arXiv:2412.19419v2 Announce Type: replace
Abstract: Graph neural networks are deep neural networks designed for graphs with attributes attached to nodes or edges. The number of research papers in the literature concerning these models is growing rapidly due to their impressive performance on a broad range of tasks. This survey introduces graph neural networks through the encoder-decoder framework and provides examples of decoders for a range of graph analytic tasks. It uses theory and numerous experiments on homogeneous graphs to illustrate the behavior of graph neural networks under different training sizes and degrees of graph complexity, with an emphasis on oversmoothing and oversquashing.
Science Journals
arXiv:2605.30748v2 Announce Type: replace
Abstract: We present Chatterbox-Flash, a zero-shot text-to-speech model obtained by fine-tuning a pretrained autoregressive TTS decoder into a block-diffusion decoder, enabling parallel token generation within each block while retaining block-by-block streaming. We find that naively transferring mainstream block-diffusion decoding to discrete speech tokens degrades quality, as a long-tail token distribution biases parallel position selection toward a few high-frequency tokens. To mitigate this without architectural modification, we introduce two inference-time techniques: prior-calibrated scoring, which subtracts the block-level marginal token distribution, and an early-decoding schedule, which adaptively terminates iteration based on calibrated confidence. On standard zero-shot TTS benchmarks, Chatterbox-Flash attains high-fidelity synthesis comparable to strong autoregressive and non-autoregressive baselines, while supporting streaming inference with time-to-first-packet on par with streaming AR systems and substantially lower real-time factor. Code and audio samples are available at https://github.com/resemble-ai/chatterbox-flash.
arXiv:2605.30706v2 Announce Type: replace
Abstract: Permacomputing is a nascent concept and community of practice concerned with developing alternative computing systems grounded in principles of resilience, reuse, sufficiency, and ecological limits. However, research engaging with permacomputing remains in an early stage of development, raising concerns about whether permacomputing can move beyond reflective critique to become a meaningful alternative practice. Through a research-through-design case study, we documented our experience moving a personal website from a data centre in Texas to a self-hosted solar-powered server built from reclaimed electronics. Guided by permacomputing principles and relational aesthesis, we explore what it takes for permacomputing to reconfigure material and perceptual relations. Our findings reveal the frictions of moving away from a maximalist techno-aesthetic while attempting to re-use already existing technologies, potential ways to overcome these challenges through building a community of practice, and the transformative potential of visibilizing and visceralizing digital infrastructures to cultivate more responsible ways of relating to technology. This paper contributes to emerging research on permacomputing and its aesthetics by bringing it into dialogue with theories of non-place and relational aesthesis. Rather than functioning as a purely symbolic gesture, permacomputing practices can cultivate greater collective autonomy, agency, and responsibility in how communities engage and create meaning within digital infrastructures. In the context of socio-ecological crises and anti-colonial transformation, our research offers a situated approach to building and relating to computing technologies in the ashes of dominant technological paradigms.
arXiv:2606.01636v1 Announce Type: new
Abstract: Post-training via Group Relative Policy Optimization (GRPO) has emerged as a powerful paradigm for aligning flow-based generative models with human preferences. However, the iterative denoising nature of flow models incurs substantial costs when generating group rollouts for policy-gradient updates, compelling existing methods to train with extremely few denoising steps. This temporal sparsity severely restricts preference optimization: reward feedback can only reach a handful of stages per trajectory, leaving the vast majority of intermediate denoising steps without direct supervision and thus compromising alignment granularity. To address this, we propose Pave-GRPO, which reformulates the GRPO objective through Principled average velocity decomposition. Rather than generating expensive high-step rollouts, we maintain efficient few-step group sampling but decompose each coarse transition into an equivalent ensemble of finer sub-trajectories spanning multiple intermediate timesteps. This propagates reward feedback to a denser set of temporal stages for more comprehensive preference alignment without additional generation cost. This design offers two benefits: (i) zero-cost horizon expansion: through the direct reuse of piece-wise group samples and their associated rewards, Pave-GRPO significantly broadens the effective optimization scope under fixed sampling budgets; and (ii) comprehensive temporal supervision: by equivalently decomposing an instantaneous velocity target into a multi-timestep ensemble, it distributes reward signals across more intermediate stages of the denoising process, enabling finer-grained and more thorough preference optimization. Extensive experiments validate that Pave-GRPO effectively advances preference alignment across different reward settings, offering comprehensive performance enhancement.
