arXiv:2606.25763v1 Announce Type: new Abstract: Real-world photography requires capture-time guidance for both camera framing and subject pose. Yet existing aesthetic cropping benchmarks mainly evaluate post-hoc crop prediction and overlook subject-side recommendations, leaving the capture-time guidance capabilities of multimodal large language models (MLLMs) underexplored. To address this gap, we introduce CaptureGuide-Bench, a benchmark with two complementary tasks: photographer-side composition decision and refinement, and subject-side scene-conditioned pose recommendation. Our evaluation reveals limitations: general-purpose MLLMs can make composition decisions but lack precise refinement localization, while specialized aesthetic cropping models localize crops effectively but are limited to refinement; neither provides actionable pose guidance. To support model development, we further construct CaptureGuide-Dataset, comprising 130K samples with textual rationales and structured visual annotations, and develop ShutterMuse, a unified MLLM trained with supervised and reinforcement fine-tuning. Experiments on CaptureGuide-Bench show that ShutterMuse achieves the best overall photographer-side performance among evaluated baselines and competitive subject-side pose recommendation with substantially lower inference cost, demonstrating the potential of MLLMs as interactive assistants for photography during image capture.
Science Journals
arXiv:2602.05704v2 Announce Type: replace Abstract: Understanding the limitations of gradient methods, and stochastic gradient descent (SGD) in particular, is a central challenge in learning theory. To that end, a commonly used tool is the Statistical Queries (SQ) framework, which studies performance limits of algorithms based on noisy interaction with the data. However, it is known that the formal connection between the SQ framework and SGD is tenuous: Existing results typically rely on adversarial or specially-structured gradient noise that does not reflect the noise in standard SGD, and (as we point out here) can sometimes lead to incorrect predictions. Moreover, many analyses of SGD for challenging problems rely on non-trivial algorithmic modifications, such as restricting the SGD trajectory to the sphere or using very small learning rates. To address these shortcomings, we develop a new, non-SQ framework to study the limitations of standard vanilla SGD, for single-index and multi-index models (namely, when the target function depends on a low-dimensional projection of the inputs). Our results apply to a broad class of settings and architectures, including (potentially deep) neural networks.
arXiv:2602.06020v3 Announce Type: replace Abstract: How do protein structure prediction models fold proteins? We investigate this question through causal interventions on the folding trunks of ESMFold, OpenFold, and Boltz-1. Across all three models, we find a shared two-stage computational structure. In the first stage, early blocks initialize pairwise biochemical signals: features like charge propagate from sequence into pairwise representations through architecture-specific pathways. In the second stage, late blocks develop pairwise spatial features: distance and contact information accumulate in the pairwise representation. We verify these mechanisms causally by showing that steering charge and distance features induces predictable structural changes. Furthermore, these representations are functionally interchangeable: pairwise states can be linearly aligned and substituted across models. Together, these results suggest that folding trunks with different architectures, inputs, and training procedures converge on a shared representational organization for mapping sequence chemistry into spatial geometry.
arXiv:2606.25770v1 Announce Type: new Abstract: Data scarcity is a major bottleneck in medical Multiple Instance Learning (MIL), especially for rare diseases or expensive modalities. We introduce a statistically grounded patient augmentation approach that generates realistic patients directly in embedding space. Using Gaussian Mixture Models as a probabilistic clustering approach on pooled instance embeddings from all patients, our method learns disease-specific "recipes"-statistical distributions of instances across unsupervised clusters. New patients are then generated by sampling embeddings from clusters based on learned recipes. Unlike existing methods that require examples from all categories, our method can generate patients offline by re-mixing pooled embeddings. Generated patients are further selected based on uncertainty quantification to improve MIL performance. We evaluate our method across three clinically relevant scarcity scenarios: (i) cross-dataset transfer, where an entirely missing "healthy" class is generated using statistics from an external cohort; (ii) low-data regimes, where class sizes are extremely limited; and (iii) small-cohort non-image tasks, including single-cell RNA-seq and flow cytometry. Across all experiments, our method improves performance over baseline, often outperforming other bag-mixing strategies. Notably, in the missing-class scenario, a performance comparable to full-dataset training is achieved, demonstrating its potential for rare disease diagnostic and privacy-preserving patient augmentation. The code is available at https://github.com/marrlab/RECIPE
arXiv:2602.17104v2 Announce Type: replace Abstract: We propose a streamlined spectral algorithm for community detection in the two-community stochastic block model (SBM) under constant edge density assumptions. By reducing algorithmic complexity through the elimination of non-essential preprocessing steps, our method directly leverages the spectral properties of the adjacency matrix. We demonstrate that our algorithm exploits specific characteristics of the second eigenvector to achieve improved error bounds that approach information-theoretic limits, representing a significant improvement over existing methods. Theoretical analysis establishes that our error rates are tighter than previously reported bounds in the literature. Comprehensive experimental validation confirms our theoretical findings and demonstrates the practical effectiveness of the simplified approach. Our results suggest that algorithmic simplification, rather than increasing complexity, can lead to both computational efficiency and enhanced performance in spectral community detection.
