increases with capacity. Scenario-conditionally, learning creates a correction benefit by reducing the initial trading-position gap relative to the full-information strategy, but also a fluctuation cost through belief revisions, so the ex ante ranking can reverse. We derive the current-belief distribution of scenario-conditional value and the associated underperformance probability, defined as the probability of negative scenario-conditional value relative to fixed-belief choice. Over sufficiently short horizons, the scenario-conditional value of learning is negative when belief error is small relative to posterior uncertainty. For a given horizon, sufficiently high attention capacity yields a higher scenario-conditional value than any fixed finite-capacity benchmark across all drift scenarios. We also derive sharp convergence rates to full-information performance under Shannon and Tsallis technologies.

We compute cohomology spaces of the ortosymplectic Lie superalgebra $\mathfrak{osp}(1|2)$ on the (1,1)-dimensional real superspace with coefficients in the superspace $\frak{D}_{{\lambda},\mu}$ of $n$-ary differential operators acting on weighted densities (${\lambda}\in\mathbb{R}^n$ and $\mu\in\mathbb{R}$).

We have devised an approach to model the empirical Point Spread Function (PSF) of the images taken with the 4-meter International Liquid Mirror Telescope (ILMT). The characteristics of the ILMT PSF are influenced by the nature of the Liquid Mirror, as well as by instrument specific and atmospheric factors. As the telescope uses Time Delay Integration (TDI) mode for tracking the passing objects, the PSF varies along the declination axis due to differential motion and star-trail curvature. The 5-element optical corrector homogenizes the PSF to some degree, but does not completely eliminate the disturbing effects mentioned above. The PSF in the RA direction is changing due to the sky conditions. We show that we can successfully model the PSF of 4k$\times$4k chunks using PSFEx, accounting for second-order X-Y variability. Our software robustly reproduces the PSF and its variation across the field.

Agentic AI moves generative AI applications toward systems that coordinate reasoning, enterprise context, model inference, and external actions. Integrating these capabilities into enterprise software introduces diverse architectural concerns, from orchestration, knowledge and memory, and model access to tools and actions, security, runtime infrastructure, and evaluation. Existing research and industry architectures address many of these concerns but differ in scope, abstraction, responsibility organization, interaction representation, and assumptions about which capabilities are required. This paper presents a technology-neutral reference architecture for enterprise agentic AI integration. We synthesize architectural capabilities from research publications, industry architecture guidance, and interoperability specifications by architectural purpose, organize them into ten responsibilities, and define boundary invariants, typed cross-responsibility interactions, and variation points that distinguish responsibilities from condition-dependent capabilities and replaceable realization mechanisms. We instantiate the architecture as an ESG portfolio-analysis and pre-trade compliance agent over an existing service-oriented ESG metric computation and reporting system, and complement the case with a mapping of an independently published agentic architecture. The case realizes seven of the ten responsibilities; different questions take different interaction paths, controls activate only under their conditions, and the probed boundary invariants hold. The independent architecture maps onto the proposed responsibilities without requiring an additional one. These results provide initial evidence for the architecture's organization and applicability rather than its completeness. The deployment and recorded results are available as a replication package.

State-of-the-art beat tracking is based on neural networks that model beats and downbeats as individual events, often followed by a separate Dynamic Bayesian Network (DBN) enforcing musical consistency. We propose a novel architecture that combines deep learning with differentiable theory-based modules: Specifically, we design a system that first predicts a tatum grid, then groups tatum events to beats and downbeats, explicitly modelling multiple options for the current tempo and time signature in a bar-pointer-like fashion. Our approach improves continuity metrics compared to existing DBN-free systems and is more versatile than the de-facto standard DBN, but still falls behind the best DBN-based methods on average. Given the many design choices and hyperparameters involved, we assume we have not yet explored the full potential of this idea and publish our code and invite others to explore this further.