arXiv:2606.02375v1 Announce Type: new
Abstract: We evaluate whether compact domain-specialized ASR models can outperform massively multilingual foundation models for conversational African speech across 19 languages in the WAXAL corpus. Fine-tuned edge models achieve a macro-averaged WER of $38.0\%$ compared to $64.9\%$ for the best zero-shot baseline, a $26.9$ percentage-point reduction using models $3-40\times$ smaller. Results confirm that domain specialization dominates scale for spontaneous African speech. Cross-domain evaluation shows that fine-tuned models recover usable performance on out-of-distribution (OOD) speech, while zero-shot models regain an advantage when the test domain matches their pretraining distribution. A distributed native-speaker audit across all surveyed languages produces a linguistically-grounded error taxonomy, showing that CTC and autoregressive architectures behave differently across language families. We further show that WER alone misrepresents performance for syllabary-script languages where CER/WER ratios reveal substantially higher character-level accuracy than headline WER suggests. Finally, to contribute to future African ASR research, we release all model weights, fine-tuning and evaluation scripts, and a cleaned WAXAL subset covering all $19$ languages.
arXiv:2508.03556v4 Announce Type: replace
Abstract: Process Reward Model (PRM) is widely used in the post-training of Large Language Model (LLM) because it can perform fine-grained evaluation of the reasoning steps of generated content. However, most PRMs lack long-term reasoning and deep thinking capabilities. On the other hand, although a few works have tried to introduce Chain-of-Thought (CoT) capability into PRMs, the annotation cost of CoT-PRM data is too expensive to play a stable role in various tasks. To address the above challenges, we propose VRPRM, a process reward model via visual reasoning, and design an efficient two-stage training strategy. Experimental results show that using only 3.6K CoT-PRM Supervised Fine-Tuning(SFT) data and 50K non-CoT PRM Reinforcement Learning (RL) training data, VRPRM can surpass the non-thinking PRM with a total data volume of 400K and achieved a relative performance improvement of up to 118\% over the base model in the BoN experiment. This result confirms that the proposed combined training strategy can achieve higher quality reasoning capabilities at a lower data annotation cost, thus providing a new paradigm for PRM training with more efficient data utilization.
arXiv:2508.06588v3 Announce Type: replace
Abstract: Vector Quantization (VQ) has recently emerged as a promising approach for learning compressed and discrete representations for graph-structured data. However, a fundamental challenge, i.e., codebook collapse, remains underexplored in the graph domain, significantly limiting the expressiveness and generalization of graph tokens.In this paper, we present an empirical study and observe that codebook collapse consistently occurs when training VQ jointly with Graph Neural Networks under graph reconstruction tasks, even with mitigation strategies proposed in vision or language domains. Moreover, we provide a diagnosis of collapse from data and optimization perspectives, showing that collapse is associated with graph data properties such as feature redundancy and connectivity density, and is further reinforced by the training dynamics of deterministic hard assignment. To address these issues, we propose RGVQ, a novel framework that integrates graph topology and feature similarity as explicit regularization signals to enhance codebook utilization and promote token diversity. RGVQ introduces soft assignments via Gumbel-Softmax reparameterization, ensuring that all codewords receive gradient updates. In addition, RGVQ incorporates a structure-aware contrastive regularization to penalize assigning the same token to dissimilar node pairs. Extensive experiments demonstrate that RGVQ substantially improves codebook utilization and consistently boosts the performance of state-of-the-art graph VQ backbones across multiple downstream tasks, enabling more expressive and transferable graph token representations.
