arXiv:2512.14888v2 Announce Type: replace-cross Abstract: We develop a probabilistic algorithm of Kronecker type for computing a Kronecker representation of a zero-dimensional linear section of an algebraic variety $V$ defined over a perfect field $k$. The variety $V$ is the Zariski closure of the set of common zeros $\{F_1=0,\ldots,F_r=0,G\not=0\}$ of multivariate polynomials $F_1,\ldots,F_r\in k[X_1,\ldots,X_n]$ outside a prescribed hypersurface $\{G=0\}$. We assume that $F_1,\ldots,F_r$ satisfy natural geometric conditions, such as regularity and radicality, in the local ring $k[X_1,\ldots,X_n]_G$. Our approach combines homotopic deformation techniques with symbolic Newton-Hensel lifting and elimination. We discuss the concept of lifting curves as intermediate geometric objects that enable efficient computation. The complexity of the algorithm is expressed in terms of the degrees and arithmetic size of the input and achieves soft-quadratic complexity in these parameters. We provide detailed complexity analyses for arbitrary perfect fields, as well as for two important cases in computer algebra: finite fields and the field of rational numbers. For each case, we obtain sharp bounds on the size of the base field or required primes.
Science Journals
arXiv:2602.02431v2 Announce Type: replace-cross Abstract: It is folklore that reusing training data more than once can improve the statistical efficiency of gradient-based learning. While this phenomenon has been extensively studied in linear regression, the benefit of multi-pass gradient descent (GD, which reuses all the data) over one-pass stochastic gradient descent (online SGD, which uses each data point only once) is not well-understood in nonlinear and non-convex settings, except for a loss modification mechanism achieved by the first two passes on the data. In this work, we consider learning a $d$-dimensional single-index model with a quadratic activation, for which it is known that one-pass SGD requires $n\gtrsim d\log d$ samples to achieve weak recovery. We first show that this $\log d$ factor in the sample complexity persists for full-batch spherical GD on the correlation loss; however, by simply truncating the activation, full-batch GD exhibits a favorable optimization landscape at $n \simeq d$ samples, thereby outperforming one-pass SGD (with the same activation) in statistical efficiency. We complement this result with a trajectory analysis of full-batch GD on the squared loss from small initialization, showing that $n \gtrsim d$ samples and $T \gtrsim\log d$ gradient steps suffice to achieve strong (exact) recovery.
arXiv:2603.07100v2 Announce Type: replace-cross Abstract: The local viscoelastic (VE) environment governs the motion of an embedded microsphere and consequently, pertinent dynamical phenomena. However, studying such phenomena with varying VE properties remains challenging for various reasons, including the strong coupling among the VE parameters and their dependence on experimental conditions, such as temperature. Here, we demonstrate the experimental realization of configurable VE media with broad variations, wherein the VE properties can be systematically and independently tuned, employing a dynamic optical trap. Specifically, the dynamics of a particle in a slowly diffusing optical trap provides the linear microrheological response of single-relaxation VE fluids, namely Jeffreys or Maxwell-Voigt (MV) fluids, where the trap strength and its diffusion coefficient regulate the elastic response and the low-frequency viscosity, respectively. The characteristic features in the mean square displacement (MSD) of the trapped particle match those of a probe particle in real MV fluids, and the simulation results following harmonically bound Brownian particle with long-time diffusion model describing single-relaxation complex fluids. Our scheme is further validated by demonstrating excellent quantitative agreement between the experimentally observed MSDs of the trapped bead and those from the corresponding analytical predictions. We extend the applicability of this scheme for realizing the microrheological response of double-relaxation VE media by incorporating appropriately correlated noise in the trap trajectory, signifying its validity for any linear VE media with multiple relaxations. Our scheme can be further extended to realize probe particle dynamics in an active VE environment, e.g., an entangled network of active polymers, by translating the trap along an active Brownian trajectory.
arXiv:2603.08977v2 Announce Type: replace-cross Abstract: We propose Universal Speech Content Factorization (USCF), a simple and invertible linear method for extracting a low-rank speech representation in which speaker timbre is suppressed while phonetic content is preserved. USCF extends Speech Content Factorization, a closed-set voice conversion (VC) method, to an open-set setting by learning a universal speech-to-content mapping via least-squares optimization and deriving speaker-specific transformations from only a few seconds of target speech. We show through embedding analysis that USCF effectively removes speaker-dependent variation. As a zero-shot VC system, USCF achieves competitive intelligibility, naturalness, and speaker similarity compared to methods that require substantially more target-speaker data or additional neural training. Finally, we demonstrate that as a training-efficient timbre-disentangled speech feature, USCF features can serve as the acoustic representation for training timbre-prompted text-to-speech models. Speech samples and code are publicly available.
