Machine Learning · Cross-listed from stat.AP
Probabilistic Deep Learning for Drought Forecasting: Role of Internal Climate Variability
Henri Funk, Cornelia Gruber, Göran Kauermann et al. (+2 more)
Predicting drought risk is essential for anticipating impacts on water resources, agriculture, ecosystems, and
climate adaptation planning. Yet drought forecasts remain uncertain because variability can substantially alter regional precipitation and evaporative demand. Treating this variability as unstructured noise ignores the fact that internal variability has spatial, seasonal, and temporal structure and thus contains information that can be used to improve drought
forecasting. We propose a deep-learning-based
forecasting framework for European drought prediction and extend it with an uncertainty-aware drought bound that explicitly incorporates internal forecast variability from a large
climate model ensemble. This bound represents a physically plausible lower-tail trajectory of future drought conditions and marks how severe drought could plausibly become under an unfavourable realisation of internal variability, giving adaptation planning a conservative, risk-averse reference. We compare the proposed bound with a lower bound derived from reanalysis data only and show that our proposed ensemble-informed bound is better calibrated across most regions and seasons. This is specifically true during anomalously dry conditions, when historical reanalysis alone underestimates lower-tail drought risk. Our results show that internal variability should be treated as a forecast quantity in its own right. More broadly, large ensembles provide a practical way to transfer physically plausible
climate variability into machine-learning drought forecasts, yielding risk-aware bounds that are more informative for drought assessment under shifting
climate conditions.
Read the full abstract ▾
All authors Henri Funk, Cornelia Gruber, Göran Kauermann, Helmut Küchenhoff, Magdalena Mittermeier
Machine Learning
Beyond Magnitude and Shape: A Direction-Aware Loss for Time Series Forecasting
Seunghan Lee, Jaehoon Lee, Jun Seo et al. (+9 more)
The direction of change — whether a series will move up or down — is often as important as its exact value in decisiondriven applications such as risk management and financial
forecasting. However, most
forecasting losses optimize either point magnitude or shape and frequency structure, and none explicitly targets the direction of change. In this paper, we find that MSE-trained forecasters fail on the direction of small moves. To address this, we propose CosDir, a simple yet effective direction-aware loss that aligns the difference vectors of the prediction and the target via cosine similarity. Being scale-invariant, CosDir keeps a directional gradient on small moves, re-injecting learning signal exactly where MSE neglects it. CosDir is a lightweight, plug-in term that attaches to any backbone without architectural modification. Since the best ratio for mixing the directional and magnitude terms differs across
datasets, we further propose CosDir-UW, an extension that makes this ratio adaptive by learning it during training, matching a per-
dataset tuned weight with no hyperparameter. We conduct over 100K experiments, demonstrating that our method consistently and significantly improves directional accuracy while preserving magnitude accuracy, and that it outperforms various loss functions. Code is available at: https://github.com/seunghan96/cosdir.
Read the full abstract ▾
All authors Seunghan Lee, Jaehoon Lee, Jun Seo, Junhyeok Kang, Sangjun Han, Sungdong Yoo, Minjae Kim, Tae Yoon Lim, Dongwan Kang, Hwanil Choi, Soonyoung Lee, Wonbin Ahn
Machine Learning
WorldDynCache: Risk-Controlled Latent Dynamics Approximation for Diffusion World Model
Leyang Chen, Junyi Wu, Shaoqiu Zhang et al. (+1 more)
Diffusion world models generate high-quality futures, but re- peated
transformer evaluations make inference prohibitively slow. Existing caches reuse intermediate features, selectively update
tokens, or reuse and extrapolate denoising outputs ac- cording to local drift or short native-space histories. These criteria can miss both approximation-induced latent transition defects that accumulate across skipped steps and phase- or condition-dependent changes in the direction of latent evo- lution. We propose WorldDynCache, a risk-controlled latent dynamics approximation framework with two core compo- nents. First, a lightweight latent-transition risk
estimator tracks the accumulated future impact of approximation defects and calibrates its predictions against counterfactual defects ob- served at exact anchors. Second, a condition- and phase- aware lifted latent surrogate approximates latent
evolution without extra
transformer evaluations. On HunyuanVoyager- 13B and Aether-5B, WorldDynCache achieves 4.92 times and 2.15 times speedups, respectively, while attaining the best gen- eration quality among the compared caching methods across WorldScore, PSNR, SSIM, and LPIPS.
Read the full abstract ▾
All authors Leyang Chen, Junyi Wu, Shaoqiu Zhang, Yulun Zhang
Machine Learning
tFUSOperator: Operator Learning for Transcranial Focused Ultrasound Digital Twins
Minjee Seo, Haris Ghafoor, Minju Seol et al. (+2 more)
Transcranial focused ultrasound (tFUS) requires accurate estimation of the intracranial acoustic field, which is distorted by skull-induced aberrations. Numerical solvers are accurate but computationally expensive for digital twins, where the field must be re-estimated repeatedly as treatment conditions change. Existing deep-learning surrogates are fast but typically use voxel-to-voxel regression on a fixed grid, with no mechanism reflecting how acoustic energy propagates through the skull. We instead cast tFUS simulation as an operator learning problem and propose tFUSOperator, a coordinate-aware neural operator that maps the free-field pressure, skull anatomy, and treatment parameters to the intracranial field within a shared physical coordinate frame. To our knowledge, this is the first operator-based formulation of tFUS field prediction. On both seen and unseen skulls, the model localizes the acoustic focus accurately-reaching about 90% and 72% Dice, respectively-and it performs nearly as well from
magnetic resonance (MR) as from computed
tomography (CT) input while running $5.6 \times 10^4$ times faster than numerical simulation. These results suggest a fast, radiation-free route to safe and practical digital twins for patient-specific tFUS treatment. The code is available at: https://github.com/CMME-Lab/tFUSOperator.git.