arXiv:2605.26436v2 Announce Type: replace
Abstract: Discrete masked diffusion language models such as LLaDA generate text through iterative denoising, where mask tokens are progressively replaced with predicted tokens. LLaDA2.1 introduced a Token-to-Token (T2T) editing mechanism that accelerates generation by directly replacing committed tokens suspected of being incorrect. However, we identify fundamental limitations of T2T editing: it couples error detection with replacement, pollutes the generation context with potentially incorrect tokens, and introduces a train-inference noise mismatch where systematic model-generated errors differ from the random perturbations seen during training. We propose Token-to-Mask (T2M) remasking, a training-free, drop-in replacement for T2T editing that resets suspected erroneous tokens back to the mask state, allowing the diffusion process to re-predict them under cleaner context. We design and empirically validate three complementary error detection strategies -- probability-based, trigger-mirrored, and temporal-difference-based -- and provide a unified theoretical analysis showing that T2M remasking purifies the generation context, converts systematic inference errors back to the model's native mask noise type, and enables delayed commitment for joint multi-position optimization. Comprehensive experiments across 12 benchmarks spanning knowledge, reasoning, mathematics, coding, and instruction following show that T2M generally improves performance on tasks requiring precise token-level output, with the largest gain on mathematics (+5.92% on CMATH). Error analysis on CMATH reveals that the dominant failure mode is last-mile token corruption -- where correct reasoning produces a corrupted final answer -- and that T2M repairs 59.4% of such cases.
arXiv:2508.10312v2 Announce Type: replace
Abstract: Recommender systems in concert with Large Language Models (LLMs) present promising avenues for generating semantically-informed recommendations. However, LLM-based recommenders exhibit a tendency to overemphasize semantic correlations within users' interaction history. When taking pretrained collaborative ID embeddings as input, LLM-based recommenders progressively weaken the inherent collaborative signals as the embeddings propagate through LLM backbones layer by layer, as opposed to traditional Transformer-based sequential models in which collaborative signals are typically preserved or even enhanced for state-of-the-art performance. To address this limitation, we introduce FreLLM4Rec, an approach designed to balance semantic and collaborative information from a spectral perspective. Item embeddings that incorporate both semantic and collaborative information are first purified using a Global Graph Low-Pass Filter (G-LPF) to preliminarily remove irrelevant high-frequency noise. Temporal Frequency Modulation (TFM) then actively preserves collaborative signal layer by layer. Note that the collaborative preservation capability of TFM is theoretically guaranteed by establishing a connection between the optimal but hard-to-implement local graph fourier filters and the suboptimal yet computationally efficient frequency-domain filters. Extensive experiments on four benchmark datasets demonstrate that FreLLM4Rec successfully mitigates collaborative signal attenuation and achieves competitive performance, with improvements of up to 8.00\% in NDCG@10 over the best baseline. Our findings provide insights into how LLMs process collaborative information and offer a principled approach for improving LLM-based recommendation systems.
arXiv:2507.09029v5 Announce Type: replace
Abstract: Pre-training large neural networks at scale imposes heavy memory demands on accelerators and often requires costly communication. We introduce Subnetwork Data Parallelism (SDP), a distributed training framework that partitions a model into structured subnetworks trained across workers without exchanging activations. We study two complementary masking regimes: backward masking, which applies sparsity only in the backward step to retain unbiased gradients, and forward masking, which also removes parameters in the forward pass to deliver stronger efficiency gains while providing additional regularization. We further explore two subnetwork construction strategies: neuron level and block level, applied across both transformers and CNNs. In experiments spanning 1B LLaMA pre-training on FineWeb to ResNet-18 on CIFAR, SDP reduces per device memory usage by 28%-60% while maintaining or improving performance under FLOP-matched settings.
arXiv:2507.09766v2 Announce Type: replace
Abstract: Accurate estimation of Remaining Useful Life (RUL) and State of Health (SoH) is essential for reliable Prognostics and Health Management (PHM), supporting timely maintenance and dependable industrial operation. However, hybrid models that combine data-driven learning with physics-based regularization often rely on fixed loss weights and therefore lose accuracy when transferred across assets with different degradation behaviors. This study introduces Reinforced Graph-based Physics-informed Networks with Dynamic Weighting (RGPD), a unified framework for spatio-temporal degradation modeling and adaptive physics-guided regularization. Graph-based representation learning captures inter-sensor degradation structure, a Soft Actor-Critic (SAC) module refines latent features under noisy conditions, and a lightweight Q-learning policy adaptively balances monotonicity, smoothness, and latent-dynamics residual losses during training. The framework is evaluated on the C-MAPSS, PHM2012, and XJTU datasets, which represent engine, bearing, and battery degradation processes. Relative to the strongest compared baselines reported in the corresponding benchmark tables, RGPD improves average RMSE by up to 12 percent on PHM2012 and C-MAPSS, and reduces average MAPE by 20 percent on XJTU compared with the second-best reported model. Performance on these heterogeneous benchmarks further suggests the model's generalizability across degradation systems. The physics-informed component is implemented through degradation-consistent priors together with a Deep Hidden Physics Model-style residual, which improves physical plausibility without requiring a full first-principles model for each asset type.