arXiv:2606.07435v2 Announce Type: replace Abstract: Visual speech recognition (VSR) models now surpass human lipreaders on benchmarks, but do such gains establish human-like visual speech perception? To explore this, we compare three VSR systems with human baselines on the MaFI word-level lipreading dataset using word, character, phoneme, and viseme-level metrics. Although models achieve higher overall accuracy, they succeed and fail on different words than humans. A text-only n-gram baseline given only a few initial phonemes rivals human lipreading. VSR word-level errors are consistently better explained by training word frequency than by the visual informativeness of words. Viseme accuracies, confusion matrices and human-model correlations further show that models gain most on visemes humans find hardest, and show much weaker dependence on visual clarity. Our work demonstrates that VSR systems rely primarily on language cues from training data rather than visual perception, failing to bind visual features into meaningful words.
arXiv:2401.02878v5 Announce Type: replace-cross Abstract: Inspired by the stochastic particle method, this paper establishes an easily implementable explicit numerical method for McKean-Vlasov stochastic differential equations (MV-SDEs) with superlinear growth coefficients. The paper establishes the theory on the propagation of chaos in the $L^{q}$ sense. The optimal {uniform-in-time} strong convergence rate $1/2$-order of the numerical solutions is obtained for the interacting particle system. Furthermore, it is proved that the numerical solutions capture the long-term dynamical behaviors of MV-SDEs precisely, including moment boundedness, stability, and ergodicity. Moreover, a unique numerical invariant probability measure is yielded, which converges to the underlying invariant probability measure of MV-SDEs in the $L^2$-Wasserstein distance. Finally, several numerical experiments are carried out to illustrate the main results.
arXiv:2604.09787v2 Announce Type: replace-cross Abstract: Data collected from the physical world is always a combination of multiple sources: an underlying signal from the physical process of interest and a signal from measurement-dependent artifacts from the sensor or instrument. This secondary signal acts as a confounding factor, limiting our ability to extract information about the physics underlying the phenomena we observe. Furthermore, it complicates the combination of observations in heterogeneous or multi-instrument settings. We propose a deep learning framework that leverages overlapping observations, a dual-encoder architecture, and a counterfactual generation objective to disentangle these factors of variation. The resulting representations explicitly separate intrinsic signals from sensor-specific distortions and noise, and can be used for counterfactual view generation, parameter inference unconfounded by measurement distortions, and instrument-independent similarity search. We demonstrate the effectiveness of our approach on astrophysical galaxy images from the DESI Legacy Imaging Survey (Legacy) and the Hyper Suprime-Cam (HSC) Survey as a representative multi-instrument setting. This framework provides a general recipe for scientific and multi-modal self-supervised pretraining: construct training pairs from overlapping observations of the same physical system, treat sensor- or modality-specific effects as augmentations, and learn invariant representations through counterfactual generation.
arXiv:2606.08676v1 Announce Type: new Abstract: AI coding assistants have significantly improved developer productivity by automatically suggesting code that aligns with user intent, and many of these tools are now integrated directly into Integrated Development Environments (IDEs). Developers interact with code in two distinct cognitive modes: Flow and Command. While developers require tools that directly complete or infill code in unfinished programs during Flow mode, they also need tools that can comprehend intentions expressed as natural-language instructions and convert them into executable code in Command mode. Although instruction-tuned Large Language Models (LLMs) dominate many application scenarios due to their abilities to infer and fulfill developers' intents, it remains unclear whether the same paradigm is equally suitable for different code-related tasks. Therefore, it is necessary to understand how instruction tuning affects the feasibility of CodeLLMs as coding assistants. To fill this gap, we conduct the first empirical study that uncovers a key trade-off caused by instruction tuning across programming modes, which we term the Instruction-Tuning Tax. Our results show that instruction tuning is not a free lunch: although instruction-tuned models are more capable of following instructions and leveraging structured guidance, these gains often come at the cost of weaker infilling performance. We further extend our study through both qualitative and quantitative analyses, including manual failure categorization, behavioral metrics that capture generation fidelity, and intermediate-checkpoint evaluation throughout the tuning process. Summarizing our results into seven findings and four implications, our study offers a new perspective on the development of AI-powered coding tools and highlights the need to carefully balance instruction-following ability with effective code generation assistance.