Read the full abstract ▾
All authors Minjee Seo, Haris Ghafoor, Minju Seol, Seonaeng Cho, Kyungho Yoon
Machine Learning · Cross-listed from cond-mat.dis-nn
Tunneling the Loss Landscape: Bypassing Memorization with Monte Carlo Parameter Swapping
Lai Shun Chan, Xiaotian Zhang, Yue Shang et al. (+2 more)
Grokking is a striking phenomenon in neural network training, where a model can undergo a prolonged period of pure memorization before abrupt generalization. While previous works have attempted to interpret it through classical machine learning mechanisms like weight norm, recent research draws an analogy from statistical physics, framing grokking as a form of computational glass relaxation. This theory defines the initial memorization as a result of `fast cooling' where the training loss is reduced so quickly that a glass state is formed, followed by a `slow relaxation' towards final generalization. Although providing a unifying framework for representative grokking theories, this perspective has remained largely at the theoretical on macroscopic level without direct empirical validation on training dynamics. Here we introduce a three-component framework to directly characterize the training dynamics via parameter mobility (PM), and two representative measurements from glassy dynamics: replica correlation (RC) and fractal dimension (FD). We demonstrate that standard optimization presents clear signatures of glass dynamics and inherently traps the grokking network in a kinetic arrested memorization state with a collapsed mobility, strong history dependence, and channel-like motions. This quantitative agreement motivates us to introduce State-Aware Monte Carlo Parameter Swapping (SAM-Swap), an optimization plug-in that can accelerate generalization, inspired by swap Monte Carlo
algorithm widely used in glass dynamics. Comparing SAM-Swap, weight decay, and Gaussian gradient noise, we find that accelerated generalization is consistently associated with random exploration in the parameter space, similar to
diffusion in physics.
Read the full abstract ▾
All authors Lai Shun Chan, Xiaotian Zhang, Yue Shang, Ge Zhang, Entao Yang
Machine Learning · Cross-listed from cs.CV
DAVET: Denoising-Aware Visual Evidence Trajectory Allocation for Diffusion Vision-Language Models
Yongkang Zhou, Xiang Xia, Cheng Yan et al. (+2 more)
Diffusion vision-
language models (dVLMs) iteratively denoise masked responses while conditioning each denoising step on visual evidence, making visual conditioning a substantial recurring inference cost. Unlike autoregressive decoding,
diffusion generation repeatedly revisits the entire response as uncertainty evolves. Our analysis reveals that visual evidence demand is strongly step-dependent, motivating adaptive allocation across denoising steps. Existing inference acceleration methods operate through decoding-side strategies or visual
token compression via
pruning and merging, but do not explicitly treat visual evidence as a resource whose demand evolves across the
diffusion process. Therefore, we present Denoising-Aware Visual Evidence Trajectory Allocation (DAVET), a training-free framework that allocates visual evidence according to the evolving generation state. Starting from a phase-conditioned evidence trajectory, the proposed allocation policy uses operation demand to set an evidence reserve whose allocation at each denoising step is modulated by trajectory risk. DAVET realizes the resulting budgets through a hierarchy of evidence views constructed from a single visual encoding, separating when and how much evidence is needed from how the evidence views are constructed. Evaluated on two representative dVLMs, LLaDA-V and LaViDa, across multiple visual-understanding
benchmarks, DAVET achieves an average speedup of 1.55$\times$ with an average relative performance drop of 1.86%, showing that denoising-aware visual evidence allocation can reduce visual conditioning cost while largely preserving generation quality.
Read the full abstract ▾
All authors Yongkang Zhou, Xiang Xia, Cheng Yan, Fan Xu, Wuyang Zhang
Machine Learning · Cross-listed from cs.HC
Reassessing the Feasibility of PPG-Based Non-Invasive Blood Glucose Level Estimation
Supraja Ramesh, Markus Neufeld, Michael Küttner et al. (+2 more)
Non-invasive blood glucose level (BGL) estimation from photoplethysmography (PPG) holds great promise for wearable health monitoring, but results across studies are hard to compare due to inconsistent
datasets, data leakage, and non-standardized evaluation metrics. We present the first reproducible, extensible evaluation pipeline and use it to reassess five representative PPG-based BGL methods on published
datasets under three increasingly strict data-split protocols: random window-level, participant-aware, and leave-some-participants-out (LSPO). Models appeared competitive under random splitting but collapsed under participant-aware and LSPO evaluation, with nearly all yielding near-zero or negative R$^2$ values comparable to a mean-prediction baseline. Critically, across every model and split, over 90% of predictions fell within clinically acceptable zones (Clarke Error Grid A+B), including the baseline. This reveals a fundamental disconnect: clinical zone metrics systematically conceal model failure in this domain. Our findings demonstrate that random train-test splits substantially overestimate the generalization of PPG-based BGL models due to sample-level data leakage, and that robust ML evaluation must precede clinical validation to meaningfully assess real-world utility.
Read the full abstract ▾
All authors Supraja Ramesh, Markus Neufeld, Michael Küttner, Tobias Röddiger, Michael Beigl
Machine Learning
Predictive Maintenance: Deep Learning-Based Remaining Useful Life Prediction for Combat Aircraft Engines
Fatih Ürgen and Doğay Alt{\i}nel
To improve the operational readiness of combat aircraft engines and reduce unplanned maintenance costs, accurately estimating the remaining useful life (RUL) is critical. Traditional maintenance often proves insufficient under dynamic mission profiles. In this study, a deep learning-based predictive maintenance model capable of autonomously extracting features from multivariate sensor data was developed. Using the NASA C-MAPSS FD001 and FD004
datasets, data were converted into sequential blocks via 50- and 30-step sliding windows, respectively. The model's architectural superiority in autonomously extracting temporal degradation features was validated against RF, CNN-LSTM, and BiLSTM baselines. On FD001, it achieved an R-squared (R2) of 0.8901, a 13.28 RMSE, and a 320.34 NASA risk score, demonstrating generalizability on the multi-regime FD004
dataset with a 15.71 RMSE. The proposed maintenance protocol achieved a 0.9973 AUC at the critical 30-cycle threshold, ensuring high reliability. Additionally, a decision-support simulator has been developed to validate this protocol under aggressive combat flight profiles.