arXiv:2508.11931v3 Announce Type: replace
Abstract: We present an oracle-efficient, near-optimal algorithm for linear contextual bandits with adversarial losses and stochastic action sets, only requiring a linear optimization oracle for the action sets in each round. Our approach reduces this setting to misspecification-robust adversarial linear bandits with fixed action sets. Without knowledge of the context distribution or access to a context simulator, the algorithm achieves $\widetilde{\mathcal{O}}(\min\{d^2\sqrt{T}, \sqrt{d^3T\log K}\})$ regret and runs in $\mathrm{poly}(d,T)$ time plus $\mathrm{poly}(d,T)$ calls to the linear optimization oracles, where $d$ is the feature dimension, $K$ is an upper bound on the number of actions in each round, and $T$ is number of rounds. This resolves the open question by Liu et al. (2023) on whether one can obtain $\mathrm{poly}(d)\sqrt{T}$ regret in polynomial time independent of the number of actions. For the important class of combinatorial bandits with adversarial losses and stochastic action sets, our algorithm is the first to achieve $\mathrm{poly}(d)\sqrt{T}$ regret in polynomial time, while no prior algorithm achieves even $o(T)$ regret in polynomial time to our knowledge. When a simulator is available, the regret bound can be improved to $\widetilde{\mathcal{O}}(d\sqrt{L^\star})$, where $L^\star$ is the cumulative loss of the best policy.
arXiv:2508.16334v2 Announce Type: replace
Abstract: Alpha mining, aimed at discovering predictive return signals, is typically formulated as symbolic regression. Traditional symbolic methods suffer from search inefficiency and biased prior knowledge. Recently, Large Language Models (LLMs) have emerged as a promising alternative, automatically generating textual thoughts and executable codes to achieve both efficient and interpretable alpha mining. However, existing approaches mostly focus on leveraging LLM's reasoning and reflection capabilities, yet largely neglect the positional bias due to the flat thought representation which restricts efficiency and diversity of the search process. This paper introduces Tree-structured thought Evolution (TreEvo), which evolves hierarchically decomposed thoughts to expand the effective search space. In addition, we propose a set of evolutionary operators tailored to structured thoughts. Experiments on four real-market datasets demonstrate that TreEvo not only obtains competitive alphas with traditional methods in up to 200 times fewer evaluations, but also consistently outperforms LLM-driven EAs across all datasets by $14.31\%$ on average.
arXiv:2508.17320v3 Announce Type: replace
Abstract: Understanding the internal representations of large language models (LLMs) remains a central challenge for interpretability research. Sparse autoencoders (SAEs) offer a promising solution by decomposing activations into interpretable features, but existing approaches rely on fixed sparsity constraints that fail to account for input complexity. We propose AdaptiveK SAE (Adaptive Top K Sparse Autoencoders), a novel framework that dynamically adjusts sparsity levels based on the semantic complexity of each input. Leveraging linear probes, we demonstrate that context complexity is linearly encoded in LLM representations, and we use this signal to guide feature allocation during training. Experiments across ten language models demonstrate that this complexity-driven adaptation outperforms fixed-sparsity approaches on reconstruction fidelity, explained variance, cosine similarity and interpretability metrics while eliminating the burden of extensive hyperparameter tuning. Our code is available at: https://github.com/hiyukie/adaptiveK.