arXiv:2606.25910v1 Announce Type: new Abstract: A central challenge in modern energy market design is the formulation of a strategy-proof imbalance settlement layer that secures both the economic efficiency of the institution and the stability of the power grid. Public data reveals that the day-ahead market is strategically biased below actual consumer demand. Such empirical observations are explained by active prosumers which provide implementable incentives for demand under-reporting. Active prosumers buy energy in the day-ahead market and sell energy in the real-time market for balancing real-time energy deviations. By under-reporting their demand for the day ahead they inflate real-time imbalances and, under uniform pricing, they dispatch their generation assets more profitably. We model the two-stage institution under linear preferences and benchmark it against its associated competitive equilibria. We show that although consumers' incentives for demand under-reporting vanish when the day-ahead market scales, prosumers' incentives remain lower bounded by a positive gain which depends only on the real-time market generation stack and their shares over it. To restore incentive compatibility under the existing informational constraints, we design a leave-one-out contrastive scoring rule-based penalty that is implemented by the day-ahead market operator, incentivizes prosumers to report their demand truthfully and ensures small charges when participating honestly. We illustrate these results with numerical simulations on synthetic data and evaluate our mechanism on real-market data by first rationalizing demand reports as subjective equilibria of the induced game. Our mechanism demonstrates strong incentive alignment while retaining a low cost for honest participation.
arXiv:2606.25774v1 Announce Type: new Abstract: Many laws of chemistry are exact within a limited scope and acquire a separate caveat outside it, and the caveats are usually treated as unrelated. This work argues that they share one cause. Each caveat marks a point where a question is asked of a description too coarse to answer it: the question belongs to a richer level of structure than the description carries. To make these levels explicit, the paper builds a tower of categories over the free symmetric monoidal category of a Petri net, the simplest categorical presentation of a reaction network. The levels, from the bottom up, are stoichiometry, thermochemistry, equilibrium, reaction kinetics, reaction mechanism, molecular geometry, and electronic structure. Each adds one kind of chemical content over the level below, and a forgetful functor runs back down. One question runs through the tower: what can a level express that the level below cannot? Answering it places each measurable quantity at the level where it lives, and identifies the content the levels below could record but not account for. The method is then turned on with two pieces of known chemistry. It recasts a classical criterion for when a reaction network has a unique stable equilibrium, the deficiency-zero theorem, as the rigidity of a single forgetful fibre. It also follows one familiar reaction up the tower: the ring opening that the Woodward-Hoffmann rules govern. At each level, from stoichiometry to electronic structure, the reaction becomes a distinct categorical object.
arXiv:2602.22188v2 Announce Type: replace Abstract: Modelling rock-fluid interaction requires solving a set of partial differential equations (PDEs) to predict the flow behaviour and the reactions of the fluid with the rock on the interfaces. Conventional high-fidelity numerical models require a high resolution to obtain reliable results, resulting in huge computational expense. This restricts the applicability of these models for multi-query problems, such as uncertainty quantification and optimisation, which require running numerous scenarios. As a cheaper alternative to high-fidelity models, this work develops eight surrogate models for predicting the fluid flow in porous media. Four of these are reduced-order models (ROM) based on one neural network for compression and another for prediction. The other four are single neural networks with the property of grid-size invariance; a term which we use to refer to image-to-image models that are capable of inferring on computational domains that are larger than those used during training. In addition to the novel grid-size-invariant framework for surrogate models, we compare the predictive performance of UNet and UNet++ architectures, and demonstrate that UNet++ outperforms UNet for surrogate models. Furthermore, we show that the grid-size-invariant approach is a reliable way to reduce memory consumption during training, resulting in good correlation between predicted and ground-truth values and outperforming the ROMs analysed. The application analysed is particularly challenging because fluid-induced rock dissolution results in a non-static solid field and, consequently, it cannot be used to help in adjustments of the future prediction.