Read the full abstract ▾
Machine Learning
LEAP: Lean Environment-Feedback via Adaptive Pruning for Code RL in GPU Kernel Generation
Tankun Li, Zhi Chen and Yaohua Tang
Post-training large
language models (LLMs) via reinforcement learning (RL) has significantly advanced code generation capabilities. To bypass the heavy
memory footprint of critic networks, current state-of-the-art frameworks leverage critic-free paradigms like Group Relative Policy Optimization (GRPO) tied to rule-based verification sandboxes. However, applying these frameworks to low-level systems programming, such as CUDA kernel generation-presents severe challenges: binary pass/fail rewards introduce severe signal sparsity, while multi-turn environmental feedback loops suffer from prohibitive compilation latencies and reward dilution across trajectories. In this work, we introduce LEAP (Lean Environment-Feedback via Adaptive
Pruning), a scalable and computationally efficient multi-turn RL framework optimized for low-level hardware accelerator
alignment. LEAP features Difficulty-Conditioned
Pruning (DCP), a dynamic gating mechanism that adaptively cuts off simple and overly catastrophic tasks from multi-turn expansion, focusing resource-heavy compilation and hardware exploration exclusively on high-value, complex tasks. To fully operationalize these paths without manual hyperparameter engineering, we propose a Rank-Based Reward formulation. By deriving scale-free relative advantages from pairwise tournament outcomes within the GRPO rollout group, our method inherently penalizes
token inefficiency on simple
prompts while maximizing learning gradients on challenging distributions. Empirical evaluations show that LEAP achieves superior first-turn proficiency and robust multi-turn debugging resilience while converging faster than unpruned multi-turn baselines, establishing a practical paradigm for low-level code RL.
Read the full abstract ▾
Machine Learning
ReFP-AD: Rectified Flow Preconditioning for Energy-Based Anomaly Detection
Camile Lendering, Erkut Akdag, Joaquín Figueira et al. (+1 more)
Unified anomaly detection requires modeling highly heterogeneous normal data without access to anomalous samples. While foundation models like DINOv2 provide rich
token representations, leveraging these spaces for explicit density estimation remains challenging. Energy-Based Models (EBMs) offer a principled formulation, but their training in high-dimensional
token spaces is unstable due to anisotropy and strong cross-dimensional correlations, which degrades finite-step Markov Chain Monte Carlo (MCMC) sampling. We identify this instability as fundamentally
geometric and introduce ReFP-AD (Rectified Flow Preconditioning for Anomaly Detection), which learns a
geometric reparameterization that maps high-dimensional
embeddings into a well-conditioned latent space via an optimal transport (OT)-coupled rectified flow. This preconditioning enables stable persistent contrastive divergence with preconditioned
Stochastic Gradient Langevin Dynamics (SGLD) in full-dimensional
token spaces. Anomaly scores are then derived from the learned energy landscape using gradient norms. Under a strict unified protocol on the MVTec-AD and VisA
datasets, ReFP-AD achieves 98.6%/97.9% Image/Pixel AUROC on MVTec-AD and 97.3%/99.0% on VisA, outperforming prior unified EBM baselines by up to +10.8% in Image AUROC. Ablation experiments demonstrate that
geometric reparameterization is critical for finite-step MCMC and accurate anomaly localization in high-dimensional
token spaces. Code is available at https://github.com/CLendering/ReFP-AD
Read the full abstract ▾
All authors Camile Lendering, Erkut Akdag, Joaquín Figueira, Egor Bondarev
Machine Learning
Multi-Source Dynamic Graph Learning for Compound-Flood Forecasting in Managed Coastal Systems
Liangjun You, Min Wu, Orlando Woods et al. (+1 more)
Compound flooding in managed coastal systems is influenced by hydrological conditions and water-management activity observed across multiple monitoring stations. Current
forecasting models can capture temporal dependencies with low average errors, but global error metrics may conceal poor reproduction of prolonged high-water plateaus that are relevant to flood early warning. Because hydrometeorological and operational signals are distributed across heterogeneous gages, single-site records do not fully represent high-water dynamics. Nevertheless, unconstrained fusion of cross-site signals can degrade the stability of local temporal forecasts. This work proposes an anchored
forecasting framework that incorporates cross-site information through state- and lead-dependent bounded residual corrections. A multi-source regime representation constructed from hydrometeorological and operational observations adaptively calibrates inter-site relationships and correction scales, enabling targeted cross-site adjustment while preserving the local temporal forecast as a stable anchor. Beyond conventional global error statistics, we evaluate event-scale high-water characteristics through the temporal
alignment of forecasted and observed high-water processes. Experiments demonstrate that selectively integrating multi-station dynamic conditions improves the prediction reliability of sustained high-water plateaus while maintaining high accuracy during routine hydrological conditions, supporting flood early warning and water-management decision support.
Read the full abstract ▾
All authors Liangjun You, Min Wu, Orlando Woods, Dongsheng Luo
Machine Learning
Heterogeneous Multi-Agent Reinforcement Learning for Radio Resource Management under Coupled Finite-Horizon Constraints
Yeonseo Jeong, Wonhyeok Ko, Sungweon Hong et al. (+1 more)
Maximizing throughput under proportional
fairness in dense wireless networks requires jointly managing user association, scheduling, base station (BS) activation, and handover control under hard finite-horizon energy and handover budgets, which induces a fundamental tension between BS-side energy management and user-side handover regulation. While multi-
agent reinforcement learning (MARL) is a natural framework for such distributed sequential control, its application here faces two difficulties: finite-horizon budget constraints cannot be evaluated at each time slot, and the nonlinear proportional
fairness utility admits no principled per-slot decomposition. We propose HeLyMARL, a Lyapunov-embedded heterogeneous MARL framework that resolves both via drift-plus-penalty decomposition with virtual queues. The energy and handover constraint pressures are internalized directly into a unified per-slot reward, converting the constrained finite-horizon problem into an unconstrained MARL problem. Comparison against two Lagrangian-based alternatives reveals a timescale separation: Lagrangian relaxation regulates constraints only across training episodes, whereas the virtual queues of HeLyMARL bound cumulative budget consumption at every partial horizon within an episode, a pacing guarantee beyond the reach of greedy Lyapunov-based control. Simulations show that HeLyMARL is the only method that sustains the throughput-
fairness balance together with uninterrupted service throughout the horizon, outperforming conventional MARL, Lyapunov-based, and constrained MARL
benchmarks without premature budget exhaustion.