arXiv:2606.08826v1 Announce Type: new Abstract: Image data regarding galactic morphology is expected to increase both in quantity and quality for the next foreseeable years; thus it is important to explore which deep learning architectures adapted for image classification tasks are cost-effective. Residual and Inception networks are ideal for exploring classification convolutional neural networks (CNNs) due to their computational efficiency, achieved through techniques such as residual connections and parallelized inception modules, enabling deeper networks without excessively increasing computational complexity. In this work, we analyze the performance of ResNet101 and InceptionV4 on a spatially-augmented Galaxy10 DECals dataset. Retaining the ten-class classification of galaxies, we modify the image count of each class. We find that ResNet101 and InceptionV4 models achieved accuracies of $\sim$ 90%, comparable with reported performance in the literature. In terms of performance metrics, ResNet101 is superior to InceptionV4. Our results indicate that either of these CNN architectures could serve as a robust foundation for specialized pipelines for classification of galaxy images from upcoming surveys.
arXiv:2606.09030v1 Announce Type: new Abstract: Clinical early warning systems built on electronic health records, in which clinical observations are recorded as irregularly sampled medical time series (ISMTS), must deliver both calibrated risk scores for patient triage and interpretable rationales that clinicians can verify. Large Language Models (LLMs) have been explored for this task, yet they collapse graded clinical risk into overconfident binary predictions. This risk polarization undermines both calibration and cross-patient comparability. To address this, we propose TRIAGE, a framework that trains an LLM to generate dialectical reasoning over competing clinical outcomes by eliciting outcome-specific rationales. This dialectical formulation mitigates risk polarization, enabling a single LLM to yield continuous risk scores grounded in explicit clinical reasoning. Evaluated on three ISMTS benchmarks, TRIAGE achieves an average AUPRC improvement of 3.3% and reduces calibration error by 81% compared to the competitive baselines. An LLM-as-a-judge assessment further shows that our rationales surpass post-hoc explanations from the baseline by 20% in clinical reasoning quality. The source code is available at https://github.com/HyeongWon-Jang/TRIAGE .
arXiv:2606.09134v1 Announce Type: new Abstract: Constructing knowledge graphs from 3D simulation scenes is essential for robot task reasoning, but the key bottleneck, grounding scene objects to formal ontology classes, still relies on manually curated dictionaries that are brittle and do not generalize across assets. We investigate whether large language models (LLMs) can automate this grounding step for Universal Scene Description (USD) scenes as a zero-shot, training-free alternative. On a kitchen scene (125 objects) with SOMA-HOME Ontology, LLMs achieve 90-96% exact-match accuracy with descriptive names and 49-89% with abbreviated names, substantially outperforming dictionary and embedding baselines. Under fully opaque names, context-augmented prompting recovers up to 48%. Feature ablation reveals that LLMs primarily exploit semantic cues in the scene graph (sibling names and parent paths); anonymizing these cues reduces accuracy to 0-6%, while geometry alone yields only 4-17%.
arXiv:2606.09156v1 Announce Type: new Abstract: Autoregressive (AR) models have demonstrated strong potential in visual generation, offering superior performance with simple architectures and optimization objectives. However, existing methods are typically limited to single-modality conditions, e.g., text, restricting their applicability in real-world scenarios that demand image synthesis from diverse controls. In this work, we present OmniGen-AR, a unified autoregressive framework for Any-to-Image generation. By discretizing various visual conditions through a shared visual tokenizer and text prompts with a text tokenizer, OmniGen-AR supports a broad spectrum of conditional inputs within a single model, including text (text-to-image generation), spatial signals (segmentation-to-image and depth-to-image), and visual context (image editing, frame prediction, and text-to-video generation). To mitigate the risk of information leakage from condition tokens to content tokens, we introduce Disentangled Causal Attention (DCA), which separates the full-sequence causal mask into condition causal attention and content causal attention. It serves as a training-time regularizer without affecting the standard next-token prediction during inference. With this design, OmniGen-AR achieves new state-of-the-art or at least competitive results across a range of benchmark, e.g., 0.63 on GenEval and 80.02 on VBench, demonstrating its effectiveness in flexible and high-fidelity visual generation.
