DeFM is a self‑supervised foundation model trained solely on depth images for robotics. It uses a DINO‑style self‑distillation objective on 60 M curated depth images to learn geometric and semantic representations that generalize across environments, tasks, and sensors. The model incorporates a novel input normalization strategy for metric awareness, can be distilled into compact versions for resource‑constrained systems, and achieves state‑of‑the‑art performance on depth‑based classification, segmentation, navigation, locomotion, and manipulation benchmarks, with strong sim‑to‑real transfer.
By Manthan Patel, Jonas Frey, Mayank Mittal, Fan Yang, Alexander Hansson, Amir Bar, Cesar Cadena, Marco Hutter
BLOKS is a Conjugate Gradient-based solver for Nyström‑sketched Kernel Ridge Regression that introduces a novel preconditioner combining block Jacobi approximation with sketching. This preconditioner’s memory cost is independent of dataset size and model parameters, enabling scaling to large models while maintaining GPU acceleration. Experiments on standard large‑scale datasets show BLOKS achieves accuracy comparable to or better than state‑of‑the‑art kernel solvers with favorable time/memory complexity.
By Giordano Vitale, Giacomo Meanti, Marco Rando, Blanca Ayuso de Dios, Lorenzo Rosasco
FedPCD introduces a priority-constrained federated optimization framework that decouples the global aggregation objective from client protection. By preserving the standard federated aggregate and applying only minimal corrections to satisfy client performance constraints, FedPCD achieves higher global accuracy and improved worst‑10% client accuracy across heterogeneous benchmarks. The method also maintains competitive trade‑offs between global accuracy and client variability while converging faster with fewer communication rounds.
By Yousef Irshaid, Dara Varam, Mohamed I. AlHajri
arXiv:2610. 13631v1 Announce Type: new Abstract: Fault diagnosis in asset-intensive industries requires multiclass models trained on noisy, sensitive data collected across geographically distributed client sites.
By Andrea P. Hernandez, Michael Ibrahim, Nagi Gebraeel
TabReliBench evaluates tabular foundation models (TFMs) on reliability metrics beyond predictive accuracy, including point prediction, distributional quality, marginal and subgroup calibration, and interval efficiency. The benchmark reveals that TFMs’ performance rankings differ from those in TabArena and vary across reliability measures. While regression TFMs often achieve near‑nominal marginal coverage with narrow intervals, this does not guarantee valid conditional coverage, and ensembling or conformal post‑processing does not consistently improve calibration. The study underscores the need for multidimensional evaluation when selecting reliable TFMs.
By Tianqi Zhao, Jia Guo, Yan Shuo Tan, Qiong Zhang
The paper argues that neural networks can exhibit modular organization even when individual neurons are not specialized. It introduces a nested hierarchy of modularity—unit-level, orthogonal, and linear—and shows that many vision, language, and RNN models lack unit-level modularity but still display clear functional modularity at the subspace level. The authors demonstrate that these subspaces can be causally validated through ablation and align with brain data, revealing distinct content/style or syntax/semantics separations.
By Shreya Saha, Benjamin Bergen, Meenakshi Khosla
LongHarness is a model‑agnostic harness that explicitly manages the interaction between acquired evidence and evolving conclusions for long‑context reasoning over corpora larger than a model’s context window. Its Evidence‑Conclusion Coevolution mechanism uses unresolved conclusions to steer evidence retrieval, while new evidence refines those conclusions, and its Evidence Lifecycle Management keeps evidence current and traceable. Evaluations on OfficeQA Pro, AA‑LCR, and LongBench v2 show substantial performance gains over direct reasoning and state‑of‑the‑art agent harnesses, with GLM‑5.3‑Flash exact match improving from 15.8% to 60.2% on V1 and from 4.4% to 61.1% on V2.
By Songtao Huang, Zhijie Zhong, Xiaohan He, Shuaiyu Zhang, Zhiyin Yu, Li Kang, Heng Zhou, Yutao Fan, Xiangchao Yan, Bo Zhang, Shiyang Feng
GraphDecisionBench is a benchmark designed to evaluate System One models—such as Jev—that make decisions directly from supplied options on graph-related tasks. The benchmark is model‑independent and incorporates structural task profiles, matched graph‑text input contrasts, and heuristic‑proposal controls to diagnose graph decision performance. Evaluations across fourteen model‑interface configurations reveal that accurate adjacency recognition does not guarantee broader structural correctness, joint graph‑text input does not consistently improve prediction, and feasible construction does not ensure high solution quality.