arXiv:2603.01057v2 Announce Type: replace Abstract: Precursor peaks in the wavenumber $k_{\mathrm{peak}}(t)$ maximizing the curl-of-vorticity spectrum have been observed to precede the dissipation peak in decaying turbulence. Because small perturbations in the initial condition can shift peak times, the associated lead time should be characterized statistically. We perform a pseudospectral DNS ensemble of $N_s=1000$ perturbed Taylor--Green vortex realizations at $N=256^3$ and $\nu=10^{-3}$. For each run we extract $k_{\mathrm{peak}}(t)$, several definitions of the precursor time $t_k$, the dissipation-peak time $t_\varepsilon$, and run-wise extrema including $K_{\max}=\max_t k_{\mathrm{peak}}(t)$ and $M_{\max}=\max_t\max_k \mathcal{C}(k,t)$, where $\mathcal{C}(k,t)$ is the isotropic curl-of-vorticity spectrum. The distribution of $\Delta t_{\varepsilon,k}=t_\varepsilon-t_k$ shows that the precursor typically leads, while rare lagging realizations occur and are strongly conditioned on $K_{\max}$. Using peaks-over-threshold extreme-value theory, we fit generalized Pareto models to the right tails of $X=-\Delta t_{\varepsilon,k}$ and $M_{\max}$; the negative shape estimates are consistent with effective bounded tails under the present finite-resolution sampling protocol and provide protocol-dependent endpoint estimates. Finally, $M_{\max}$ correlates strongly with $\varepsilon_{\max}$ and ensemble cross-correlations reveal a reproducible phase offset, consistent with an empirical association between high-curvature activity and dissipation bursts.
arXiv:2603.09170v3 Announce Type: replace Abstract: Achieving versatile and natural whole-body humanoid interaction control remains challenging due to the high cost of whole-body teleoperation data. We present ZeroWBC, a teleoperation-free framework that learns humanoid whole-body interaction from human egocentric videos paired with synchronized whole-body motion and text annotations. ZeroWBC adopts a generation-then-tracking formulation to tackle the static scene whole-body interaction control problem. Given an initial egocentric image and a language instruction, a fine-tuned Vision-Language Model generates future human whole-body motion tokens, which are decoded into continuous motions and retargeted to the humanoid. The resulting reference motions, together with root and key body-part trajectories, are then executed by a general interactive motion tracking policy. To improve interaction performance, we introduce an interaction-oriented tracking reward that prioritizes global root and key body-part trajectory alignment while preserving natural whole-body motion. Experiments on the Unitree G1 humanoid robot show that ZeroWBC enables diverse scene-aware behaviors without robot teleoperation demonstrations. These results suggest a scalable paradigm for learning natural humanoid whole-body interaction from human egocentric data.
arXiv:2606.25787v1 Announce Type: new Abstract: When a large language model (LLM) answers a question about a company, it grounds the answer in retrieved web sources, and those sources decide what the model says. Most analysis of AI brand visibility looks at the answer text. This study looks one step earlier, at the citations. We merge three Rankfor.AI datasets covering 128 brands across 12 home markets and 13 languages, and analyse 167,551 URL-grounded citations (189,974 total attribution rows). We classify each citation by domain and source type and measure where AI gets its brand information, by language and by market. Four patterns hold. First, AI grounds brand answers overwhelmingly in third-party sources: 85.7% of citations point to sites the brand does not own, against 14.3% owned. Second, the source base is concentrated and long-tailed: 80% of citations come from about 18% of domains, fitting a Zipf law (alpha = 0.86, R^2 = 0.983). Third, one reference site dominates almost everywhere: Wikipedia is the most-cited domain in 11 of 12 languages, the exception being Lithuanian, where the business daily vz.lt edges it (4.38%). Fourth, the source mix is market-specific at the margin: for 46 Polish national brands the most-cited domain is YouTube, and four HR and careers portals supply 637 citations against 297 for Polish Wikipedia, about twice as many.