arXiv:2606.09162v1 Announce Type: new Abstract: Video semantic segmentation for low-altitude UAVs requires temporal consistency, yet dense optical flow introduces spatially structured noise in the planar regions that dominate aerial imagery. We propose a zero-parameter geometric gate that uses RANSAC homography inlier ratios on a $16\times16$ spatial grid to route each region to either homography or optical flow warp before fusion via Semantic Similarity Propagation. The gate requires no learned parameters -- only a median-threshold binary decision on RANSAC statistics -- adding only 211K trainable parameters (the SSP fusion layer) to a frozen backbone. On synthetic UAVid, the method achieves +4.24--4.91\% mIoU improvement over base models across two architectures (SegFormer-b2 and Hiera-S+UPerNet). Mechanism diagnostics reveal that flow residuals in planar regions are spatially autocorrelated (Moran's I = 0.32, $p < 0.001$), predict boundary instability (Spearman $\rho = 0.66$), and that rigidification recovers temporal consistency from 62\% to 92\% (+29.5pp) in homography-valid regions.
arXiv:2606.09173v1 Announce Type: new Abstract: We establish sharp, uniform-in-time strong error estimates for a nonlinearity-explicit tamed finite element method (FEM) applied to a class of superlinear stochastic partial differential equations (SPDEs) driven by multiplicative noise, including the stochastic Allen--Cahn equation with a moderately thick interface. This tamed-FEM was first introduced in [Z. Liu and J. Shen, arXiv:2502.19117] to ensure long-time unconditional stability and to preserve the Lyapunov structure of this class of SPDEs. We further prove that the scheme is exponentially ergodic and derive the convergence rate between the exact invariant measure and its numerical counterpart in the Wasserstein-2 distance. Finally, we present numerical experiments that verify the ergodicity as well as the sharpness and time-independence of the strong convergence rates for this tamed-FEM.
arXiv:2606.09175v1 Announce Type: new Abstract: Recently, mobile edge computing (MEC)-enabled collaborative deep neural network (DNN) inference has emerged as a promising approach for delivering intelligent services to resource-constrained mobile devices. A representative scenario is multi-user collaborative edge inference, where distinct devices independently partition their DNN models and offload backend computation to a common edge server over wireless networks. However, determining the optimal DNN partition for each device is challenging due to unknown and time-varying system conditions, including fluctuating wireless links and diverse device capabilities. To address this problem, we propose Cooperative Autodidactic NeuroSurgeon (CANS), a collaborative edge inference framework that enables devices to adaptively learn optimal DNN partitions by sharing informative feedback during online inference. To handle the challenge of device heterogeneity and better leverage offline inference experience, we integrate a novel FedLinUCB-DW algorithm that groups devices of the same type and warm-starts online exploration using local offline early-exit inference experience. Furthermore, we provide theoretical guarantees for FedLinUCB-DW by deriving the regret upper bound. We also validate our method on both a simulated environment and a hardware prototype system. Empirical evaluations demonstrate that CANS achieves lower inference latency compared to state-of-the-art baselines. Especially, in prototype experiments on two edge devices, the proposed CANS reduced average inference latency by up to 50% compared to the non-cooperative baseline.
arXiv:2606.09195v1 Announce Type: new Abstract: Understanding and reasoning over abstract visual content remains a challenge for current multi-modal large language models (MLLMs). In this paper, we explore a novel abstract data type termed complex visual query (CVQ), designed to probe symbolic and abstractive reasoning, which is a critical yet underexplored dimension of human-like neuro-symbolic reasoning for MLLMs. We present a comprehensive investigation from three perspectives: \textbf{Data $\times$ Paradigm $\times$ Exploration}. Specifically, we propose a scalable pipeline for synthesizing CVQs grounded in large-scale multi-modal knowledge graphs, generating a diverse dataset encompassing 14 distinct query types via systematic combinations of first-order logic operators. We further introduce a two-stage training framework that progressively equips MLLMs with robust visual reasoning capabilities. We conduct extensive experiments to rigorously evaluate MLLMs across multiple dimensions, including reasoning performance on CVQs, as well as cross-task and cross-scenario generalization. We believe our work opens new perspectives and avenues for advancing the reasoning frontiers of MLLMs.