Read the full abstract ▾
All authors Yeonseo Jeong, Wonhyeok Ko, Sungweon Hong, Songnam Hong
Machine Learning
Toward Plasticity-Preserving KL Regularization for Capability Retention in LLM Reinforcement Learning
Li Wang, Xiaodong Lu, Xiaohan Wang et al. (+4 more)
Reinforcement learning (RL) has become a central paradigm for large
language model (LLM) post-training, but optimization toward new objectives can degrade capabilities already present in the base model. KL regularization is widely used to mitigate such forgetting by constraining policy drift toward a reference model. However, standard full-policy KL regularization constrains the entire response distribution and may unnecessarily restrict exploration and target-task learning. This raises a natural question: can a more precise constraint preserve existing capabilities while minimizing interference with learning new tasks? To this end, we propose \underline{Co}rrectness-Conditioned \underline{KL} Regularization (CoKL), a conditional regularization framework that narrows the preservation constraint from the full output distribution to correctness-conditioned response distributions. We instantiate CoKL with forward KL divergence and derive a practical finite-group training objective for RL-based LLM post-training. At the population level, CoKL decouples the total probability assigned to correct responses from their correctness-conditioned distribution, thereby regularizing the relative probability allocation among reference-supported correct responses without directly anchoring incorrect outputs or total correctness mass. We further show that full-policy forward and reverse KL regularization induce a strict optimal correctness gap when the reference policy is imperfect, whereas CoKL avoids this limitation. Experiments in controlled multi-solution environments and continual post-training settings across multiple model scales demonstrate that CoKL achieves a more favorable balance between target-task improvement and prior-capability retention than existing regularization methods. Our code is available at https://github.com/Lumina04/CoKL.
Read the full abstract ▾
All authors Li Wang, Xiaodong Lu, Xiaohan Wang, Jiajun Chai, Wei Lin, Tianhao Peng, Guojun Yin
Machine Learning
Disagree to Accelerate: Closing the Loop on Diffusion Feature Forecasts
Yanchao Li, Jiaqing Xie, Ben Gao et al. (+6 more)
Training-free feature
forecasting accelerates
diffusion sampling by predicting features at skipped denoising steps. Recent work has mainly focused on designing stronger forecasters. Yet forecast error varies sharply across steps, and open-loop caches trust the forecast in full at every skipped step. This fixed trust is what breaks as acceleration turns aggressive. The missing question is not only how to forecast better, but when and how much to trust a forecast. We show that reliability can be observed from the cache itself. Two forecasts agree where the feature trajectory is smooth, and they diverge where prediction turns hard. Their disagreement is a cheap runtime signal, and it costs no extra denoiser evaluation. Based on this signal, we introduce RACER, a training-free closed-loop controller with two responses. It continuously shrinks uncertain forecasts toward the last computed feature. At the riskiest steps, RACER refreshes the feature and repays the added evaluation by skipping a later scheduled one. We derive a deterministic error bound for the shrinkage and empirically evaluate its validity and tightness across acceleration regimes. At the same number of denoiser evaluations, RACER improves the strongest open-loop baseline across SD3.5-Large, FLUX.1-dev, Wan2.1-14B, and HunyuanVideo on DrawBench, VBench, and COCO. On SD3.5, we further show that RACER samples faster at equal quality. RACER generalizes across
forecasting designs as well. For example, it recovers much of the quality lost on a Taylor base. These results show that reliable
diffusion acceleration also depends on how forecasts are used. Code is available at https://github.com/LiZaiyuan0619/RACER
Read the full abstract ▾
All authors Yanchao Li, Jiaqing Xie, Ben Gao, Wanhao Liu, Yanbo Wang, T. Y. Tsui, Jinfei Liu, Yuqiang Li, Tianfan Fu
Machine Learning
LLM-Guided Retrieval for Prediction of Molecular Perturbation Responses
Betty Xiong, Jan-Christian Huetter, Gabriele Scalia et al. (+2 more)
Predicting transcriptomic responses to small-molecule perturbations across
cell lines is central to drug discovery, but exhaustive profiling of drug-
cell combinations is infeasible. We frame molecular perturbation prediction as retrieve-and-aggregate: approximate an unmeasured drug's response in a
cell line by aggregating measured responses of a small set of biologically related compounds. We propose LLM-Guided
Retrieval (LGR), where a large
language model (LLM) ranks candidate neighbor drugs (restricted to those profiled in the target
cell line); after which a fixed mean aggregator combines their observed expression deltas to form the prediction. We evaluate on the Tahoe-100M single-
cell perturbation atlas under unseen-drug, unseen-
cell-line, and open-world regimes. LGR consistently improves over drug mean, ChemCPA, and chemistry-based kNN baselines, with the strongest gains for unseen
cell-line generalization, where it achieves higher correlation and lower error than mean baselines. Across settings, LGR improves directional (sign) accuracy of gene regulation, indicating better recovery of biologically meaningful perturbation effects even when magnitude-based metrics are similar. These results suggest that
retrieval quality, rather than predictor complexity, is a key driver of zero-shot molecular perturbation prediction, and that LLMs can provide a useful biological prior when used as constrained
retrieval modules.