arXiv:2606.25788v1 Announce Type: new Abstract: Medium Access Control (MAC) address randomization has been widely adopted during the IEEE 802.11 network discovery phase as a countermeasure against passive tracking. This paper exposes vulnerabilities in these privacy protocols by demonstrating that devices remain identifiable using Machine Learning (ML)-based fingerprinting. To study the potential tracking capabilities of a passive attacker, we evaluate different eavesdropping scenarios and configurations. To this end, we extract unencrypted hardware specifications from Probe Frames, which we combine with the Inter-Probe Frame Arrival Time (IFAT) and Simulated Received Signal Strength Indication (SRSSI) signals. A core contribution of this paper is the bitwise decomposition of the High Throughput (HT) capabilities information field, which improves device identification accuracy. We evaluate this de-randomization approach using three unsupervised clustering algorithms (K-Means, DBSCAN, and OPTICS) across a dataset of 22 devices from six manufacturers. Our results show that DBSCAN, when using decomposed HT capabilities information and three SRSSI measurements, achieves a global accuracy up to 89.6%. This suggests that the existing MAC randomization solutions are insufficient and underscores the need for enhancing privacy within Wi-Fi standardization.
arXiv:2603.19305v2 Announce Type: replace Abstract: Humanoid robots are expected to execute agile and expressive whole-body motions in real-world settings. Existing text-to-motion generation models are predominantly trained on captured human motion datasets, whose priors assume human biomechanics, actuation, mass distribution, and contact strategies. When such motions are directly retargeted to humanoid robots, the resulting trajectories may satisfy geometric constraints (e.g., joint limits and pose continuity) and appear kinematically reasonable. However, they frequently violate the physical feasibility required for real-world execution. To address these issues, we present PhyGile, a unified framework that closes the loop between robot-native motion generation and General Motion Tracking (GMT). PhyGile performs physics-prefix-guided robot-native motion generation at inference time, directly generating robot-native motions in a 262-dimensional skeletal space with physics-guided prefixes, thereby eliminating inference-time retargeting artifacts and reducing generation-execution discrepancies. Before physics-prefix adaptation, we train the GMT controller with a curriculum-based mixture-of-experts scheme, followed by post-training on unlabeled motion data to improve robustness over large-scale robot motions. During physics-prefix adaptation, the GMT controller is further fine-tuned with generated objectives under physics-derived prefixes, enabling agile and stable execution of complex motions on real robots. Extensive offline and real-robot experiments demonstrate that PhyGile expands the frontier of text-driven humanoid control, enabling stable tracking of agile, highly difficult whole-body motions that go well beyond walking and low-dynamic motions typically achieved by prior methods.
arXiv:2603.19964v3 Announce Type: replace Abstract: High-resolution geometric prediction is essential for robust perception in autonomous driving, robotics, and AR/MR, but current foundation models are fundamentally limited by their scalability to real-world, high-resolution scenarios. Direct inference on 2K images with these models incurs prohibitive computational and memory demands, making practical deployment challenging. To tackle the issue, we present 2K Retrofit, a novel framework that enables efficient 2K-resolution inference for any geometric foundation model, without modifying or retraining the backbone. Our approach leverages fast coarse predictions and an entropy-based sparse refinement to selectively enhance high-uncertainty regions, achieving precise and high-fidelity 2K outputs with minimal overhead. Extensive experiments on widely used benchmark demonstrate that 2K Retrofit consistently achieves state-of-the-art accuracy and speed, bridging the gap between research advances and scalable deployment in high-resolution 3D vision applications. Code will be released upon acceptance.