This report studies passage retrieval and re-ranking on the MS MARCO benchmark using lightweight Qwen3-0.6B models. Across three experimental settings — BM25-then-reranking task, corpus-wide dense retrieval task, and dense retrieval then re-ranking task — the results demonstrate that competitive ranking performance can be achieved without complex architectural modifications, custom loss functions, or cross-encoder distillation. Instead, substantial gains stem from data-centric strategies: corpus-targeted continued pre-training, rigorous data curation, hard-negative mining, training–inference distribution alignment, and weight merging via Model Soup. In particular, Model Soup delivers marked performance gains while preserving the inference efficiency of a single model. These findings highlight how data curation and weight merging can push sub-billion-parameter models to state-of-the-art performance. My models can be downloaded from Hugging Face: https://huggingface.co/shuguangh/models

Industrial IoT sensors produce continuous, high-volume data streams that predictive maintenance needs to analyse in real time. Cloud-centric architectures send all raw telemetry to central servers, which costs latency, bandwidth and resilience. This paper presents a hybrid cloud-edge architecture for manufacturing predictive maintenance built on AWS IoT Core and AWS IoT Greengrass v2, with local machine learning inference at the edge and selective cloud synchronization. We deploy two configurations on real AWS infrastructure and compare them: a cloud-only baseline (Configuration A) and the proposed hybrid edge architecture (Configuration B). We measure end-to-end inference latency, bandwidth consumption and inference accuracy. Configuration B cuts cloud-bound bandwidth by 86.8% (63.6 KB to 7.2 KB per 200-reading window) and decision latency by 256x (23.1 ms cloud round trip to 0.09 ms local inference). At the edge, with no cloud involvement, it keeps an anomaly-class F1 score of 0.84 and 91% recall. The evaluation also instruments system uptime and Stream Manager queuing under cloud disconnection; the quantitative disconnection experiment is left as planned work. Our review of the literature points to four gaps that this work addresses: the lack of empirically validated Greengrass v2 deployments for predictive maintenance, unquantified bandwidth savings from selective edge synchronization, unmeasured resilience under cloud disconnection, and the open MLOps feedback loop at the edge.

The TF-cone conjecture of Asai and Iyama predicts that the TF equivalence class of an integral $g$-vector is the open cone generated by the canonical summands of all its positive multiples. We give a counterexample to this conjecture. For the tensor product of the three-arrow Kronecker algebra with the path algebra of $s\to t$, a vector $w$ has a two-dimensional TF equivalence class, whereas every positive integral multiple $mw$ is generically indecomposable. The cone generated by these canonical summands is therefore a single ray. The same example disproves the relative-interior version proposed by Haerizadeh and Yassemi.

Dengue is a viral infection caused by the dengue virus (DENV) and a major public health concern in Bangladesh, where symptom overlap with other febrile illnesses complicates early diagnosis. This study compares Logistic Regression, Random Forest and XGBoost to predict dengue from clinical and hematological data using two Bangladeshi hospital datasets (2024, n=1,511; 2025, n=1,283) with SMOTE. Extended models outperformed benchmarks within 2024 (Random Forest ROC-AUC 0.640 to 0.661), with RDW-CV the most consistently influential added predictor. 2025 benchmark performance was higher (ROC-AUC 0.904), unexplained by class imbalance and possibly reflecting data quality. External validation is needed.

Three-dimensional (3D) velocity model building (VMB) remains computationally challenging with conventional approaches and native 3D neural networks, largely due to the prohibitive memory and training cost of full 3D formulations. Therefore, we introduce a depth-progressive framework that achieves 3D VMB using 2D generative diffusion models through a depth-as-channel formulation, in which the depth dimension of each 3D velocity patch is mapped to the channel dimension of a 2D U-Net. This allows the model to synthesize overlapping depth patches sequentially from shallow to deep while conditioning the prediction of the deeper patch on the shallower one, as well as on any available geophysical constraints, including well logs and structural attributes. During training, paired shallow-deep patch pairs with randomized overlaps are used to teach the network how deeper velocity structures relate to shallower information, and a classifier-free guidance strategy enables the model to handle missing constraints gracefully. At inference, starting from a shallow velocity prior, the model recursively generates deeper patches in a top-down manner, and a depth-attenuated Gaussian blending scheme merges overlapping predictions into spatially coherent 3D volumes while accounting for the progressive accumulation of uncertainty at greater depths. Numerical experiments on both in-distribution benchmark models, including Overthrust, SEAM Arid, and SEG/EAGE, and the out-of-distribution Marmousi model demonstrate that the method effectively propagates shallow velocity priors downward and produces high-quality 3D velocity volumes across a range of geological settings and conditioning scenarios. Compared to a 3D benchmark of equivalent architecture, the proposed method significantly reduces parameter count, training cost, and inference cost, while maintaining superior VMB accuracy.

