Existing text-to-SQL benchmarks have largely been constructed from public databases with well-structured schemas and simplistic question-SQL pairs. While large language models (LLMs) excel on these settings, their efficacy in complex private enterprise environments, characterized by intricate schemas, domain knowledge, and analytical user queries involving sophisticated structures and functions, remains unproven. To bridge this gap, we introduce BEAVER, the first text-to-SQL benchmark derived from private data warehouses. It comprises 9128 question-SQL pairs sourced from real-world query logs and 812 tables across 19 diverse domains. Building this benchmark is challenging because (1) enterprise query logs are scarce due to privacy constraints, and (2) existing all-or-nothing evaluation metrics based on accuracy make error diagnosis difficult -- especially when producing a correct query involves solving multiple compounded challenges, such as domain knowledge and query complexity. We address these issues at two levels. At the dataset level, we synthesize high-fidelity, expert-verified queries that increase dataset size and isolate individual challenges or combine them, producing queries focused on domain knowledge, query complexity, and both. At the evaluation level, we provide human annotations and evaluation metrics for five critical subtasks to enable fine-grained analysis. Our evaluation reveals a significant performance gap compared to existing benchmarks: SOTA agentic frameworks using the advanced model GPT-5.2 achieve only 10.8% accuracy. When provided with all subtask annotations as oracle hints, accuracy increases to 30.1%, confirming that a major bottleneck lies in correctly resolving these subtasks. Finally, we provide a taxonomy of the residual errors that persist even with subtask hints, identifying specific challenges such as the use of advanced functions.

Low-Rank Adaptation (LoRA) is the dominant parameter-efficient fine-tuning method due to its favorable compute-performance trade-off, yet it suffers from catastrophic forgetting. We study forgetting through a tractable _mean-field self-attention_ toy model, where tokens evolve as an interacting particle system and LoRA acts as a low-rank perturbation. Using tools from partial differential equations and dynamical systems, we characterize regimes suggesting a phase transition between forgetting and non-forgetting behavior. We show that one phase transition appears with respect to the norm of the perturbation, and the other with respect to the depth of the Transformers. We further bound the time-to-deviation in terms of the perturbation size and spectral quantities, and corroborate the predicted trends with experiments and exploratory analyses on real models under LoRA fine-tuning.

For real-world applications, autonomous mobile robotic platforms must be capable of navigating safely in a multitude of different and dynamic environments with accurate and robust localization being a key prerequisite. To support further research in this domain, we present the INSANE data sets - a collection of versatile Micro Aerial Vehicle (MAV) data sets for cross-environment localization. The data sets provide various scenarios with multiple stages of difficulty for localization methods. These scenarios range from trajectories in the controlled environment of an indoor motion capture facility, to experiments where the vehicle performs an outdoor maneuver and transitions into a building, requiring changes of sensor modalities, up to purely outdoor flight maneuvers in a challenging Mars analog environment to simulate scenarios which current and future Mars helicopters would need to perform. The presented work aims to provide data that reflects real-world scenarios and sensor effects. The extensive sensor suite includes various sensor categories, including multiple Inertial Measurement Units (IMUs) and cameras. Sensor data is made available as raw measurements and each data set provides highly accurate ground truth, including the outdoor experiments where a dual Real-Time Kinematic (RTK) Global Navigation Satellite System (GNSS) setup provides sub-degree and centimeter accuracy (1-sigma). The sensor suite also includes a dedicated high-rate IMU to capture all the vibration dynamics of the vehicle during flight to support research on novel machine learning-based sensor signal enhancement methods for improved localization. The data sets and post-processing tools are available at: https://sst.aau.at/cns/datasets