Read the full abstract ▾
All authors Betty Xiong, Jan-Christian Huetter, Gabriele Scalia, Tommaso Biancalani, Sepideh Maleki
Machine Learning · Cross-listed from cs.NI
CENTILE: A Telemetry Foundation Model Evaluated by the Decisions It Drives
Zifan Zhang, Zhichao Hou, Tingxiang Ji et al. (+1 more)
Modern computing and networking infrastructure emits telemetry continuously, yet operators convert it into decisions with a separate predictor per task, entity, and horizon. One generative model, pretrained once over an operator's own event streams, could replace this fleet, an approach that already scales to high-cardinality streams in recommendation systems. However, point-forecast error on operational telemetry saturates near simple last-value baselines, so lower error alone need not improve the decisions it feeds. To close this gap, we present \sys, a generative foundation model for network and systems telemetry, evaluated by replaying the decisions its calibrated conditional
quantiles drive. \sys treats heterogeneous telemetry as event-driven, irregularly timed entity streams and serves flexible forecast horizons in a single pass, requiring no future timestamps. To our knowledge, \sys is the first pretrained telemetry model to improve both HPC scheduling and network provisioning decisions under replay, its runtime
estimator transferring zero-shot across months and its pretrained weights across domains from hours of target data. Extensive experiments on HPC job logs and network traffic confirm that \sys lowers the mean bounded slowdown of backfilling by up to approximately $77\%$ over deployed user estimates and roughly halves the deployed rule's violation rate. Our code is available at https://github.com/ZzZTripleZzZ/all-in-one.
Read the full abstract ▾
All authors Zifan Zhang, Zhichao Hou, Tingxiang Ji, Yuchen Liu
Machine Learning
Beyond On-Policy Exploration: Integrating External Policy Rollouts for Reinforcement Learning in Diffusion Language Models
Wonseok Lee, Jimyeong Kim, Jungmin Ko et al. (+1 more)
Recent reinforcement learning methods for
diffusion large
language models (dLLMs) commonly rely on on-policy rollouts generated by the target dLLM itself. When successful on-policy rollouts are scarce, however, on-policy training may receive little positive reward and make only limited progress. To mitigate this problem, we explore incorporating higher-reward rollouts generated by a stronger external policy alongside on-policy rollouts from the target dLLM. However, directly incorporating these external rollouts introduces two practical challenges: differences in rollout length and instability when jointly processing rewards from on-policy and external rollouts. To address these challenges, we propose External Rollout Integration with Length Control and Source-Specific Processing (ERILS), which controls external-rollout length and processes the rewards of on-policy and external rollouts separately. Experiments on Sudoku, Countdown, and MATH500 under zero-shot evaluation show that ERILS improves multi-sample performance across all three tasks, with the largest gains on Sudoku. On Sudoku, ERILS achieves 98.4% best-of-4 completion accuracy, compared with 40.3% for the strongest baseline. ERILS also maintains approximately 90% deterministic single-completion accuracy on Sudoku across generation lengths of 128, 256, and 512
tokens. Our component analysis further shows that length-controlled external rollouts are more effective than uncontrolled external rollouts, and that source-specific reward processing avoids the training collapse observed with joint reward processing. These results show that rollout construction and reward processing are important design dimensions when integrating external rollouts into dLLM reinforcement learning.
Read the full abstract ▾
All authors Wonseok Lee, Jimyeong Kim, Jungmin Ko, Wonjong Rhee
Machine Learning
Beckmann Transport Models: From Autonomous Flows to One-Step Maps
Lee Cheuk-Kit, Florentin Coeurdoux, Yuyuan Chen et al. (+5 more)
We propose an instantiation of flow matching that relies on a time-independent velocity field (an autonomous flow) to exactly map between two distributions, so long as the target is singular, i.e.\ supported on a lower-dimensional data
manifold. We also show that the one-step generative map associated with this flow is the unique solution of a simple conservation equation, which can be used to learn the map directly from samples. These autonomous flows and maps give a dynamical meaning to the flux constraint of Beckmann's transportation problem. Their construction provides a unifying framework that recovers, for instance, the closed-form Poisson-flow generative model and
equilibrium matching with a quadratic flow-matching regression loss. We illustrate how this theory corrects inconsistencies in existing methods and demonstrate the effectiveness of the autonomous flow and the one-step map on ImageNet 256x256.
Read the full abstract ▾
All authors Lee Cheuk-Kit, Florentin Coeurdoux, Yuyuan Chen, Sophia Tang, Peter Potaptchik, Yilun Du, Michael Samuel Albergo, Eric Vanden-Eijnden
Machine Learning
Progressive Agent Skill Generation via Reinforcement Learning
Junhao Shen, Chengzhi Piao, Zhanqiu Zhang et al. (+2 more)
Recent large
language model
agents often use external skills as modular procedural units that condition inference and improve complex task solving. Thus, automatically generating high-quality skills from documents or experience has become an important problem. Existing skill generation methods largely rely on heuristics or pipeline-style consolidation, which must be specially designed for different evidence sources. In contrast, learning-based approaches offer a more unified way to model skill generation across heterogeneous sources. However, learning-based skill generation remains challenging because skills lack a natural supervision signal based on relevance or correctness; their value can largely be determined only by whether they improve the behavior of the
agent on downstream tasks. To address this challenge, we propose Skill-$\alpha$, a reinforcement learning method that learns a unified policy for progressive skill generation. Specifically, we construct each skill by repeatedly applying the learned policy to successive source evidence and introduce a novel rollback reward that evaluates each edit by comparing downstream execution under the original and edited skills on an anchored query. Extensive experiments show that Skill-$\alpha$ generates more effective skills than methods based on heuristics or pipelines in both document-to-skill and experience-to-skill settings. Under the main GPT-4o worker, Skill-$\alpha$ improves average downstream success rates over the strongest skill-generation baseline by 3.1 points on CL-Bench and 6.7 points on tau2-bench. Further ablations and analysis validate the importance of rollback reward and progressive generation.