We determine the best constants in estimates of the $L^\infty$ norm of functions on the sphere with bounded Laplacian whose first spherical harmonic components vanish. In every dimension $d\geq2$, these constants are expressed through weighted $L^1$ extremal polynomials. On $S^2$ the constant is an explicit positive series, and the best coefficient of $n^{-2}$ is $7\zeta(3)/4$. On $S^3$ the best coefficient of $n^{-2}$ is a positive integral involving the Hörmander–Bernhardsson extremal function. In higher dimensions the rescaled extremal polynomials converge to weighted entire extremals, which give asymptotic formulas for the Bohr constants on $S^4$ and $S^6$. The most unexpected and, in our view, interesting observation is that, for every fixed $n\geq2$ on $S^2$ and $S^3$, the constants in best approximation estimates are strictly smaller than the corresponding Bohr–Favard constants.

Enterprise AI agents can execute altered versions of approved actions, concurrently reuse the same permission, or retry operations whose responses were lost. These failures arise when approval and resource execution lack a shared contract. KedgeFlow specifies language-independent contracts for trusted controllers to authorize state-changing operations and privileged gateways to enforce that authority independently of agent reasoning. Inspired by OpenFlow, it connects policy checks supported by evidence, signed permission for an exact action, a shared allowance that can be consumed only once, target-side conditions for applying the change, and retained outcome records. Under trusted issuance of exact-action authority, two conditional properties establish at most one successful claim per allowance and link every governed dispatch to a matching grant, claim, and subsequent dispatch authorization, even when the agent is compromised. A reference implementation exposes this execution path to agent systems through the Model Context Protocol (MCP). In the single-resource evaluation, 6,560 concurrent ledger claim attempts yielded exactly one successful claim for each of 82 allowances. Boundary tests rejected unauthorized and altered requests before consumption. A 32-call gateway race produced one committed mock effect, and a retry after a lost response produced no additional dispatch. Deliberately permissive policy nevertheless authorized an undesirable effect, exposing the limits of execution authorization. Evaluation uses three members of the etcd coordinated key-value store on one host, synthetic authentication and authorization inputs, and a mock target that rejects stale requests. Distributed reservations, enforcement of every production access route, native crash recovery, exactly-once effects at real targets, and interoperability between independent implementations remain untested.

Influence functions provide an efficient alternative to retraining for assessing the effects of removing or reweighting training data. For groups, however, the standard first-order estimate reduces individual contributions to a sum that can conceal substantial differences among members, including opposing effects that cancel. We establish a Bregman centroid formulation of the linearized group influence objective, interpreting this aggregate as a single representative and using Bregman information (BI) to quantify the variation it omits. Extending this formulation to multiple clusters reveals distinct contributions whose sum recovers the original estimate. We further propose an estimator that jointly optimizes the weights of these cluster contributions, rather than simply summing them. Experiments uncover cluster-specific influence patterns and a positive association between target-specific BI and approximation error. The proposed estimator achieves higher $R^2$ than standard group influence in predicting target changes under the nonlinear proximal Bregman response function. Simplified clustering and correction variants reduce computational cost. The centroid view thus explains what additive group influence leaves out and provides a basis for using that information to improve estimation.