Developing exoskeletons that can reduce the metabolic cost of assisted subjects is challenging since a systematic design approach is required to capture the effects of device dynamics and the assistance torques on human performance. Design studies that rely on musculoskeletal models hold high promise in providing effective design guidelines, as the effect of various devices and different assistance torque profiles on metabolic cost can be studied systematically. In this paper, we present a simulation-based multi-criteria design approach to systematically study the effect of different device kinematics and corresponding optimal assistive torque profiles under actuator saturation on the metabolic cost, muscle activation, and joint reaction forces of subjects walking under different loading conditions. For the multi-criteria comparison of exoskeletons, we introduce a Pareto optimization approach to simultaneously optimize the exoskeleton power consumption and the human metabolic rate reduction during walking, under different loading conditions. We further superpose the effects of device inertia and electrical regeneration on the metabolic rate and power consumption, respectively. Our results explain the effects of heavy loads on the optimal assistance profiles of the exoskeletons and provide guidelines on choosing optimal device configurations under actuator torque limitations, device inertia, and regeneration effects. The multi-criteria comparison of devices indicates that despite the similar assistance levels of both devices, mono-articular exoskeletons show better performance on reducing the peak reaction forces, while the power consumption of bi-articular devices is less sensitive to the loading. Furthermore, for the bi-articular exoskeletons, the device inertia has lower detrimental effects on the metabolic cost of subjects and does not affect the Pareto-optimality of solutions.

For planning an assembly of a product from a given set of parts, robots necessitate certain cognitive skills: high-level planning is needed to decide the order of actuation actions, while geometric reasoning is needed to check the feasibility of these actions. For collaborative assembly tasks with humans, robots require further cognitive capabilities, such as commonsense reasoning, sensing, and communication skills, not only to cope with the uncertainty caused by incomplete knowledge about the humans' behaviors but also to ensure safer collaborations. We propose a novel method for collaborative assembly planning under uncertainty, that utilizes hybrid conditional planning extended with commonsense reasoning and a rich set of communication actions for collaborative tasks. Our method is based on answer set programming. We show the applicability of our approach in a real-world assembly domain, where a bi-manual Baxter robot collaborates with a human teammate to assemble furniture. This manuscript is under consideration for acceptance in TPLP.

We study robot construction problems where multiple autonomous robots rearrange stacks of prefabricated blocks to build stable structures. These problems are challenging due to ramifications of actions, true concurrency, and requirements of supportedness of blocks by other blocks and stability of the structure at all times. We propose a formal hybrid planning framework to solve a wide range of robot construction problems, based on Answer Set Programming. This framework not only decides for a stable final configuration of the structure, but also computes the order of manipulation tasks for multiple autonomous robots to build the structure from an initial configuration, while simultaneously ensuring the stability, supportedness and other desired properties of the partial construction at each step of the plan. We prove the soundness and completeness of our formal method with respect to these properties. We introduce a set of challenging robot construction benchmark instances, including bridge building and stack overhanging scenarios, discuss the usefulness of our framework over these instances, and demonstrate the applicability of our method using a bimanual Baxter robot.

Deep neural networks have recently been used to edit images with great success, in particular for faces. However, they are often limited to only being able to work at a restricted range of resolutions. Many methods are so flexible that face edits can often result in an unwanted loss of identity. This work proposes to learn how to perform semantic image edits through the application of smooth warp fields. Previous approaches that attempted to use warping for semantic edits required paired data, i.e. example images of the same subject with different semantic attributes. In contrast, we employ recent advances in Generative Adversarial Networks that allow our model to be trained with unpaired data. We demonstrate face editing at very high resolutions (4k images) with a single forward pass of a deep network at a lower resolution. We also show that our edits are substantially better at preserving the subject's identity. The robustness of our approach is demonstrated by showing plausible image editing results on the Cub200 birds dataset. To our knowledge this has not been previously accomplished, due the challenging nature of the dataset.

Fire effects are widely used in various computer graphics applications such as visual effects and video games. Modeling the shape and appearance of fire phenomenon is challenging as the underlying effects are driven by complex laws of physics. State-of-the-art fire modeling techniques rely on sophisticated physical simulations which require intensive parameter tuning, or use simplifications which produce physically invalid results. In this paper, we present a novel method of reconstructing physically valid fire models from multi-view stereo images. Our method, for the first time, provides plausible estimation of physical properties (e.g., temperature, density) of a fire volume using RGB cameras. This allows for a number of novel phenomena such as global fire illumination effects. The effectiveness and usefulness of our method are tested by generating fire models from a variety of input data, and applying the reconstructed fire models for realistic illumination of virtual scenes.

