Hyperbolic Prototype Routing (HyPro) is a rehearsal‑free framework for class‑incremental learning that assigns a dedicated LoRA‑Expert module to each incremental task, enabling isolated representation learning. It projects routing features onto a Poincaré ball and uses geodesic nearest‑prototype matching for reliable task‑level discrimination. Experiments on standard and few‑shot CIL benchmarks demonstrate that HyPro consistently improves both average and final‑stage accuracy compared to strong baselines.
By HongWei Zhao (Beihang University), Rui Liu (Beihang University), Yong Chen (Beijing University of Posts,Telecommunications)
The paper introduces two hybrid imputation methods, NuclearForest and SoftForest, that combine a low‑rank initialization (using Singular Value Thresholding or SoftImpute) with a single Random Forest refinement. These methods capture global covariance patterns and local nonlinear dependencies, respectively, and are benchmarked against seven existing techniques across MCAR, MAR, and MNAR scenarios. Results show comparable or superior imputation accuracy to state‑of‑the‑art iterative methods while achieving significant speedups of roughly 5.8× and 9.5× over MissForest.
By Jinwei Li, Michelle Bruch, Daniel Tenbrinck
The paper extends the abstract gradient training (AGT) framework to provide tighter differential privacy guarantees for both private prediction and private learning. It introduces Abstract Gradient Sampling (AGS) to analyze privacy in continuous, unbounded regression and offers theoretical and empirical evidence that these methods yield tighter bounds than global-sensitivity baselines, even in previously unbounded settings. The authors also demonstrate that their private learning algorithm can outperform standard private learners under comparable conditions.
By Mihnea Ghitu, Matthew Wicker
The paper investigates test‑time scaling for language models, noting that sequential methods often stall due to entering attractors—sets of answers that block further exploration. Across 27 experiments, over half of sequential trajectories hit an attractor within four iterations. Introducing a simple model‑mixing intervention reduces attractor incidence by 21.2 percentage points, expands solution coverage beyond a compute‑matched parallel baseline, and boosts recursive self‑aggregation accuracy by at least 2.2 percentage points.
By Joseph Rance, Fabio Pizzati, Juil Sock, Woody Bayliss, Marc G\'orriz Blanch, Philip Torr, Adel Bibi
The Nixtlaverse is an open‑source Python ecosystem that unifies statistical, machine‑learning, and neural forecasting models by sharing a common long‑format panel data structure and keyed forecast outputs while preserving each model family’s specialized implementation. Using the M5 competition data, the authors demonstrate the framework’s ability to evaluate diverse models in a single rolling‑origin setup, profile runtime and memory usage across thousands of series, and reconcile forecasts from multiple engines, including an external one, over the full M5 hierarchy. The paper highlights the design’s cost, boundary, and utility advantages and notes its growing adoption and permissive licensing.
By Olivier Sprangers, Max Mergenthaler Canseco, Marco Peixeiro, Saul Caballero Ramirez, Mariana Menchero Garc\'ia, Jing-Qiang Goh, Han Wang, Nikhil Gupta, Rogelio Melo, Senbong Gee, Cristian Challu
The paper introduces CG-OMatG, an equivariant Riemannian flow-based generative model that predicts molecular crystal structures using a coarse-grained, hierarchical representation. It treats molecules as rigid bodies, performs inter- and intra-molecular message passing, and learns to reconstruct molecule centroids, orientations, and lattice parameters conditioned on chemical species and conformer geometry. The model is trained on OMC25 and CSD datasets, fine-tuned with policy gradient reinforcement learning to favor low-energy structures, and validated against a CSP blind test benchmark using COMPACK packing-similarity analysis.
By Thomas Egg, Harry Winston Sullivan, Maya M. Martirossyan, Philipp H\"ollmer, Cheng Zeng, Adrian Roitberg, Mingjie Liu, Richard Hennig, Sapna Sarupria, Ellad B. Tadmor, Stefano Martiniani
PMosFM introduces a preconditioned manifold matching framework that enables one‑step physics‑constrained generation by encoding constraints in a manifold decoder. The method learns transport in intrinsic coordinates, eliminating the need for residual losses or trajectory unrolling, and employs a geometric preconditioner and covariance transform to improve conditioning. Experiments demonstrate reduced training and sampling time compared to multi‑step baselines while maintaining comparable physical and distributional fidelity.