arXiv:2605.30443v2 Announce Type: replace
Abstract: Multilingual large language models can generate figurative language, but whether the internal signals driving this behavior are language-specific or reusable across languages is unclear. Using activation steering as a probe, we estimate a direction for a figurative category from figurative--literal activation differences in one language and apply it during generation. Across five figurative categories, six languages, and four multilingual LLMs, these directions steer reliably within their own language, most robustly for metaphor and simile. More importantly, they transfer across languages: a direction learned in one increases the target behavior when applied to another, with German among the most receptive targets. Going further, directions assembled from other languages can match or even surpass a target language's own native direction, while removing this shared component weakens native steering. Together, these results provide direct evidence of a reusable but target-dependent cross-lingual signal for figurative generation.
arXiv:2605.30380v2 Announce Type: replace
Abstract: Deep convolutional and transformer-based detectors achieve strong performance for SAR ship detection but are often computationally prohibitive for real-time or onboard deployment. Lightweight models offer improved efficiency yet struggle to capture the complex structural relationships inherent in SAR backscatter. Most existing SAR knowledge-distillation approaches rely on feature or logit matching, which enforces localized activation similarity while neglecting the geometric relationships among object representations. We propose a Structured Unified Relational knowledGE distillation framework for SAR Ship detection (SURGE) that transfers relational geometry from a powerful teacher detector to a compact student detector using a contrastive InfoNCE objective in a shared projection embedding space. To the best of our knowledge, this work presents the first transformer-based SAR ship detector knowledge distillation framework in SAR domain. The framework is architecture-agnostic in the sense that it provides a common region-level distillation interface for two-stage, one-stage and transformer-based detectors without modifying their deployed architectures. Experiments on the SSDD and HRSID benchmarks demonstrate that the proposed method yields substantial improvements for two-stage detectors, achieving up to 6.2 mAP and 8.0 AP75 gains over baseline student and even surpassing teacher performance
arXiv:2605.30355v2 Announce Type: replace
Abstract: We introduce Collective Switched Motion (Cosm), a dynamical system-based heuristic algorithm. Cosm combines locally interacting continuous circular variables with novel global coordination rules that facilitate collective dynamics. Pairwise interactions occur sequentially over a set of conflict-free edge partitions, resulting in an interaction network that switches periodically. Unlike conventional gradient-based approaches, Cosm employs structured, non-smooth switching dynamics with finite-magnitude interactions that sustain collective fluctuations and promote exploration beyond local minima. A correlated perturbation mechanism further promotes coordinated cluster motion in the circular phase space. On the three largest Ising problems from the Gset suite, which have 10,000-20,000 variables and represent 2D spin glasses, Cosm attains the optimal solutions (verified with an exact solver) heuristically for the first time. On two large bounded-degree non-lattice graph instances, Cosm reduces the state-of-the-art times-to-target from hundreds of hours to 36-303 s. Results on benchmark problems with tuned hardness suggest favorable scaling relative to previously characterized dynamical solvers. These results suggest that Cosm's synthesis of local interactions, structured switching dynamics, and global coordination provides an effective computational framework for sparse optimization.
arXiv:2509.02107v2 Announce Type: replace
Abstract: We develop high-order numerical schemes to solve random hyperbolic conservation laws using linear programming. The proposed schemes are high-order extensions of the existing first-order scheme introduced in [{\sc S. Chu, M. Herty, M. Luk\'a\v{c}ov\'a-Medvi{\softd}ov\'a, and Y. Zhou}, SIAM J. Sci. Comput., 48 (2026)], where a novel structure-preserving numerical method using a concept of generalized, measure-valued solutions to solve random hyperbolic systems of conservation laws is proposed, yielding a linear partial differential equation concerning the Young measure and allowing the computation of approximations based on linear programming problems. The second-order extension is obtained using piecewise linear reconstructions of the one-sided point values of the unknowns. The fifth-order scheme is developed using the finite-difference alternative weighted essentially non-oscillatory (A-WENO) framework. These extensions significantly improve the resolution of discontinuities, as demonstrated by a series of numerical experiments on both random (Burgers equation, isentropic Euler equations) and deterministic (discontinuous flux, pressureless gas dynamics, Burgers equation with non-atomic support) hyperbolic conservation laws.