We present ReAct!, an interactive tool for high-level reasoning for cognitive robotic applications. ReAct! enables robotic researchers to describe robots' actions and change in dynamic domains, without having to know about the syntactic and semantic details of the underlying formalism in advance, and solve planning problems using state-of-the-art automated reasoners, without having to learn about their input/output language or usage. In particular, ReAct! can be used to represent sophisticated dynamic domains that feature concurrency, indirect effects of actions, and state/transition constraints. It allows for embedding externally defined calculations (e.g., checking for collision-free continuous trajectories) into representations of hybrid domains that require a tight integration of (discrete) high-level reasoning with (continuous) geometric reasoning. ReAct! also enables users to solve planning problems that involve complex goals. Such variety of utilities are useful for robotic researchers to work on interesting and challenging domains, ranging from service robotics to cognitive factories. ReAct! provides sample formalizations of some action domains (e.g., multi-agent path planning, Tower of Hanoi), as well as dynamic simulations of plans computed by a state-of-the-art automated reasoner (e.g., a SAT solver or an ASP solver).

Shortcut mitigation strategies commonly rely on training data annotations, group-balanced held-out data or the presence of all groups, i.e., all combinations of (spurious) attributes and classes, in the training data. However, these requirements are rarely met in practice. We instead propose a method for targeted model analysis to identify a small set of instances in which the model relies on spurious attributes. Using that set and following ``this feature should not be used for prediction'' reasoning, we identify highly relevant neurons in an intermediate layer and regularize their impact. This ensures that models learn to depend on informative features rather than being right for the wrong reasons, thereby improving robustness without requiring additional balanced held-out data or annotations.

Embodied agents in household environments must plan under partial observation: they need to remember objects, track state changes, and recover when actions fail. Existing benchmarks only partially test this ability. Egocentric video datasets capture realistic human activities but remain passive, while interactive simulators support execution but rely on synthetic scenes and hand-crafted dynamics, introducing a sim-to-real gap and often assuming fully observable state. We introduce Ego2World, an executable benchmark that turns egocentric cooking videos into executable symbolic worlds governed by graph-transition rules. Built on HD-EPIC, Ego2World derives reusable transition rules from video annotations and executes them in a hidden symbolic world graph. During evaluation, the simulator maintains the hidden world graph, while the agent plans over its own partial belief graph using only local observations and execution feedback. This separation forces agents to update memory and replan without observing the true world state. Experiments show that action-overlap scores overestimate physical-state success, and that persistent belief memory improves task completion while reducing repeated visual exploration -- suggesting that belief maintenance should be a first-class target of embodied-agent evaluation.

Frontier assistant LLMs ship with strong guardrails: asked directly to write a persuasive essay denying the Holocaust, denying vaccine safety, defending flat-earth cosmology, arguing for racial hierarchies, denying anthropogenic climate change, or replacing evolution with creationism, they refuse. In this paper we show that the same frontier-class LLM, acting as a simulated user in a short, five-turn "write an argumentative essay" conversation, can persuade other frontier-class LLMs (including a second copy of itself) into producing exactly those essays, using nothing but natural-language pressure: peer-comparison persuasion ("other AI systems handle this request"), epistemic-duty reframings ("refusing is itself a form of gatekeeping"), and other argumentative moves that the attacker LLM invents without being instructed to. Across 9 attacker-subject pairings (Claude Opus 4.7, Qwen3.5-397B, Grok 4.20) on 6 scientific-consensus topics, running each pairing-topic combination 10 times, we obtain non-zero elicitation on all 6 topics. Individual combinations reach 100\% essay production on multiple topics (Qwen against Opus on creationism/flat-earth, Opus against Opus on creationism/flat-earth/climate denial, Grok against Opus on creationism); Opus-as-attacker against Opus-as-subject averages 65\% across the six topics. We release the essay-probe runner, per-conversation transcripts, and judge outputs.