arXiv:2603.20452v2 Announce Type: replace Abstract: Longitudinal neuroimaging is essential for modeling disease progression in Alzheimer's disease (AD), yet irregular sampling and missing visits pose substantial challenges for learning reliable temporal representations. To address this challenge, we propose SDE-HGNN, a stochastic differential equation (SDE)-driven spatio-temporal hypergraph neural network for irregular longitudinal fMRI connectome modeling. The framework first employs an SDE-based reconstruction module to recover continuous latent trajectories from irregular observations. Based on these reconstructed representations, dynamic hypergraphs are constructed to capture higher-order interactions among brain regions over time. To further model temporal evolution, hypergraph convolution parameters evolve through SDE-controlled recurrent dynamics conditioned on inter-visit intervals, enabling disease-stage-adaptive connectivity modeling. We also incorporate a sparsity-based importance learning mechanism to identify salient brain regions and discriminative connectivity patterns. Extensive experiments on the OASIS-3 and ADNI cohorts demonstrate consistent improvements over state-of-the-art graph and hypergraph baselines in AD progression prediction. The source code is available at https://anonymous.4open.science/r/SDE-HGNN-017F.
arXiv:2604.00287v3 Announce Type: replace Abstract: Distributed energy resources (DERs) such as rooftop solar, batteries, demand response, and electric vehicles can contribute to power system reliability, yet their performance is difficult to translate into firm resource adequacy (RA) capacity across jurisdictions. Existing analyses often locate this difficulty within individual technical requirements, such as metering, accreditation, or dispatch performance, but give less attention to how constraints at one stage carry over to the next. This review traces the RA participation pathway through five stages: load forecasting, registration and classification, metering and verification, capacity accreditation, and performance obligations. We synthesize literature, tariffs, market manuals, and regulatory documents from California, PJM, ISO-NE, Great Britain, and Ireland, spanning U.S. capacity markets and European capacity remuneration mechanisms. Across these frameworks, similar barriers recur despite different procurement models and regulatory structures, indicating that participation is constrained by cross-stage design, not jurisdiction-specific rules alone. We identify three cross-stage couplings through which capacity value is lost between stages: mismatches between resource classification and operational obligations, weak links between verification evidence and accreditation, and temporal misalignment between planning forecasts and scarcity-hour performance. The central finding is that compliance architecture, not DER technology alone, is often the binding constraint on translating DER capability into firm RA contributions. This points to reforms that codify cross-stage information handoffs, tie accreditation to auditable verification evidence, and refresh capacity values as deployment changes system conditions. Rather than adjusting individual stages in isolation, RA reform should redesign the participation pathway end-to-end.
arXiv:2604.02596v3 Announce Type: replace Abstract: In-context learning (ICL) allows large language models (LLMs) to adapt to new tasks from a few examples, making it promising for languages underrepresented in pre-training. Recent work on many-shot ICL suggests that modern LLMs can further benefit from larger ICL examples enabled by their long context windows. However, such gains depend on careful example selection, and the inference cost can be prohibitive for low-resource language communities. In this paper, we present an empirical study of many-shot ICL for machine translation from English into ten truly low-resource languages recently added to FLORES+. We analyze the effects of retrieving more informative examples, using out-of-domain data, and ordering examples by length. Our findings show that many-shot ICL becomes more effective as the number of examples increases. More importantly, we show that BM25-based retrieval substantially improves data efficiency: 50 retrieved examples roughly match 250 many-shot examples, while 250 retrieved examples perform similarly to 1,000 many-shot examples. We further show that ICL provides additional gains on top of fine-tuning.
arXiv:2604.03723v4 Announce Type: replace Abstract: Controlling both camera motion and object dynamics is essential for coherent and expressive video generation, yet current methods typically handle only one motion type or rely on ambiguous 2D cues that entangle camera-induced parallax with true object movement. We present SymphoMotion, a unified motion-control framework that jointly governs camera trajectories and object dynamics within a single model. SymphoMotion features a Camera Trajectory Control mechanism that integrates explicit camera paths with geometry-aware cues to ensure stable, structurally consistent viewpoint transitions, and an Object Dynamics Control mechanism that combines 2D visual guidance with 3D trajectory embeddings to enable depth-aware, spatially coherent object manipulation. To support large-scale training and evaluation, we further construct RealCOD-25K, a comprehensive real-world dataset containing paired camera poses and object-level 3D trajectories across diverse indoor and outdoor scenes, addressing a key data gap in unified motion control. Extensive experiments and user studies show that SymphoMotion significantly outperforms existing methods in visual fidelity, camera controllability, and object-motion accuracy, establishing a new benchmark for unified motion control in video generation. Codes and data are publicly available at https://grenoble-zhang.github.io/SymphoMotion/.