Read the full abstract ▾
All authors Junhao Shen, Chengzhi Piao, Zhanqiu Zhang, Yiwen Guo, Hong Cheng
Machine Learning · Cross-listed from cs.CV
Generative Brownian Bridge Diffusion In Motion Space For Enhanced Myocardial Strain Analysis
Rishov Paul, Frederick H. Epstein and Miaomiao Zhang
Myocardial strain analysis of cardiac
magnetic resonance (CMR) images provides an important tool for evaluating cardiac function. However, current techniques require either human-adjusted post-processing with suboptimal regional accuracy, or specialized acquisitions with limited availability. In this paper, we propose to leverage the power of generative models to synthesize high-quality motion-derived strain values from routinely acquired CMR sequences. Specifically, we develop a novel Brownian bridge
diffusion model in motion space to learn the probabilistic mapping between standard CMR motion estimated from widely adopted registration methods and highly accurate motion provided by advanced strain imaging techniques. To promote the fidelity of anatomical structure in the generation process, our model is conditioned on the corresponding CMR images. We validate our method on large-scale multi-center CMR
datasets including subjects of paired standard cine CMR and advanced strain imaging acquisitions. Experimental results demonstrate that our framework significantly improves the accuracy of motion prediction and strain analysis from standard CMRs compared to existing learning-based approaches. Our research represents a new paradigm for potentially developing cost-effective, clinically deployable AI tools for cardiac function assessment with enhanced strain accuracy in busy clinical workflows. Our code is publicly available at https://github.com/Rishov-MIA/Brownian-Bridge-strain-analysis.
Read the full abstract ▾
Machine Learning · Cross-listed from cs.CL
Understanding Sparse Attention Selectivity in Long-Context Foundation Models via Counterfactual Evaluation
Xingyu Ren, Youran Sun, Chugang Yi et al. (+1 more)
Sparse
attention is widely deployed in long-context serving stacks, yet no framework audits how discarding blocks changes the influence of specific content on model output. We first establish that the phenomenon is real and causal: Block Sparse Flash
Attention (BSFA) route replay across four architectures changes output decisions in 13 of 16
cells, with zero identity-replay label flips. We then introduce a dense-calibrated counterfactual audit using matched probe cards—Gold (carrying the correct answer label), Poison (carrying a target wrong label), and Benign (filler only)—under six-layout position
symmetry, isolating the sparsification-specific effect. Two patterns compete. Signal concentration: the selector preserves Gold and Poison blocks far above filler-matched Benign blocks (G$\approx$P$\gg$B across all model–task pairs). Integration loss: discarding blocks severs cross-block
attention—confirmed by an ablation where isolating the probe block collapses its influence from 4.48 logits to zero. Compression ratio governs the balance: a full sweep from mild ($c=0.25$) to aggressive ($c=0.75$) compression across four model–task pairs reveals that three of four
cells move toward stronger sparse amplification at higher compression, with two exhibiting sign reversals. Three independent arms—BSFA route replay, controlled block-top-$k$, and KV-cache eviction—converge: sparsification changes content influence in ways aggregate accuracy cannot detect. We provide an open measurement framework deployable on any model exposing block identities.
Read the full abstract ▾
All authors Xingyu Ren, Youran Sun, Chugang Yi, Haizhao Yang
Machine Learning · Cross-listed from cs.CL
Learning What to Remember: Test-Time Training via Context Distillation
Zixuan Wang, Xingyu Dang, Rui-Jie Zhu et al. (+4 more)
Effective long-context modeling is not merely about retaining more of the past, but about preserving the information that may prove relevant later. Test-time training (TTT) is an appealing approach that performs online parameter updates for long-context modeling, yet existing TTT methods only optimize either reconstruction or online adaptation objectives without considering the future utility of retained information. In this work, we propose Test-Time Context
Distillation (TTCD), a TTT framework that introduces a self-supervised objective for allocating limited
memory capacity for future use. Specifically, TTCD uses a long-window teacher to supervise the fast weights of a short-window student, where the hidden-state discrepancy between them offers a dense, self-supervised signal guiding the model to memorize the contextual information crucial for future
token predictions. We focus on an in-place variant: In-Place TTCD (IP-TTCD), which uses the existing MLP parameters as the fast weights. Experiments on long-context
language modeling tasks show IP-TTCD consistently outperforms DeltaNet, Gated DeltaNet, sliding-window
attention, and TTT when pre-trained from scratch. Furthermore, IP-TTCD allows pre-trained
transformer models to adapt their parameters during inference through continual pre-training, gaining long-context capabilities with only a lightweight architectural augmentation. Our results position TTCD as a step toward architectural continual learning.
Read the full abstract ▾
All authors Zixuan Wang, Xingyu Dang, Rui-Jie Zhu, Zixin Wen, Hengyu Fu, Wenhao Chai, Jason D. Lee
Machine Learning
Sharp Root Anti-Concentration via Projective Incidence and Ordered Root Laws
Zijun Wang, Yuchen Miao, Yifan Hu et al. (+1 more)
This paper answers the one-dimensional local root anti-concentration questions posed by Balcan, Pegden, and Sharma in the context of online optimization of piecewise-Lipschitz functions. For a homogeneous feature curve and coefficients whose density relative to the uniform law on a symmetric convex body $K$ is bounded by $A$, we show that the worst-case interval-hitting constant equals $A$ times a section-averaged projective incidence speed. For cube-supported coefficients, this speed is equivalent, up to universal constants, to the projective Lipschitz constant. This yields a sharp, dimension-free characterization and removes the previous $\sqrt N$ loss. For monic degree-$d$
polynomials under arbitrary coefficient laws, we prove that the interval-hitting constant is finite if and only if the ordered real-root laws have bounded densities, with a factor-$d$ comparison that is sharp. Conditional and joint coefficient-space area formulas, together with a two-chart certificate, make this criterion verifiable for dependent and singular coefficient laws. We also give two graph-learning applications that complete the transition-to-regret chain. A cost-sensitive Gaussian-RBF harmonic classifier uses the projective incidence
theorem and achieves expected regret $\widetilde O((An^2D e^{BD}/\ell+1)\sqrt T)$. A common-offset
polynomial-kernel model uses rigid translation of the ordered roots and achieves $\widetilde O((qn^2\kappa+1)\sqrt T)$ regret, even when the induced coefficient law is singular in the ambient coefficient space.