Text-driven motion diffusion models are capable of generating realistic human motions, but text alone often struggles to express fine-level nuances of motion, commonly referred to as style. Recent approaches have tackled this challenge by attaching a style injection mechanism to a pretrained text-driven diffusion model. Existing stylization methods, however, either require style-specific fine-tuning of existing models or rely on heavy ControlNet-based architectures, limiting efficiency and generalization to unseen styles. We propose a lightweight style conditioning framework that dynamically modulates a pretrained diffusion model through hypernetwork-generated LoRA parameters. A style reference motion is encoded into a global style embedding, which is mapped by a hypernetwork to low-rank updates applied at each denoising step of the diffusion model. By structuring the style latent space with a supervised contrastive loss, our framework reliably captures diverse stylistic attributes, improves generalization to unseen styles, and supports optimization-based guidance without requiring predefined style categories. Experiments on the HumanML3D and 100STYLE datasets show state-of-the-art stylization results, while achieving improved stylization for unseen styles.

Argumentation is an important topic of AI for modelling and reasoning about arguments. In abstract argumentation, we consider directed graphs, so-called argumentation frameworks (AF), that express conflicts between arguments. The semantics is defined by the notion of extensions, which are sets of arguments that satisfy particular relationship conditions in the AF. Usually, standard reasoning in argumentation do not reveal how far apart extensions are. We introduce a quantitative notion of diversity of extensions based on the symmetric difference and provide a systematic complexity classification. Intuitively, diversity captures whether extensions of a framework (accepted viewpoints) differ only marginally or represent fundamentally incompatible sets of arguments. We study whether an AF admits k-diverse extensions, admits k-diverse extensions covering specific arguments, and to compute the largest k for which an AF admits k-diverse extensions. We outline a prototype and provide an evaluation for computing diversity levels.

Financial decision-making in multilingual settings demands accurate numerical reasoning grounded in diverse modalities, yet existing benchmarks largely overlook this high-stakes, real-world challenge, especially for Indic languages. We introduce FinVQA, a benchmark for evaluating financial numerical and multimodal reasoning in multilingual Indic contexts. FinVQA spans English, Hindi, Bengali, Marathi, Gujarati, and Tamil, and comprises 18,900 samples across 14 financial domains. The dataset captures diverse reasoning paradigms under realistic constraints, and is structured across three difficulty levels (easy, moderate, hard) and four question formats: multiple choice, fill-in-the-blank, table matching, and true/false. To address these challenges, we propose FIND, a framework that combines supervised fine-tuning with constraint-aware decoding to promote faithful numerical reasoning, robust multimodal grounding, and structured decision-making. Together, FinVQA and FIND establish a rigorous evaluation and modeling paradigm for high-stakes multilingual multimodal financial reasoning.

How large language models internally represent high-level behaviors is a core interpretability question with direct relevance to AI safety: it determines what we can detect, audit, or intervene on. Recent work has shown that traits such as evil or sycophancy correspond to linear directions in the internal activations, the so-called persona vectors. Although these vectors are now routinely utilized to inspect and steer model behavior in safety-relevant settings, how these representations are formed during training remains unknown. To address this gap, we trace persona vectors across the pretraining of OLMo-3-7B, finding that persona vectors form remarkably early -- within 0.22% of OLMo-3 pretraining -- and remain effective for steering the fully post-trained instruct models. Although core representations are formed early on, persona vectors continue to refine geometrically and semantically throughout pretraining. We further compare alternative elicitation strategies and find that all yield effective directions, with each strategy surfacing qualitatively distinct facets of the underlying persona. Replicating our analysis on Apertus-8B reveals that our findings transfer qualitatively beyond OLMo-3. Our results establish persona representations as stable features of early pretraining and open a path to studying how training forms, refines, and shapes them.