By Zhangyong Liang, Haibin Ling
The study shows that previously reported gains in decoding spoken words from non‑invasive brain recordings can largely be replicated without any brain data, due to a timing shortcut where overlapping windows reveal word durations. By processing each window independently rather than jointly, the authors eliminate this shortcut, enabling the neural network to learn genuine brain‑based word information. This change improves performance and makes two existing strategies—aggregating predictions from multiple neural responses and incorporating a pretrained language model—significantly more effective, achieving a word error rate of 36.6% on a perceived speech benchmark.
By Dulhan Jayalath, Oiwi Parker Jones
The paper introduces VirusCascade, a black‑box attack that exploits the collaborative reflection mechanism in LLM‑powered recommender agents. By leveraging reflective persistence and cross‑agent propagation, the attack promotes a target item through semantic rationalisation and structural positioning, achieving state‑of‑the‑art targeted exposure on four real‑world datasets. The study demonstrates that existing static‑pipeline attacks cannot exploit this recurrent, multi‑agent amplification pathway.
By Yurong Hao, Wen Zhou, Guowei Guan, Tiantong Wu, Fuyao Zhang, Wei Yang Bryan Lim
The paper presents C2C (From Codebase to Culprit), a bug‑localization framework that narrows the debugging search space through semantic retrieval and hierarchical reinforcement learning. It first retrieves likely buggy files using contrastively fine‑tuned CodeBERT embeddings and then progressively localizes bugs from files to functions to lines of code. Experiments on Java and Python datasets show that this multi‑resolution approach improves both retrieval precision and localization accuracy, with ablation studies underscoring the importance of hierarchical decomposition and reward shaping.
By Ankur Garg, Corey Yang-Smith, Rishav Rishav, Ahmad Abdellatif, Samira Ebrahimi Kahou
Diffusion-2BC is a hybrid training method that combines a diffusion denoising objective with an auxiliary deterministic behavior‑cloning loss over a shared visual encoder. The auxiliary loss is used only during training, while inference remains diffusion‑based. Experiments on the Claw environment and CARLA navigation show that Diffusion‑2BC reduces mean mask‑distance error by about 10% compared to a diffusion baseline and by 85% compared to standard deterministic behavior cloning, and it enables the agent to travel farther and exhibit multimodal route choices.
By Bruno Maciel Machado, Eric Aislan Antonelo
SHIFT‑Truck is the first high‑fidelity aerodynamics dataset and benchmark specifically for pickup trucks, comprising 1,000 Spalart‑Allmaras delayed detached‑eddy simulations of a reference truck geometry morphed across 17 shape parameters. Each simulation uses a ~100‑million‑cell mesh at a Reynolds number of 1.4×10⁷ and provides time‑averaged surface pressure, wall shear stress, volumetric pressure, and velocity data, verified by grid refinement, repeated runs, and wind‑tunnel measurements. The benchmark defines geometry‑grouped splits and evaluates four neural surrogates—DoMINO, GeoTransolver, AB‑UPT, and SMART—on surface and volume tracks, while also introducing controlled distribution shifts in operating point, surface discretization, and vehicle archetype to test generalization.
By Riddhiman Raut, Yin Yu, Aashwin Anand Mishra, Michael Emory, Thomas Economon, Peter Lyu, Juan J. Alonso
The paper introduces the Posterior Concentration Phenomenon (PCP), a length‑dependent failure mode where probability‑based rewards collapse to a narrow interval for long reasoning traces, destabilizing verifier‑free reinforcement learning. To address this, the authors propose RLCPR, a framework that uses uncertainty‑aware data sampling and concentration‑aware regularization to mitigate PCP, improving token efficiency and outperforming state‑of‑the‑art baselines on multiple reasoning benchmarks.