arXiv:2606.09242v1 Announce Type: new Abstract: The rapid diffusion of generative artificial intelligence (genAI) systems reshapes how individuals engage with information systems, requiring users to monitor, assess, and adapt their interaction with non-deterministic systems. Existing constructs capture elements of this engagement but do not account for the situated dynamics of the entire evaluative process in genAI use. This research-in-progress, situated in a larger endeavour towards a scale development, derives an initial conceptualisation of reflective use: a behavioural-knowledge capability that unfolds across pre-use, in-use, and post-use phases, reinforced through situated reflective knowledge gained in practice. Drawing on expert interviews and a focus group, we identify four core components of reflective use and show how they form an iterative capability cycle anchored within the motivational needs outlined in self-determination theory. Understanding reflective use is essential to ensure appropriate reliance and high decision quality, and thus provides a foundation for promoting responsible and effective human-AI interaction.
arXiv:2606.09303v1 Announce Type: new Abstract: The rapid development of pretrained foundation models has enabled more general image segmentation. Multimodal large language models (MLLMs) have been widely explored for image segmentation with complex queries that require high-level reasoning. Despite promising progress, existing methods are often constrained by limited training data and the gap between MLLMs and mask generation modules. To better transfer MLLMs' perception and reasoning ability to complex reasoning-based segmentation tasks, we propose a two-stage framework Rea2Seg for mask generation and selection. Specifically, the framework first identifies potential regions as candidate masks based on the attention maps of a segmentation MLLM. It then employs an MLLM to reason over the question and candidate masks and assign scores to each mask. The final segmentation result is obtained by reranking the candidates and selecting the highest-scoring mask, reformulating image segmentation as candidate discovery followed by discriminative mask selection. We also notice that a large portion of questions in existing benchmarks focus on commonsense reasoning, and these questions usually do not fully require joint visual observation and reasoning. To address this issue, we introduce a new benchmark called ReasonSeg-SGDR that comprehensively evaluates a model's perception, grounding, and reasoning abilities across multiple dimensions, including discriminative recognition, spatial reasoning, geometric reasoning, and multi-step reasoning, with fine-grained mask generation. In addition, we collect training data to enhance MLLMs' ability to jointly understand multimodal queries and candidate masks, and to assign scores through reasoning. Experimental results on the proposed benchmark and ReasonSeg demonstrate the effectiveness of the unified mask generation and selection framework.
arXiv:2606.09338v1 Announce Type: new Abstract: Large Language Models fail at implicit multi-hop reasoning: a model answers "When was $X$ born?" and "Who is $Y$'s closest friend?" correctly but fails on "When was $Y$'s closest friend born?" in a single forward pass, even when both facts are perfectly memorized and individually retrievable. We study this failure in a controlled natural language setting with a strict separation between individuals exposed to compositional contexts during pretraining and those that never appear in any such context. We confirm that compositional failure persists even at 97% 1-hop accuracy, establishing the gap as a pretraining failure rather than a knowledge absence. We propose and test nine data-centric augmentation formats and find that compositional pretraining transfers to unseen questions for exposed individuals, but never to individuals absent from compositional pretraining, suggesting that exposure to compositional contexts during pretraining is a necessary condition for implicit multi-hop reasoning.
arXiv:2606.09389v1 Announce Type: new Abstract: As large language models (LLMs) are increasingly applied to real-world legal tasks, evaluating the reliability of their open-ended legal responses has become essential. These tasks require context-sensitive answers and allow little room for error, motivating fine-grained and diagnostic evaluation that can identify specific sources of response quality failures. We introduce LexRubric, a rubric-based benchmark for evaluating open-ended Chinese legal tasks. LexRubric contains 649 instances from legal consultation and judicial examination, which reflect both everyday legal needs and professional legal reasoning and cover 14 legal scenarios. It further includes 12,337 expert-written atomic scoring criteria organized under a unified six-dimensional framework, enabling accurate evaluation and diagnostic analysis across tasks and evaluation dimensions. To validate the reliability of the evaluation, we test multiple judge models and compare model-based judgments with human judgments. We further evaluate 18 recent general and legal-domain LLMs on LexRubric. Results show that different models exhibit distinct capability profiles, and that open-ended legal question remains challenging for current LLMs. Data is available at: https://github.com/foggpoy/LexRubric.