Read the full abstract ▾
All authors Zijun Wang, Yuchen Miao, Yifan Hu, Huanmin Liu
Machine Learning · Cross-listed from cs.CV
FAU at ImageCLEF 2026 Task on Multimodal Reasoning Robust Candidate Scoring and Concise Multilingual Visual Answering
Mohamed Basem and Vincent Christlein
We present our ImageCLEF 2026
Multimodal Reasoning system for the Visual Multiple Choice Question Answering (Visual MCQ) and Visual Open Question Answering (Visual OpenQA) subtasks. The challenge requires reliable
reasoning over
multilingual educational and scientific images with dense text, diagrams, charts, tables, formulas, and units, while enforcing strict answer formats. Our central finding is that robust output control is as important as model choice. For Visual MCQ, we replace fragile free-form generation with direct candidate label scoring from
vision-
language model logits, then combine complementary runs through score fusion and voting. For Visual OpenQA, we use image enhancement, concise final answer prompting, deterministic decoding, and targeted post-processing to remove
reasoning traces and formatting artifacts. Without task-specific model training, our official submissions achieved third place in Visual MCQ with 0.7108 accuracy and first place in Visual OpenQA with 0.6488 COMET, 0.1391 BLEU, 0.2762 ROUGE L, and 0.2383 METEOR. The results highlight the practical value of inference engineering: careful scoring, ensembling, prompting, and cleanup can turn strong VLMs into reliable competition systems.
Read the full abstract ▾
Machine Learning · Cross-listed from cs.AI
LongCat Sparse Attention: Taming the Lightning via Streaming-aware Hierarchical Cross-Layer Indexing
Wen Zan, Jiaqi Zhang, Jianchao Tan et al. (+11 more)
DeepSeek Sparse
Attention (DSA) enables efficient long-context modeling through its Lightning Indexer. However, practical deployment remains constrained by the indexer's expensive $O(L^2)$ scoring overhead and the hardware-inefficient, discontinuous
memory-access patterns induced by its outputs. To address these system-level bottlenecks, we introduce LongCat Sparse
Attention (LSA), a hardware-
algorithm co-designed framework comprising three complementary and orthogonal strategies: (1) Streaming-Aware Indexing, which selectively converts scattered KV entries into hardware-aligned contiguous layouts to enable coalesced HBM access; (2) Cross-Layer Indexing, which amortizes indexing overhead by reusing the results produced by a single layer across consecutive layers, supported by cross-layer
distillation; and (3) Hierarchical Indexing, which adopts a coarse-to-fine scoring scheme to progressively narrow the candidate set for each query, thereby substantially reducing indexing computation. Extensive scaling experiments, ranging from 69B-A3B to 560B-A27B models, demonstrate that LSA consistently achieves performance on par with full
attention across both general-purpose and long-context
benchmarks. Moreover, LSA supports native training with context lengths of up to one million
tokens and underpins the development of LongCat-2.0 (1.6T-A48B). To facilitate further research, we also introduce and open-source LongCat-Flash-Lite-Sparse (69B-A3B), which integrates LSA into LongCat-Flash-Lite and incorporates an updated long-context training corpus.
Read the full abstract ▾
All authors Wen Zan, Jiaqi Zhang, Jianchao Tan, Hong Liu, Cunguang Wang, Xiang Li, Duyue Ma, Guanyu Wu, Yifan Lu, Fengcun Li, Yerui Sun, Peng Pei, Yuchen Xie, Xunliang Cai
Machine Learning · Cross-listed from math.OC
Non-KKT Accumulation in Entropic Mirror Descent
Kuangyu Ding and Kim-Chuan Toh
For mirror descent generated by a Legendre kernel, perhaps one of the most basic question in optimization is this: must every accumulation point of a bounded mirror descent sequence be Karush–Kuhn–Tucker (KKT) stationary under proper stepsizes? We show that the answer is no. A longstanding obstacle to resolving this question is the boundary blow-up of the Legendre gradient: it keeps every mirror step in the interior, while at a boundary limit, the inverse
entropy metric vanishes on active coordinates and can erase the dual-feasibility in the KKT system. We construct $C^\infty$ objectives and bounded sequences generated by the Shannon-entropic mirror descent on the nonnegative orthant $\R_+^n$, for every $n\geq 3$, and on the probability simplex $\Delta_n$, for every $n\geq 4$, such that, in each case, the set of accumulation points is a smooth boundary circle containing a nonempty relatively open arc of non-KKT points. The steps satisfy $\alpha_k\asymp k^{-\beta}$ with $\beta\in(1/2,1)$, the objective values are nonincreasing, and the objectives are
entropy-relatively smooth. Hence the pathology stems from the degeneracy of the Bregman geometry at the boundary, rather than from failure of descent, or improper stepsizes. To the best of our knowledge, these provide the first counterexamples to KKT accumulation for bounded mirror descent sequences with nonincreasing objective values.