arXiv:2509.05367v5 Announce Type: replace
Abstract: Large Language Model safety alignment predominantly operates on a binary assumption that requests are either safe or unsafe. This classification proves insufficient when models encounter ethical dilemmas, where the capacity to reason through moral trade-offs creates a distinct attack surface. We formalize this vulnerability through TRIAL, a multi-turn red-teaming methodology that embeds harmful requests within ethical framings. TRIAL achieves high attack success rates across most tested models by systematically exploiting the model's ethical reasoning capabilities to frame harmful actions as morally necessary compromises. Building on these insights, we introduce ERR (Ethical Reasoning Robustness), a defense framework that distinguishes between instrumental responses that enable harmful outcomes and explanatory responses that analyze ethical frameworks without endorsing harmful acts. ERR employs a Layer-Stratified Harm-Gated LoRA architecture, achieving robust defense against reasoning-based attacks while preserving model utility.
arXiv:2509.06948v3 Announce Type: replace
Abstract: Supervised fine-tuning (SFT) and reinforcement learning with verifiable rewards (RLVR) are two widely used post-training paradigms for improving the reasoning ability of large language models (LLMs). Recent methods attempt to integrate SFT and RLVR in a single stage by reweighting or scheduling their objectives. However, such coupling can be counterproductive because supervised updates are not uniformly beneficial for reward optimization. To address this, we propose BRIDGE, a scalable framework in which SFT learns to supervise RL by selectively transferring knowledge that improves reward optimization. Specifically, BRIDGE alternates two updates at each meta-training step: a base-model update that fuses the SFT and RL gradients, and an update to a lightweight low-rank adapter (LoRA) that coordinates the two objectives by maximizing a cooperative-gain signal, defined as the reward of joint SFT-RL training over an RL-only baseline. Across five mathematical reasoning benchmarks, BRIDGE consistently outperforms two-stage cold start, naive mixing, and representative single-stage integration baselines, yielding over three points average absolute improvement and more stable training dynamics. We further show that BRIDGE extends to logical reasoning and generalizes out-of-distribution to code and science without additional training, while staying robust under noisy rewards.
arXiv:2509.11056v2 Announce Type: replace
Abstract: Artificial intelligence (AI) is anticipated to emerge as a pivotal enabler for the forthcoming sixth-generation (6G) wireless communication systems. However, current research efforts regarding large AI models for wireless communications primarily focus on fine-tuning pre-trained large language models (LLMs) for specific tasks. This paper investigates the large-scale AI model designed for beamforming optimization to adapt and generalize to diverse tasks defined by system utilities and scales. We propose a novel framework based on bidirectional encoder representations from transformers (BERT), termed BERT4beam. We aim to formulate the beamforming optimization problem as a token-level sequence learning task, perform tokenization of the channel state information, construct the BERT model, and conduct task-specific pre-training and fine-tuning strategies. Based on the framework, we propose two BERT-based approaches for single-task and multi-task beamforming optimization, respectively. Both approaches are generalizable for varying user scales. Moreover, the former can adapt to varying system utilities and antenna configurations by re-configuring the input and output module of the BERT model, while the latter, termed UBERT, can directly generalize to diverse tasks, due to a finer-grained tokenization strategy. Extensive simulation results demonstrate that the two proposed approaches can achieve near-optimal performance and outperform existing AI models across various beamforming optimization tasks, showcasing strong adaptability and generalizability.