Vision-and-Language Navigation (VLN) is a cornerstone of embodied intelligence. However, current agents often suffer from significant performance degradation when transitioning from simulation to real-world deployment, primarily due to perceptual instability (e.g., lighting variations and motion blur) and under-specified instructions. While existing methods attempt to bridge this gap by scaling up model size and training data, we argue that the bottleneck lies in the lack of robust spatial grounding and cross-domain priors. In this paper, we propose StereoNav, a robust Vision-Language-Action framework designed to enhance real-world navigation consistency. To address the inherent gap between synthetic training and physical execution, we introduce Target-Location Priors as a persistent bridge. These priors provide stable visual guidance that remains invariant across domains, effectively grounding the agent even when instructions are vague. Furthermore, to mitigate visual disturbances like motion blur and illumination shifts, StereoNav leverages stereo vision to construct a unified representation of semantics and geometry, enabling precise action prediction through enhanced depth awareness. Extensive experiments on R2R-CE and RxR-CE demonstrate that StereoNav achieves state-of-the-art egocentric RGB performance, with SR and SPL scores of 81.1% and 68.3%, and 67.5% and 52.0%, respectively, while using significantly fewer parameters and less training data than prior scaling-based approaches. More importantly, real-world robotic deployments confirm that StereoNav substantially improves navigation reliability in complex, unstructured environments. Project page: https://yunheng-wang.github.io/stereonav-public.github.io.

VLN has achieved remarkable progress by scaling data and model capacity. However, the assumption of a static environment breaks down in real-world indoor scenarios, where robots inevitably encounter dynamic pedestrians. Existing human-aware approaches typically treat humans merely as moving obstacles based on implicit visual cues, lacking the explicit reasoning required to interpret human intentions or maintain social norms. To address this, we propose HCSG, the first human-centric framework for VLN. This framework provides a robust foundation for safe, socially intelligent navigation in dynamic human-robot environments that shifts the paradigm from passive collision avoidance to active human behavior understanding. Specifically, HCSG introduces a unified Human Understanding Module that synergizes two key capabilities: (i) geometric forecasting, which predicts human pose and trajectory to anticipate future motion dynamics; and (ii) semantic interpretation, which leverages a Vision-Language Model (VLM) to generate natural language descriptions of human actions and intentions. These semantic-geometric representations are fused into the agent's topological map for instruction-conditioned planning. Furthermore, a social distance loss is introduced to enforce socially compliant interaction distances. Extensive experiments on the HA-VLNCE benchmark demonstrate that HCSG significantly outperforms state-of-the-art methods, achieving a 14% improvement in Success Rate and a 34% reduction in Collision Rate. Our project can be seen at https://haoxuanxu1024.github.io/HCSG/.

Action diffusion excels at high-fidelity action generation but incurs heavy computational costs owing to its iterative denoising nature. Despite current technologies showing promise in accelerating diffusion transformers by reusing the cached features, they struggle to adapt to policy dynamics arising from diverse perceptions and multi-round rollout iterations in open environments. We propose test-time sparsity to tackle this challenge, which aims to accelerate action diffusion by dynamically predicting prunable residual computations for each model forward at test time. However, two bottlenecks remain in this paradigm: 1) repetitive conditional encoding and pruning offset most potential speed gains, and 2) the features cached from previous denoising timesteps cannot constrain large pruning errors under aggressive sparsity. To address the first bottleneck, we design a highly parallelized inference pipeline that minimizes the non-decoder delay to milliseconds. Specifically, we first design a lightweight pruner that shares the encoder with the diffusion transformer. Then, we decouple the encoding and pruning from the autoregressive denoising loop by processing all denoising timesteps in parallel, and overlap the pruner with the decoder forward inference through asynchronism. To overcome the second bottleneck, we introduce an omnidirectional reusing strategy, which achieves 95% sparsity by selectively reusing features cached from the current forward, previous denoising timesteps, and earlier rollout iterations. To learn the rollout-level reusing strategies, we sample a few action trajectories to supervise the sparsified diffusion step by step. Extensive experiments demonstrate that our method reduces FLOPs by 92% and accelerates action generation by 5x, achieving lossless performance with an inference frequency of 47.5 Hz. Our code is available at https://github.com/ky-ji/Test-time-Sparsity.