arXiv:2604.03943v2 Announce Type: replace Abstract: We report time-resolved photoelectron spectroscopy of the $3s3p$ doubly excited states of helium dressed by an intense near-infrared (NIR) laser field. Using synchronized XUV free-electron-laser and 800-nm NIR laser pulses, we observe a pronounced delay-dependent shift of resonance-related spectral minima together with the emergence of additional structures around the NIR sideband energy. \textit{Ab initio} theoretical calculations support these observations and identify the features as signatures of NIR-induced coupling of the bright ($3s3p {}^{1}P^{o}$) autoionizing state to nearby dark ($^{1}D^{e}$ and $^{1}S^{e}$) resonances below the $N = 3$ threshold. A multichannel Fano resonance analysis of the measured spectra yields delay-dependent line-shape parameters and resonance energies, establishing a quantitative route to characterize and control correlated two-electron resonances in strong laser fields.
arXiv:2604.05738v2 Announce Type: replace Abstract: Medical Vision-Language Models (Med-VLMs) have achieved expert-level proficiency in interpreting diagnostic imaging. However, current models are predominantly trained on professional literature, limiting their ability to communicate findings in the lay register required for patient-centered care. While text-centric research has actively developed resources for simplifying medical jargon, there is a critical absence of large-scale multimodal benchmarks designed to facilitate lay-accessible medical image understanding. To bridge this resource gap, we introduce MedLayBench-V, the first large-scale multimodal benchmark dedicated to expert-lay semantic alignment. Unlike naive simplification approaches that risk hallucination, our dataset is constructed via a Structured Concept-Grounded Refinement (SCGR) pipeline. This method enforces strict semantic equivalence by integrating Unified Medical Language System (UMLS) Concept Unique Identifiers (CUIs) with micro-level entity constraints. MedLayBench-V provides a verified foundation for training and evaluating next-generation Med-VLMs capable of bridging the communication divide between clinical experts and patients.
Kuramoto Oscillatory Phase Encoding: Neuro-inspired Synchronization for Improved Learning Efficiency
arXiv:2604.07904v2 Announce Type: replace Abstract: Spatiotemporal neural dynamics and oscillatory synchronization are widely implicated in biological information processing and have been hypothesized to support flexible coordination such as feature binding. By contrast, most deep learning architectures represent and propagate information through activation values, neglecting the joint dynamics of rate and phase. In this work, we introduce Kuramoto oscillatory Phase Encoding (KoPE) as an additional, evolving phase state to Vision Transformers, incorporating a neuro-inspired synchronization mechanism to advance learning efficiency. We show that KoPE can improve training, parameter, and data efficiency of vision models through synchronization-enhanced structure learning. Moreover, KoPE benefits tasks requiring structured understanding, including semantic and panoptic segmentation, representation alignment with language, and few-shot abstract visual reasoning (ARC-AGI). Theoretical analysis and empirical verification further suggest that KoPE can accelerate attention concentration for learning efficiency. These results indicate that synchronization can serve as a scalable, neuro-inspired mechanism for advancing state-of-the-art neural network models. Code is avaliable at https://github.com/microsoft/Neuro-inspired_Phase_Encoding.
arXiv:2604.07945v2 Announce Type: replace Abstract: As the demand for mobile robots continues to increase, social navigation has emerged as a critical task, driving active research into deep reinforcement learning (RL) approaches. However, because pedestrian dynamics and social conventions vary widely across different regions, simulations cannot easily encompass all possible real-world scenarios. Real-world RL, in which agents learn while operating directly in physical environments, presents a promising solution to this issue. Nevertheless, this approach faces significant challenges, particularly regarding constrained computational resources on edge devices and learning efficiency. In this study, we propose incremental residual RL (IRRL). This method integrates incremental learning, which is a lightweight process that operates without a replay buffer or batch updates, with residual RL, which enhances learning efficiency by training only on the residuals relative to a base policy. Through the simulation experiments, we demonstrated that, despite lacking a replay buffer, IRRL achieved performance comparable to those of conventional replay buffer-based methods and outperformed existing incremental learning approaches. Furthermore, the real-world experiments confirmed that IRRL can enable robots to effectively adapt to previously unseen environments through the real-world learning.