By Xianliang Yang, Yapu Zhang, Li Zhao
The paper examines the commonly used structure‑disjoint split for evaluating molecular machine learning models, which relies on single‑linkage clustering with the Maximum Common Edge Subgraph (MCES) distance at a threshold of 10. The authors show that this criterion, originally calibrated on 29,000 molecules, has never been tested on the current public libraries of 227,690 molecules due to its NP‑hard complexity. They propose an optimized algorithm that reduces the number of required integer programs from $2.6 imes10^{10}$ to $1.15 imes10^{7}$, enabling lossless partitioning at scale, and reveal that the resulting graph percolates—over 95 % of the library lies in a single component, making held‑out compounds systematically larger and more complex than training ones.
"whyItMatters":"The study highlights that large‑library benchmarks using the traditional split conflate structural novelty with extrapolation to larger, more complex chemistry, potentially misleading assessments of model generalization."
By Tomo Oga, Devesh Shah, Cailum M. K. Stienstra, Gabriel Asher, Antonio H. de O. Fonseca, Niall O'Connor, Michael Widrich
CAST (Chunk-aware credit Assignment with Selective Token Anchoring) is a new framework for cross-tokenizer on-policy distillation that tackles two key problems: credit alignment and unaligned token positions. It redistributes chunk-level advantage signals across student tokens using predictive uncertainty and teacher-signal smoothness, while anchoring unaligned tokens to the initial policy via a KL penalty. Experiments on six mathematics, science, and code benchmarks show that CAST improves average performance over the strongest baseline by up to 2.45 points in cross-tokenizer transfer scenarios.
By Jingru Yang, Jinhang Wu, Minjing Shi, Shengfeng He, Xuyan Li, Man Tan, Biye Li
EarLog is a real‑time, embedded system that opportunistically uses headphone playback as a probe to detect the acoustic reflex (AR), an involuntary ear response indicating protective activation during high‑load listening. By employing content‑conditioned residual sensing, a resonance‑constrained sparse FIR model, and dual‑microphone observations, EarLog suppresses noise from nonstationary content, individual ear acoustics, and device variations to identify weak AR signals. In a study with 81 participants, 16 headphone models, and over 400 hours of recordings, EarLog achieved 99.0% sensitivity and 98.7% specificity against a clinically referenced benchmark, enabling continuous, listener‑specific AR monitoring during everyday listening.
By Yongzhi Huang, Jiayi Zhao, Kaishun Wu
Neurograf v1 is an experimental end‑to‑end Romanian‑first language‑model pipeline that includes a 635 GB Romanian corpus (RoSfera), an 80k byte‑level BPE tokenizer, and five decoder‑only models ranging from 160 M to 6 B parameters trained on 131 B tokens. The pipeline also produces three instruction‑tuned assistants and two contrastively trained embedding models, all trained on ELI‑NP/IFIN‑HH infrastructure. Evaluation shows consistent scaling of validation loss and improved Romanian factual and cultural knowledge with larger models, achieving 69.89 % accuracy on the 1,000‑question WWTBM benchmark for the 6 B model.
"whyItMatters":"The study demonstrates that it is feasible to train Romanian‑first language models from scratch, highlighting practical challenges such as diacritic preservation and tokenization efficiency while providing a unified training pipeline for base, instruction‑following, and embedding models."
By Ionut Adrian Rotaru
The paper introduces UkrMeme, a dataset of 200 Ukrainian wartime memes annotated for persuasive techniques, and PTM-UA, a Ukrainian translation of a subset of the Propaganda 950 benchmark that retains the original visual content. The authors evaluate vision‑language models in a zero‑shot prompting setup across image, text, and multimodal inputs using hierarchical F1 metrics. They also compare Layer‑wise Relevance Propagation and model‑generated token attributions to analyze the semantic categories of evidence emphasized by each explanation method.
By Nazar Andrushko, Vera Schmitt, Veronika Solopova
The paper introduces Kernelized Parameter Sensitivity Atlases (KPSA), a method that uses a block decomposition of the empirical neural tangent kernel to estimate input‑specific parameter sensitivities efficiently. KPSA embeds parameter groups in a query space and applies a lightweight kernel to predict sensitivities for unseen inputs, achieving a high fraction of the perturbation susceptibility of full back‑propagation while dramatically reducing computation and memory usage. Experiments across vision, forecasting, and protein‑modeling tasks show that KPSA identifies sparse, depth‑spanning sensitivity circuits and improves throughput and memory efficiency in activation‑attribution problems.