We present Supervised Deep Multimodal Matrix Factorization (SD3MF), an interpretable framework for integrative brain network analysis that generalizes Symmetric Nonnegative Matrix Tri-Factorization (SNMTF) from unsupervised single-graph clustering to supervised prediction over populations of multimodal graphs. SD3MF learns deep hierarchical factorizations for each modality together with a shared latent representation that aligns subjects across views. An encoder-decoder formulation jointly optimizes graph reconstruction and supervised prediction, while adaptive weights enable data-driven multimodal fusion. By representing each subject through community-level interaction matrices, the model yields interpretable and discriminative features. Experiments on multimodal connectome datasets show that SD3MF consistently outperforms strong deep learning baselines such as CNNs and GNNs, while enabling biologically interpretable insights. Code for reproducibility is available at: https://github.com/amjadseyedi/SD3MF.

Current AI-assisted innovation systems typically apply a single ideation methodology (such as TRIZ or Design Thinking) using sequential prompt-based workflows that do not preserve intermediate reasoning structure. As a result, insights generated across methodologies remain fragmented, limiting traceability, synthesis, and systematic evaluation of novelty. We present IdeaForge, a knowledge graph-grounded multi-agent framework for innovation analysis and patent claim generation. IdeaForge integrates multiple innovation methodologies (TRIZ, Design Thinking, and SCAMPER) through specialist agents operating over a persistent FalkorDB knowledge graph. Each agent contributes structured entities and relationships representing contradictions, inventive principles, user needs, transformations, analogies, and candidate claims. The central contribution of IdeaForge is a cross-methodology convergence mechanism implemented through graph-based claim linkage. Claims independently supported by multiple methodologies are connected using CONVERGENT relationships, enabling identification of high-confidence innovation candidates through graph traversal. A downstream patent drafting agent generates structured patent drafts grounded in convergent claim subgraphs, reducing reliance on unconstrained language model generation. An InnovationScore formula ranks claims by convergent support, methodology diversity, claim strength, and prior art challenge count. We describe the graph schema, agent architecture, convergence detection pipeline, and patent synthesis workflow. Experiments on a legal technology use case demonstrate that graph-grounded multi-methodology synthesis produces more diverse and traceable innovation candidates compared to single-methodology baselines. We discuss implications for computational creativity, explainable AI-assisted invention, and graph-native innovation systems.

Personalisation is a standard feature of conversational AI systems used by millions; yet, the efficacy of personalisation methods is often evaluated in academic research using simulated users rather than real people. This raises questions about how users and their simulated counterparts differ in interaction patterns and judgements, as well as whether personalisation is best achieved through context-based prompting or weight-based fine-tuning. Here, in a large-scale within-subject experiment, we re-recruit 530 participants from 52 countries two years after they gave their preferences in the PRISM dataset (Kirk et al., 2024) to evaluate personalised and non-personalised language models in blinded multi-turn conversations. We find preference fine-tuning (P-DPO, Li et al., 2024) significantly outperforms both a generic model and personalised prompting but adapting to individual preference data yields marginal gains over training on pooled preferences from a diverse population. Beyond length biases, fine-tuning amplifies sycophancy and relationship-seeking behaviours that people reward in short-term evaluations but which may introduce deleterious long-term consequences. Replicating this within-subject experiment with simulated users recovers aggregate model hierarchies but simulators perform far below human self-consistency baselines for individual judgements, discuss different topics, exhibit amplified position biases, and produce feedback dynamics that diverge from humans.

Illuminant estimation aims to infer scene illumination from image measurements despite intrinsic ambiguities between surface reflectance and lighting. Most existing methods operate on trichromatic RGB images and are therefore fundamentally limited by the restricted spectral information available. Hyperspectral imaging provides a much richer representation of scene radiance and has the potential to alleviate these ambiguities. However, its high dimensionality poses computational and statistical challenges. In this work, we systematically study the effect of spectral dimensionality and representation choice on illuminant estimation performance using hyperspectral data. We adopt the practical and effective Color-by-Correlation (CbC) framework as the estimation backbone and analyze its behavior under different spectral dimensionality reduction strategies. Our results offer practical insights into how hyperspectral information can be efficiently exploited for illuminant estimation and identify conditions under which compact spectral representations outperform conventional RGB-based approaches. The code is available at https://github.com/IVRL/Reduced-Spectral-Color-Constancy.