Read the full abstract ▾
Machine Learning · Cross-listed from cs.DC
Bole: Efficient Tree Speculation for Hybrid-Attention Language Models
Li Wang, Yi Su, Xiabao Wu et al. (+9 more)
Hybrid-
attention large
language models combine full
attention with recurrent linear
attention to reduce long-context inference costs, yet their autoregressive decoding remains
memory-bound. Tree speculative decoding offers an attractive acceleration path, but existing tree-speculation systems are designed around the key–value caches of full-
attention models. On hybrid models, they traverse recurrent layers branch by branch and materialize a full state for every proposal node, causing verification latency and transient
memory to scale poorly with tree and batch sizes. We present Bole, a kernel–runtime co-design that enables efficient tree speculation for hybrid-
attention LLMs. Bole transforms the linear-
attention recurrence into a tree-structured closed form and realizes it with a resource-efficient GPU kernel, verifying all proposal nodes in parallel and accelerating linear-
attention tree verification by 3.4–7.7$\times$. It losslessly encodes speculative state updates as
token-level factors and reconstructs only the state selected after sampling, reducing transient state
memory by 82–99$\times$ and freeing GPU capacity for KV caches. Its integration into SGLang, a widely deployed production LLM serving engine, couples efficient state management with a batch-wide verification budget calibrated to the complete hybrid forward. Across four models, two GPU platforms, and diverse
datasets, Bole delivers up to $4.72\times$ the offline decode throughput of autoregressive decoding and up to $2.03\times$ that of the strongest tree-speculative baseline. Under online
agent workloads, it reduces TTFT and TPOT by up to $67.6%$ and $49.9%$, respectively, over the strongest tree-speculative baseline.
Read the full abstract ▾
All authors Li Wang, Yi Su, Xiabao Wu, Chiran You, Yongchao Liu, Zhan Qiu, Juelu Zhang, Jiajun Zheng, Fangxin Liu, Jie Zhang, Chen Tian, Chengying Huan
Machine Learning
Evaluating Forecasting Techniques for Hardware Errors on a Large-scale HPC System
Kaiyuan Liao, Xiwei Xuan, Tanwi Mallick et al. (+3 more)
Hardware error logs in high-performance computing (HPC) systems provide early signals of abnormal behavior, yet there remain challenges in effectively
forecasting these errors using modern predictive methods. This work investigates the boundaries of applying time series
forecasting to HPC hardware error dynamics. We use seven years of production logs from the Theta supercomputer to evaluate the predictive efficacy of classical statistical and deep learning models. Our results show that
forecasting effectiveness depends strongly on the temporal structure of the error series: regularly occurring and structurally stable errors can be modeled accurately, particularly by LSTM and
Transformer architectures with temporal features, while sparse and burst-dominated errors remain difficult to predict. Rather than proposing a deployment-ready failure prediction framework, this study provides empirical guidance on when
forecasting is effective and highlights potential directions for improving
forecasting accuracy in HPC hardware error analysis.
Read the full abstract ▾
All authors Kaiyuan Liao, Xiwei Xuan, Tanwi Mallick, Kevin Brown, Christopher D. Carothers, Kwan-Liu Ma
Machine Learning
GraphIR: Architecture-Level Search States for LLM-Guided Neural Architecture Evolution
Zhen Liu, Wanqi Zhou, Shuanghao Bai et al. (+3 more)
Large
language models (LLMs) enable neural architecture search (NAS) directly over executable neural network programs. However, code-level flexibility does not provide the architecture state needed for effective mutation: LLMs must infer tensor dependencies, editable components, and compatibility constraints from implementation details. To address this representation mismatch, we propose GraphIR, an architecture-aware intermediate representation that supplements executable programs with a mutation-aligned candidate state. GraphIR organizes each candidate through three complementary views: a computation skeleton describing tensor flow, a mutation surface exposing editable modules and operations, and a validity envelope capturing interface contracts, propagated shapes, and downstream dependencies. To evaluate our method, we construct NAS-Dependency, a 120-question
benchmark covering six complementary dependency-
reasoning dimensions. The diagnostic shows that GraphIR is particularly effective at identifying exact producer occurrences, tracing dependency propagation, and diagnosing interface and failure risks. Across six downstream
benchmarks including CLRS, GraphIR achieves the best overall search performance while maintaining comparable model size and favorable end-to-end NAS efficiency when integrated into OpenEvolve. These results show that a mutation-oriented architecture state provides an effective interface between executable neural programs and LLM-guided architecture
evolution.
Read the full abstract ▾
All authors Zhen Liu, Wanqi Zhou, Shuanghao Bai, Yuhan Liu, Jinjun Wang, Jingwen Fu
Machine Learning
QWRF-Net: A Quantum-Wavelet Framework with Rectified Flow for Short-Term Precipitation Nowcasting
Zhuo Wang, Chaorong Li, Wenjie Luo et al. (+1 more)
Short-term precipitation nowcasting is important for hydrometeorological early warning, especially when intense convective rainfall may trigger urban flooding, flash floods, and other high-impact hazards. A key challenge in warning-oriented nowcasting is that radar precipitation fields contain strongly coupled multi-scale structures, while forecast quality often degrades at later lead times, making it difficult to preserve intense precipitation cores and their spatial organization over the full warning-relevant horizon. To address this problem, we propose QWRF-Net, a
quantum-wavelet framework with rectified flow for short-term precipitation nowcasting. The core idea is to improve the conditional representation of precipitation by explicitly decomposing latent features into wavelet sub-bands and then performing differentiated
quantum-inspired modulation in the decomposed latent space, before generating future sequences through a rectified-flow-based non-autoregressive decoder. Experiments on the KNMI radar and SEVIR
benchmarks under a unified evaluation protocol show that QWRF-Net achieves favorable overall performance, with relatively consistent gains at medium-to-high precipitation thresholds, on an extreme-event subset, and in preserving intense precipitation cores and fine-scale structures. Ablation results further indicate that wavelet-based scale disentanglement, differentiated sub-band modulation, and flow-based generation provide complementary benefits within the proposed framework. Overall, these results suggest that jointly enhancing multi-scale precipitation representation and stable multi-step generation is a promising direction for warning-oriented short-term precipitation nowcasting. The observed improvements may also provide a more useful precipitation basis for downstream hydrological and warning-related applications.
Read the full abstract ▾
All authors Zhuo Wang, Chaorong Li, Wenjie Luo, Chuanhu Deng