arXiv:2606.09470v1 Announce Type: new Abstract: Automated L2 speech assessment can assign proficiency labels, but often lacks interpretability. We propose a rubric-guided SpeechLLM for multi-aspect, multi-granular assessment, trained with a hybrid objective combining supervised fine-tuning and Bounded Direct Preference Optimization. The model jointly predicts ordinal labels at the sentence-level (accuracy, fluency, prosody), word/phoneme-level accuracy, and generates a natural-language rationale in the same response. On SpeechOcean762, our approach matches or outperforms single-granularity models while remaining competitive with prior approaches. We analyze rationale reliability along two axes: self-consistency with model predictions and alignment with ground-truth labels, using sentiment consistency (plausibility) and mention-based agreement (faithfulness). Rationales are plausible at the sentence level, but faithfulness degrades at the word/phoneme level: references are sparse and weakly aligned with token-level labels.
arXiv:2606.09509v1 Announce Type: new Abstract: Local search is a well-known heuristic method used in optimization. In this thesis, we explore its capabilities on the vertex coloring problem, an $NP$-hard problem with relevance in both theoretical analysis and practical application. To recognize limitations in the applicability of local search of the vertex coloring problem, we analyze local search landscapes on differently-structured bipartite graphs. We identify structures that ensure only global optima can exist as well as ones that enable the existence of non-global local optima, showing that on general bipartite graphs, it is possible for local search to return arbitrarily bad results. Further, we analyze the capabilities of local search on graphs where a local optimum can be found. To do so, we introduce a gray-box local search mutation operator that removes less frequent colors with higher probability and prove that it finds an optimal coloring on complete bipartite graphs in an expected run time of $\Theta(n \log n)$. This is a drastic improvement to the exponential tun time of the black-box Random Local Search, showing that gray-box mutation operators can improve the run time of local search.
arXiv:2606.09514v1 Announce Type: new Abstract: Large language models (LLMs) incur high inference cost due to their depth and parameter scale. Depth pruning can reduce latency by skipping redundant Transformer blocks, but existing methods (i) provide limited control under user-specific compute budgets and (ii) typically fix the routing path, failing to adapt as the context grows during decoding. We propose Buddy, a budget-driven dynamic depth routing framework. Buddy uses a lightweight Decision Module to score intermediate layers conditioned on the input and deterministically executes the top-k layers to satisfy a given budget. To support decode-time adaptation, Buddy reuses the first-layer KV cache as a low-overhead global context source and pools it together with the newest token representation before each routing decision. When no explicit budget is provided, an optional Budget Predictor estimates an input-dependent compute level to balance quality and efficiency. Experiments on Llama-family and Qwen models show that Buddy is competitive with strong static pruning baselines and often improves the accuracy-compute trade-off, while uniquely supporting strict budget control, decode-time rerouting, and multiple budgets within a single trained model.
arXiv:2606.09589v1 Announce Type: new Abstract: AI minidramas (also known as fruit dramas) are short, algorithmically distributed generative AI video series featuring anthropomorphized characters that have recently emerged as a widespread phenomenon on social media platforms. This paper argues that despite their seemingly innocuous aesthetic, these videos reproduce deeply gendered narrative structures in which female characters are systematically associated with moral transgression, sexual betrayal, and reproductive capacity, and that several plots also encode the logic of racialization, i.e., the process by which visible bodily difference is morally loaded. Drawing on feminist film theory, critical race theory, and platform studies, it further argues that the generative AI aesthetic of these videos, characterized by softness, roundness, and visual cuteness, functions as a mechanism of aesthetic laundering, neutralizing the ideological weight of these narratives and enabling their circulation despite content moderation systems. This paper approaches these questions through personal observation and close reading, reflecting on the specific affordances of generative AI that make this phenomenon both possible and culturally consequential for the field of computational creativity.
arXiv:2606.09610v1 Announce Type: new Abstract: Cooperative object transportation is essential in numerous domains, including industrial to domestic services. A popular transportation strategy is to carry objects on top of multi-robot systems. The corresponding task is typically solved by decomposing it into three interconnected subproblems: formation control, cooperative navigation, and collision avoidance. A particular challenge posed by real-world objects is their potentially arbitrary shape and non-uniform mass distribution, necessitating robot formations that securely support the object. In this work, we address the challenge of pattern formation control for transporting such real-world objects by proposing a novel multi-agent reinforcement learning approach. Our approach enables a multi-robot system to autonomously position itself underneath an object to support its weight while avoiding obstacles during the formation process. Our evaluations with diverse environments and varying numbers of robots show that our approach leads to policies that reliably produce balanced formations and generalize to cluttered scenes and objects with complex geometry and non-uniform mass distribution.