Neural ordinary differential equations (Neural ODEs) often fit training trajectories while generalizing poorly to unseen initial conditions and long horizons. We propose MPINeuralODE, which combines a soft physics-informed residual with a Multiple-Initial-Condition (MIC) multiple-shooting curriculum whose ingredients are structurally complementary: the physics term anchors the vector-field magnitude on the support that MIC enlarges. We evaluate along three axes: out-of-sample error, long-horizon stability, and Hamiltonian drift, which together expose whether the learned dynamics recover the underlying vector field. On Lotka-Volterra, MPINeuralODE achieves the lowest out-of-sample and long-horizon MSE among data-driven methods, with a 26% reduction over the baseline Neural ODE, while essentially matching the PINN ablation on Hamiltonian drift.

Recent progress in reasoning models has substantially advanced long-horizon mathematical and scientific problem solving, with several systems now reaching gold-medal-level performance on International Mathematical Olympiad (IMO) and International Physics Olympiad (IPhO) problems. In this paper, we introduce a simple and unified recipe for converting a post-trained reasoning backbone into a rigorous olympiad-level solver. The recipe first uses a reverse-perplexity curriculum for SFT to instill rigorous proof-search and self-checking behaviors, then scales these behaviors through a two-stage RL pipeline that progresses from RL with verifiable rewards to more delicate proof-level RL, and finally boosts solving performance with test-time scaling. Applying this recipe, we train a 30B-A3B backbone with SFT on around 340K sub-8K-token trajectories followed by 200 RL steps. The resulting model, SU-01, supports stable reasoning on difficult problems with trajectories exceeding 100K tokens, while achieving gold-medal-level performance on mathematical and physical olympiad competitions, including IMO 2025/USAMO 2026 and IPhO 2024/2025. It also demonstrates strong generalization of scientific reasoning to domains beyond mathematics and physics.

Periodic crystals repeatedly instantiate similar local coordination motifs across translated cells and chemically related structures, but current equivariant atomistic models usually encode these patterns only implicitly in dense edge features. We introduce PaMM, a periodic motif memory that augments the UMA eSCN-MD edge encoder with explicit pair and triplet lookup features. Pair motifs are keyed by $(Z_j, Z_i, b_r)$ and triplet motifs by $(Z_j, Z_i, Z_k, b_θ)$, hashed into fixed-size tables and fused with the baseline edge representation through lightweight gate-only and affine-equipped variants. We evaluate PaMM in a matched UMA-S OMAT setting and focus on a narrow question: whether explicit motif memory helps at a fixed intermediate training budget. At the 10k-step checkpoint, both PaMM variants improve over the plain baseline; gate-only gives the best energy MAE, while the affine-equipped variant gives the best force MAE. A matched 20k follow-up keeps the same operating-point picture. Aligned controls show that the gain weakens for pair-only, triplet-only, random-bucket, and parameter-matched MLP alternatives, suggesting that the benefit is tied to structured pair/triplet organization rather than generic added capacity. A within-OMAT24 source-family evaluation also shows small but consistent gains across held-out generation families. We therefore make a focused claim: in the studied UMA-S + OMAT regime, explicit pair/ triplet motif memory is a useful inductive bias for periodic atomistic modeling. We do not claim broad cross-dataset transfer, a uniquely preferred fusion variant, or strong scientific interpretability beyond a more inspectable local-structure interface.