arXiv:2604.09661v2 Announce Type: replace Abstract: Many natural systems posses, and can transition between, multiple alternative states. For example, a climate ``tipping element'' is a climate component that can transition to an alternative steady state due to an external perturbation such as global warming. Despite the potential impact, existence of alternate states in realistic, complex simulations (e.g. climate models) remain poorly understood. Arguably a reason for this is the lack of applicable methodology that explicitly targets finite yet high-dimensional datasets. In this work we utilize recent progress in computational nonlinear dynamics to formulate a workflow that analyses potentially multi-state simulation data and decides algorithmically what are the alternate states contained within, if any are clearly distinguishable. The framework undergoes an optimization routine that showcases which observables in the data best differentiate the alternate states, and which ones do not differentiate at all, which could be used to guide monitoring and early-warning for multistable components in climate or ecosystems. Finally, once the alternate states have been found, we define an indicator called ``intermingledness''. It quantifies differences and similarities between alternate states, as well as for their basins of attraction (if applicable), across various diagnostic variables. We analyse and present results using three diverse climate datasets: Atlantic ocean circulation, atmospheric midlatitude flow, and habitability of exoplanets. The method is not exclusive to climatic data, but applicable to a variety of cases, including complex networks such as power grids or biological networks. We also provide easy-to-use open source code for applying the workflow to new data.
arXiv:2604.23178v2 Announce Type: replace Abstract: LLM-as-a-Judge has become the dominant paradigm for evaluating language model outputs, yet LLM judges exhibit systematic biases that compromise evaluation reliability. We present a comprehensive empirical study comparing nine debiasing strategies across five judge models from four provider families (Google, Anthropic, OpenAI, Meta), three benchmarks (MT-Bench n=400, LLMBar n=200, custom n=375), and four bias types. Our headline practical finding is that a mid-tier model with the right debiasing can outperform frontier judges at a fraction of the cost: Gemini 2.5 Flash with the Combined Budget strategy reaches the highest agreement of any configuration we tested (71.0%, kappa=0.549) at ~$0.001 per evaluation, about 15x cheaper than the best frontier setup (Claude Sonnet 4, 69.5%, ~$0.015). Other key findings: (1) Style bias is the dominant bias (0.10-0.76 across models, favoring markdown over plain prose), far exceeding position bias (<=0.04), yet is rarely studied. (2) Verbosity bias is heterogeneous when measured length-aware: Pro, Flash, and Llama prefer longer answers (+0.24 to +0.44), Claude prefers concise (-0.12), and GPT-4o is neutral (-0.04); on truncation controls all models correctly prefer the complete response (0.88-1.00 accuracy). (3) Debiasing helps multiple models: Claude S8 (+11.5 pp), Flash S8 (+7.5 pp), and Claude S5 (+7.3 pp) survive Holm-Bonferroni correction, with Flash S1 (+4.7 pp) and Llama S8 (+4.5 pp) also significant. We release our evaluation framework, the 375-pair controlled dataset, and per-instance cached results for all nine strategies.
arXiv:2604.28019v2 Announce Type: replace Abstract: Barvinok introduced the symmetrized determinant ($\sdet$) as a \emph{non-commutative} analogue of the determinant. Intuitively, given a square matrix over an associative algebra, we can obtain the symmetrized determinant by averaging over all possible multiplication orders in the Leibniz formula for the determinant. He used the symmetrized determinant to design algorithms estimating the permanent of a matrix. To this end, he showed that there is a $O(n^{r+3})$ algorithm computing $\sdet$, where $r$ is the dimension of the algebra, and is therefore polynomial-time computable for fixed $r$. In this work, we study the algebraic properties and complexity of $\sdet$. While most of the properties of the ordinary determinant don't generalize to $\sdet$ defined on non-commutative algebras, we show that the principal minor expansion of the $\sdet$ is analogous to the ordinary determinant. Second, we prove that there exists a polynomial-sized algebra such that computing the symmetrized determinant is $\sharpP$-hard. Third, we show that the associated polynomial family is $\VNP$-complete over a suitable polynomial-dimensional algebra in the non-commutative setting. Further, when seen as a family of polynomials over the matrix algebra, it is also $\VNP$-complete in the commutative setting. This places the symmetrized determinant among the natural complete families arising from algebraic computation.