arXiv:2605.30290v2 Announce Type: replace
Abstract: Self-improvement at scale has been a longstanding goal for reasoning models, and there are two natural places to do it: at test time, through verification-refinement (V-R) loops; and at training time, through self-training methods. Both are gated by the same bottleneck: the verifier. V-R loops stall when verifier scores inflate while accuracy stagnates, and when feedback is too generic to act on; self-training fails similarly when bad self-generated data are added to training. Better verification would unlock both, but the capability we want to train, i.e., catching self-generated errors, lacks training signal. To address this challenge, we propose self-trained verification (STV). Our key observation is that, while a model cannot catch these errors alone, it can when shown the reference solution. We turn this asymmetry into a supervision target and train the verifier to imitate a more informed version of itself. At test time, STV substantially improves V-R loops on hard problems, while alternatives (e.g., SFT, RL on verifier scores, and even meta-verifiers) do not. STV roughly doubles accuracy on hard math and lifts it 14x on scientific reasoning tasks (1.5% to 21%). At training time, we additionally train the generator using RL with STV verifier's feedback inside the V-R loop - a procedure we call verifier-in-the-loop training (ViL). Starting from an RL-converged generator, ViL yields a further 33% gain in pass@1. More notably, the generator's standalone pass@1, with no verifier at test time, climbs 30% relative past where standard RL had converged. Hence, the next frontier in reasoning on hard problems may lie in how we train for and with verification. Website: https://ar-forum.github.io/stv-webpage
arXiv:2605.30190v2 Announce Type: replace
Abstract: Diffusion-based planning has achieved strong results in single-agent offline reinforcement learning, yet scaling to many-agent systems remains intractable due to the curse of dimensionality in the joint trajectory space. We introduce MF-Diffuser, a framework that lifts trajectory planning to the Wasserstein space of trajectory distributions, where the propagation of chaos ensures a small representative subset of agents captures the full population dynamics. Our approach features a value-weighted chaotic entropy objective that reconciles generative fidelity with return maximization, and a hierarchical coarse-to-fine strategy that progressively grows the agent population during denoising. We establish end-to-end suboptimality bounds with four interpretable terms, revealing that mean-field approximation error scales as $O(H^2/\sqrt{N})$ while offline distribution shift provably does not grow with population size $N$, and prove the generated policy is an approximate mean-field Nash equilibrium with explicit convergence guarantees. Experiments on three mean-field RL benchmarks -- spanning stage games, sequential dynamics, and adversarial team competition -- show MF-Diffuser achieves the best return in the majority of settings, with the largest gains on suboptimal offline data and at extreme scales ($N \geq 10^3$).
arXiv:2605.30188v2 Announce Type: replace
Abstract: Reliable probability estimates are critical in many machine learning applications, yet modern classifiers are often poorly calibrated. Post-hoc calibration provides a simple and widely used solution, but the large number of proposed methods, combined with small-scale and inconsistent evaluations, makes it difficult to determine which approaches are truly effective in practice. We introduce a large-scale, standardized benchmark for post-hoc calibration, covering nearly 2000 experiments across tabular and computer vision tasks, including binary, multiclass, and large-scale classification settings. Our benchmark aggregates predictions from a diverse set of classical models, modern deep learning architectures, and foundation models, and provides unified, reproducible implementations of dozens of calibration methods within a common evaluation framework. We argue that Post-Hoc Improvement (PHI) in proper scoring rules offers a principled alternative to traditional calibration error estimators for comparing post-hoc methods, capturing both calibration quality and potential degradation to the model's predictive performance. Using this framework, we conduct the most comprehensive empirical study of post-hoc calibration to date. Our results reveal consistent patterns across domains: smooth calibration functions outperform binning-based approaches, dedicated multiclass methods are essential in high-dimensional settings, and generic machine learning models are not competitive without calibration-specific design. To facilitate future research, we release all data, code, and evaluation tools, providing a plug-and-play benchmark for developing and comparing calibration methods.
arXiv:2605.26068v3 Announce Type: replace
Abstract: Weakly supervised anomaly detection (WSAD) has developed in three primary directions: incomplete, inexact, and inaccurate supervision. However, these directions remain isolated, lacking a unified framework to assess whether they address unique challenges or share fundamental mechanisms. This paper introduces WSADBench, the first benchmark that unifies evaluation across distinct weakly supervised scenarios, benchmarking diverse approaches from specialized WSAD methods to advanced tabular foundation models. WSADBench establishes standardized protocols to evaluate 36 algorithms across 4 modalities by systematically varying label quantity, granularity, and quality, revealing the performance boundaries of various methods. Based on over 700K experiments, WSADBench reveals four critical insights: (i) Strong intrinsic correlations exist between these weak supervision scenarios, challenging the isolation of current research directions. (ii) Specialized WSAD algorithms excel only in extreme label-scarcity regimes but are quickly dominated by tabular foundation models and general classification methods as supervision increases or in OOD scenarios. (iii) Unlabeled data shows inconsistent utility across settings, with marginal gains compared to label refinement. (iv) Models exhibit asymmetric sensitivity to different types of label noise. We release WSADBench as an open-source benchmark with code and datasets to facilitate future WSAD research: https://github.com/SUFE-AILAB/WSADBench.