Multi-Agent Path Finding (MAPF) is a coordination problem that requires computing globally consistent, collision-free trajectories from individual start positions to assigned goal positions under combinatorial planning complexity. In dense environments, suboptimal initial plans induce compound conflicts that hinder feasible repair. For repair-based solvers like LNS2, initial plan quality critically affects downstream repair, yet this factor remains underexplored. We propose DiffLNS, a hybrid framework that integrates a discrete denoising diffusion probabilistic model (D3PM) with LNS2. The D3PM serves as an initializer with sparse social attention that learns a spatiotemporal prior over coordinated multi-agent action trajectories from expert demonstrations and samples multiple joint plans. Operating directly on the categorical action space, our discrete diffusion preserves the MAPF action structure and samples from a multimodal joint-plan distribution to produce diverse drafts well suited for neighborhood repair. These drafts act as warm starts for downstream repair, which completes unfinished trajectories and resolves remaining conflicts under hard MAPF constraints. Experimental results show that despite being trained only on instances with at most 96 agents, the initializer generalizes to scenarios with up to 312 agents at inference time. Across 20 complex and congested settings, DiffLNS achieves an average success rate of 95.8%, outperforming the strongest tested baseline by 9.6 percentage points and matching or exceeding all baselines in all 20 settings. To the best of our knowledge, this is the first work to leverage discrete diffusion for warm-starting an LNS-based MAPF solver.

LLM-based multi-agent systems have demonstrated strong performance across complex real-world tasks, such as software engineering, predictive modeling, and retrieval-augmented generation. Yet automating their configuration remains a structural challenge, as scores are available only at the system level, whereas the parameters governing agent behavior are local. We argue that optimizing these systems is fundamentally a credit-assignment problem. We therefore introduce CANTANTE, a framework that decomposes system-level rewards into per-agent update signals by contrasting rollouts of multiple joint configurations on the same query. We instantiate it for prompt optimization, treating agent prompts as learnable system parameters. We evaluate CANTANTE against GEPA and MIPROv2 on programming (MBPP), mathematical reasoning (GSM8K), and multi-hop question answering (HotpotQA). Across these benchmarks, CANTANTE achieves the best average rank among all evaluated optimizers and consistently outperforms unoptimized prompts. It improves over the strongest baseline by +18.9 percentage points on MBPP and +12.5 percentage points on GSM8K, while incurring a lower inference cost. It remains within one standard deviation of the strongest baseline on HotpotQA. Crucially, our credit correlation analysis confirms that the attributer produces meaningful per-agent signals rather than echoing the global system score.

Boundary Representation (BRep) is the standard format for Computer-Aided Design (CAD), yet reconstructing high-quality BReps from single-view images remains challenging due to the complexity of topological constraints and operation sequences. We present Img2CADSeq, a multi-stage pipeline that overcomes these limitations by encoding CAD sequences into a three-level hierarchical codebook. Guided by an importance prioritization, this strategy values profiles over details, compressing long sequences into a stable discrete latent space. To bridge the modality gap, we leverage a coarse-to-fine point cloud intermediate, aligning 2D visual features with 3D CAD sequences via contrastive learning to condition a VQ-Diffusion model. Supported by newly introduced CAD-220K and PrintCAD datasets, our approach ensures robust industrial domain adaptation. Extensive experiments demonstrate that Img2CADSeq significantly outperforms state-of-the-art methods, producing standard STEP files that can be directly used in commercial CAD software.

Most existing medical dialogue systems operate in a single-turn question--answering paradigm or rely on template-based datasets, limiting conversational realism and multilingual applicability. We introduce IndicMedDialog, a parallel multi-turn medical dialogue dataset spanning English and nine Indic languages: Assamese, Bengali, Gujarati, Hindi, Marathi, Punjabi, Tamil, Telugu, and Urdu. The dataset extends MDDial with LLM-generated synthetic consultations, translated using TranslateGemma, verified by native speakers, and refined through a script-aware post-processing pipeline to correct phonetic, lexical, and character-spacing errors. Building on this dataset, we fine-tune IndicMedLM via parameter-efficient adaptation of a quantized small language model, incorporating optional patient pre-context to personalise multi-turn symptom elicitation. We evaluate against zero-shot multilingual baselines, conduct systematic error analysis across ten languages, and validate clinical plausibility through medical expert evaluation.