By Shiu-Hong Kao, Yubo Zhao, Zhenyu Tian, Pengzhan Sun, Yicong Li, Angela Yao
The paper introduces the Generalized Geometry Block Proximal Linearized (GGBPL) method, which uses a generalized geometry proximal linearized operator to update block variables in multiblock nonconvex and nonsmooth optimization problems. By allowing arbitrary inner products and admissible metrics, GGBPL adapts to the geometric structure of various problems, improving numerical efficiency over traditional Euclidean‑based methods. An inertial variant (iGGBPL) is also presented, and a unified convergence framework is established, proving objective convergence, global convergence to a critical point, and an ε‑stationary point iteration complexity of φ(ε−−). The methods are applied to sparse nonnegative matrix factorization and sparse nonnegative CP decomposition with λ₀ constraints, showing superior performance in numerical experiments.
By Weifeng Yang
ThunderEP is a new communication design for expert parallelism on PCIe-connected consumer GPUs that eliminates relay hops, uses DMA engines to avoid GPU compute contention, and reduces CPU polling overhead. Integrated into vLLM, it outperforms NCCL on RTX 4090 and RTX 5090 GPUs, delivering average speedups of 2.00× for dispatch, 1.53× for combine, and up to 1.66× end‑to‑end over existing MoE inference frameworks.
By Jaehwan Lee, Sangmin Lee, Chaewon Kim, Junsik Shin, Jaejin Lee
The paper investigates budget pacing in online advertising, where an advertiser participates in a sequence of second-price auctions with time-varying value and bid distributions. It asks how many historical samples are needed to learn an effective expenditure plan that maximizes utility under a budget constraint. The authors prove that a single sample per distribution suffices to achieve near‑optimal ∫O(√T) regret, improving on the previous T log T requirement and maintaining robustness to noisy sampling.
By Santiago Balseiro, Rachitesh Kumar, Vahab Mirrokni, Balasubramanian Sivan, Di Wang
The paper introduces SEAR, a four-task benchmark that evaluates audio language models (ALMs) for audio deepfake detection by requiring acoustic evidence identification, quantification, deepfake detection, and forensic rationale generation. It also proposes BAEA, an acoustic evidence agent that equips a frozen ALM with controlled acoustic tools under fixed or adaptive evidence-acquisition policies. Experiments with six ALMs show a gap between plausible rationales and verifiable acoustic evidence reasoning, and demonstrate that BAEA‑Fixed improves final verdicts and forensic rationales while revealing that misleading evidence harms detection and grounding performance.
By Rong Wan, Suliu Qin, Jiaxi Li, Wei Xie, Wenwu Wang, Xiaolong Han, Lu Yin, Xilu Wang
The paper introduces a reference‑free diagnostic for physics‑informed neural networks that learn partial differential equation coefficients from noisy data. It presents three tests—an information‑matrix residual test, a heterogeneity statistic across window refits, and a Fisher‑rank statistic on the design—that can detect whether the operator used in the model is incorrect or whether the design is blind to certain rates, all without retraining the network. The authors demonstrate the method on a parabolic benchmark with 2 % noise, showing that the diagnostics reliably flag misspecified operators and blind designs while remaining silent when the model is correctly specified.
By Eric Fock
The study shows that electronic structure datasets used for machine learning contain significant redundancy due to low intrinsic dimensionality. Random pruning and a coverage-based strategy can reduce dataset size by up to two orders of magnitude while preserving chemical accuracy and generalizability, cutting training time by a factor of three or more. The essential information lies on a low‑dimensional, non‑linear manifold, suggesting that large datasets may be largely overlapping and that minimal, representative datasets could suffice for accurate predictions.
By Sazzad Hossain, Ponkrshnan Thiagarajan, Shashank Pathrudkar, Stephanie Taylor, Abhijeet S. Gangan, Amartya S. Banerjee, Susanta Ghosh
The paper introduces RateAR, a benchmark of AR images and videos that captures perceptual factors such as object placement, scale, and shadow consistency. Eleven commercial vision‑language models (VLMs) are evaluated on this benchmark, with VLM‑based quality predictions showing strong correlation with human judgments (Spearman’s ρ up to 0.8695). Using contextual prompting, the authors build an automated AR content adjustment system that, in a 21‑participant study, improved placement and size coherence of virtual content for over 90% of users.
By Elias Rotondo (Duke University), Lin Duan (Duke University), Yanming Xiu (Duke University), Sangjun Eom (Duke University), Conrad Li (Duke University), Maria Gorlatova (Duke University)