By Davide Wiest, Ahanaf Hasan Ariq, Abdul Monaf Chowdhury
CoGameReg is a cooperative‑game‑based semi‑supervised regression framework that fuses heterogeneous regressors using contribution‑weighted predictions to generate shared pseudo‑labels for seismic reservoir property estimation. It employs a Tri‑Game Shapley mechanism to assign non‑negative weights from labeled data, keeps these weights fixed during collaborative training, and screens candidate pseudo‑labels with a local k‑nearest‑neighbor error‑reduction criterion before retraining. Evaluations on ten public datasets, a synthetic 3‑D P‑wave velocity task, and a field water‑saturation application show that CoGameReg achieves superior blind‑well metrics, especially at low labeled‑data ratios, with the highest Pearson correlation among compared field methods.
By Zhiyu Guo, Jiwei Jia, Lei Li
The paper introduces Data-Adaptive Mahalanobis Cross-Head Attention (DAMCHA), a new attention mechanism that extends Multi-Head Attention by learning a full cross-head metric matrix and adapting it to input data. DAMCHA captures inter-head interactions beyond the block-diagonal structure of standard MHA and uses input-dependent parameterization for dynamic attention geometry. The authors provide regularization techniques, such as stack-wise parameter sharing, to maintain efficiency and stability, and report that DAMCHA-based Transformers outperform baselines on several benchmarks.
By Zhenke Duan, Xin Li, Jiqun Pan, Dong Xiaofei, Hanwen Ning, Xinyuan Song
PlacePulse is a structural geospatial foundation model that transforms 572 attributes—expanded to 2,288 through H3 neighborhood aggregates—into 256‑dimensional unit‑norm embeddings for over a million H3 resolution‑7 cells across the United States. The model achieves high intrinsic reconstruction quality (mean $R^2 = 0.853$) and strong spatial organization (Moran's $I = 0.93$), while outperforming a dimensionality‑matched baseline on 74 held‑out targets from 12 external sources and competing with Google’s Population Dynamics Foundation Model on various health, economic, election, mobility, and social‑vulnerability outcomes. These results demonstrate that static structural composition can serve as a distinct and complementary modality for geospatial foundation models.
By Luc\'ia Garc\'ia-Duarte, Miguel \'Alvarez, Diego Vicente, Javier P\'erez Trufero, Javier de la Torre
BLEND is a framework for federated vision‑language models that balances personalization and generalization. It uses selective fine‑tuning with global vision and text adapters shared across clients and a personalized vision projection adapter for each client. A novel loss function blends outputs from global and personalized adapters, while an anchor design in the projection heads promotes generalization, leading to superior performance over existing baselines.
By Payam Abdisarabshali, Fardis Nadimi, Kasra Borazjani, Naji Khosravan, Richeng Jin, Seyyedali Hosseinalipour
JitterVLA is a replay benchmark designed to evaluate motion quality in vision‑language‑action (VLA) driven human‑robot manipulation. It replays identical command and arrival traces through different controllers to isolate controller effects, and is applied to synthetic reaches, hand trajectories from 20 people, three open VLA pipelines (838 episodes), and a physical UR3 robot. The benchmark demonstrates that interpolation reduces modeled p99 jerk by 83% for human motion with a 200 ms delay, while jerk‑limited generation further lowers jerk at a higher cost, and shows varying effects of entropy‑based techniques across pipelines.
By Faith Ogunwoye, Didar Seyidov, Konstantin Gubernatorov, Dzmitry Tsetserukou
VLA-FAIL introduces a lightweight framework for detecting failures in vision-language-action models used in robotic manipulation. It combines two novel detectors: the first-layer Mahalanobis distance (LLMD) identifies out-of-distribution states by comparing token-wise deviations in last-layer features to training data, while the second, action chunk consistency (ACC), detects failures by spotting inconsistencies between consecutive action chunks in receding-horizon control. The authors also propose a threshold‑independent metric, “AUCPDT,” to evaluate detection accuracy and latency, and show through real‑world and simulation experiments that LLMD and ACC together provide reliable and early failure detection, outperforming more expensive baseline methods.
By Florian Seligmann, Emiliyan Gospodinov, Enes Ulas Dincer, Gerhard Neumann