Agentic approaches for automated test generation can reach high-quality results, but they improve a given test suite incrementally, so a weak starting point may lead to many iterations or insufficient tests. To address this, we present Testuary, which provides the agentic phases with an improved starting point. Testuary first generates multiple test suites through independent, parallel zero-shot requests, checks and repairs them deterministically, and combines them into a single suite with a novel structural merging approach. A runtime-weighted set-cover reduction removes redundant tests without losing covered lines, branches, or killed mutants. After that, two agentic phases target the remaining gaps in coverage and mutation score. We compare Testuary with established search and LLM-based test generation approaches. The non-agentic first phase alone already significantly outperforms the current state-of-the-art. After the agentic phases, Testuary achieves on average more than 99.9% line and branch coverage as well as a mutation score of 98.6%, significantly outperforming all compared baseline approaches. Testuary needs fewer model requests and its test suites are on average smaller and faster than those of the compared agentic approach.

Evaluating autonomous agents requires examining whether they can translate goals into coherent action across extended interactions. In persistent environments, this requires coordinating interdependent actions under incomplete observations, accounting for delayed effects, and adapting as earlier decisions shape subsequent opportunities. We introduce ASIWorld, a survival sandbox that combines persistent world and character state with character-scoped observations and a programmable asynchronous interface to study sustained goal execution. Agents pursue interdependent production and exploration tasks through direct actions or agent-authored programs that continue running between model calls. The evaluation assesses task success, the attainment of development achievements, and survival costs, complemented by an analysis of execution behavior. In an exploratory study of nine agent configurations using a common harness, eight complete the production task and six complete both tasks. Analyses of execution trajectories and performance across difficulty levels reveal different preparation, combat, and recovery behaviors among configurations with similar task outcomes. Incomplete preparation is associated with stalled progression, while configurations that complete the tasks differ in survival costs and the fulfillment of additional achievement criteria. These findings illustrate the value of assessing sustained goal execution through both task outcomes and the behaviors and costs associated with achieving them.

We investigate the gravitational waves generated by first-order phase transitions (PTs) in an expanding Universe. Using a scalar field-fluid coupled model, we perform three-dimensional lattice numerical simulations of slow first-order PTs with strong strength. We observe that the contribution from sound waves to the gravitational wave (GW) signal significantly dominates over that from bubble collisions. The sound wave spectra exhibit two peaks located at the mean bubble separation and the sound-shell width. Rotational modes are generated in the fluid during the bubble collisions, and their magnitude grows with the PT strength. The lifetime of the sound wave source increases with decreasing $\alpha$: for weaker transitions the source can persist for longer than a Hubble time, while for stronger transitions it decays within a Hubble time. When the walls propagate as deflagrations, the GW spectral shape changes drastically, with steeper ultraviolet tails and a shorter-lived sound wave source. We also observe that the magnitude of the normalized GWs is roughly an order of magnitude smaller than obtained in the Minkowski spacetime simulations.

Natural habitats confront microorganisms with heterogeneous, mechanically complex surroundings that shape how they move and explore. Most studies of motility in complex media treat the environment as either a homogeneous fluid or a rigid obstacle field. Here we instead study a medium that is simultaneously soft and reconfigurable — deformable by the organism moving through it. We track the unicellular biflagellate microalga Chlamydomonas reinhardtii swimming in a soft, disordered suspension of microgel particles comparable in size to the cell itself, across a range of gel packing fractions. With increasing packing fraction, trajectories change from long straight runs to short, frequently reorienting ones, and the effective translational diffusivity decreases monotonically. The bimodal run/tumble speed distribution, characteristic of the free-swimming gait, collapses toward a single mode as the packing fraction increases, by near-proportional slowing of both gait states rather than a redistribution of cells between them. Recasting trajectories as alternating hops and traps reveals the signature of a deformable medium: hop lengths halve as the gel densifies, approaching the microgel aggregate scale, while trapping times do not increase with increasing packing fraction. Independent force estimates place the cell's propulsive force close to an order of magnitude above that needed to deform a single grain, so remodeling is expected rather than merely observed. A minimal active remodeler model — an active Brownian particle among mobile, rigid obstacles — shows that obstacle mobility alone is sufficient to relocate the onset of permanent trapping. Thus, our results establish obstacle mobility as a control parameter for motility in soft, disordered matter.