跳到正文
研踪PaperTrail

论文库

检索计算机领域论文,按研究方向、作者或关键词筛选。

部分采集异常
正在读取账号与研究记录…

全站更新概况

+1959新增论文
+1550其中 AI 相关
+212已有论文修订
+67新增 AI 导读

arXiv 最近完成扫描:2026-11-17 23:20 · 全站最近入库:2026-11-17 23:35各信源进度 →

统计口径与数据新鲜度

统计 2026-11-17 01:13 — 2026-11-18 01:13(北京时间) · 点击数字查看对应简报;AI 相关包含在新增论文内。

最近完成检查:2026-11-17 23:20最近发现新增版本:2026-11-17 23:35库内论文最新更新时间:2026-11-17 02:59

每分钟刷新已保存的数据。空检查不代表采集异常;新收录可能包含历史论文。AI 导读按新增版本计数。

25,786 已收录论文21,491 AI 相关7,725 有修订112 HF 社区发现

浏览论文库

时间范围跟随排序:版本更新、本站收录或首次发表。最新收录可能包含历史论文。部分信源异常,查看状态 →

找到 377 篇论文 · 按版本更新时间排序上轮扫描起点 2026-11-17 23:19(北京时间)
新论文 · v1arXiv:2611.21477

Knowledge Fermentation: Learning from Shared Variation in Few-Shot Classification

Ali Mohammadi Ruzbahani, Abbas Yazdinejad, Hadis Karimipour

原文摘要

A classifier built from a handful of labeled examples knows where each class sits in feature space but not how it varies, which is why prototype rules and linear probes overfit in the low-data regime. Knowledge Fermentation (KF) recovers this second-order knowledge by letting classes exchange what they know about variation. Residuals around class prototypes act as carriers; the variance of each direction is measured on classes that took no part in estimating it, and directions above the level of the surrounding medium define a structured noise that acts on a linear classifier as a closed-form change of geometry. A convex objective trains the classifier at a noise strength calibrated on the distances between prototypes, and unlabeled examples, when available, join as carriers in repeated rounds. We prove that the held-out variances are unbiased for the directions on which they are measured, that the fermented map never expands a direction, and that damping caused by wrong assignments limits itself. On 14 datasets from vision and language and 8 frozen encoders, with all constants fixed, KF is on par with the best of six tuned inductive baselines. With unlabeled carriers it raises one-shot accuracy from 52.4% to 65.8% on average and attains the highest average accuracy among nine transductive methods, significantly so on vision encoders, while information maximization stays ahead on sentence encoders. Controlled studies attribute the gain to the task-level geometry and show that it grows with the within-class variation that the classes of a task share.

cs.IT

版本更新 · v3arXiv:2611.15592

Small Heads, Strong Decisions: A Controlled Study of Typed-Decision Models

Ritvik Rastogi

原文摘要

In September 2026, TypeSafe announced Jev as the first model in its System One category: unstructured state goes in, and typed, calibrated decisions that software can use directly come out. Jev's architecture and training data are unpublished, but many open typed-decision models have followed, including Nimble, the Decision 1.0-3.0 family and Clef. They answer structured questions about a piece of text, such as a yes/no check, a choice among labelled options or a graded score, by reading the text with a language model and scoring the options with a small readout. Each pairs a different readout with a different backbone, training mix and recipe, so it is unclear which design choices matter. We re-implement five readouts and train all of them under one protocol: the same backbone, the same 15 public English training sources, the same recipe with a per-head learning-rate sweep, and the same seed. Our experiments demonstrate that: The complexity of the head does not affect accuracy: on Qwen3-0.6B every readout reaches 81.3-82.0% in-domain accuracy, and a 0.5M-parameter cosine head, which scores each option by its cosine similarity to the question, matches the 27M-parameter joint schema head adapted from Clef in-domain at half its inference time. Reading all of a text's questions from one prompt also matches one prompt per question in accuracy, with 1.6 times faster inference and 2.1-2.5 times faster training on one H100. The backbone's size matters more than either choice: moving from Qwen3-0.6B to Qwen3-4B adds 4 points in-domain and 11 points on held-out sources, while the base, instruct and thinking checkpoints of the same model tie. We recommend spending effort on the backbone and keeping the head minimal.

cs.AI

新论文 · v1arXiv:2611.21370

Federated TD Learning in Heterogeneous Environments with Average Rewards: A Two-timescale Approach with Polyak-Ruppert Averaging

Ankur Naskar, Gugan Thoppe, Abbasali Koochakzadeh, Vijay Gupta

原文摘要

Federated Reinforcement Learning (FRL) provides a promising way to speedup training in reinforcement learning when one has access to multiple edge devices that can operate in parallel. Recently, it has been shown that even when these edge devices have access to different dynamic models---such as cars or robots with differing system configurations---an optimal convergence rate that has a linear speedup proportional to the number of devices is achievable. However, this result requires that the algorithm's stepsize be chosen in a manner that depends on the unknown model parameters. Also, it applies only to a discounted setting, which has been argued to fit episodic tasks better than continuing control tasks. In this paper, we obtain finite-time bounds for heterogeneous FRL with average rewards. We show that optimal convergence rate and linear speedup are achievable even with a universal stepsize choice, independent of the underlying dynamics. To achieve our result, we modify the existing one-timescale FRL method to a novel two-timescale variant that additionally incorporates iterate averaging. We remark that this idea also extends to other settings already studied in the literature, such as the homogeneous FRL or the heterogeneous case with a discounted cost function.

cs.LG

新论文 · v1arXiv:2611.21344

TEMPO: Regime-Aware Operator Learning with Locally Adapted POD Bases

Rodion Akinzhala, Alexander Terentyev, Alexey Golubev, Yury Maximov

原文摘要

Operator learning approximates the solution map of a parametric partial differential equation from data, enabling fast prediction at new parameter values without repeated simulation. The dominant approach fits a single global low-rank basis, typically a proper orthogonal decomposition (POD), to all solution snapshots. This fails when the snapshots span qualitatively different solution shapes, from smooth diffusive profiles to near-discontinuous shocks. We propose TEMPO (Trajectory Expectation-Maximization Mixture of POD Operators), a two-phase framework for regime-aware operator learning. In the offline phase, TEMPO recovers latent regimes via expectation-maximization, fitting a dedicated linear or nonlinear basis to each discovered regime. In the online phase, a gating network and per-regime branch networks are trained jointly, routing new inputs to the regimes at inference. We test TEMPO on Burgers' flow, Sod shock tubes, Coanda flow, Darcy flow, Navier-Stokes flow and the heterogeneous Helmholtz equation. With 64 total POD modes, TEMPO gives 18.61% test error on Burgers', compared with 19.61% for global POD. It also gives lower errors on Coanda and Helmholtz.

cs.LG

版本更新 · v4arXiv:2401.16095

On the Separability Problem of VASS Reachability Languages

Eren Keskin, Roland Meyer

原文摘要

We show that the regular separability problem of VASS reachability languages is decidable and $\mathbf{F}_ω$-complete. At the heart of our decision procedure are doubly-marked graph transition sequences, a new proof object that tracks a suitable product of the VASS we wish to separate. We give a decomposition algorithm for DMGTS that not only achieves perfectness as known from MGTS, but also a new property called faithfulness. Faithfulness allows us to construct, from a regular separator for the $\mathbb{Z}$-versions of the VASS, a regular separator for the $\mathbb{N}$-versions. Behind faithfulness is the insight that, for separability, it is sufficient to track the counters of one VASS modulo a large number that is determined by the decomposition.

cs.FL

新论文 · v1arXiv:2611.21160

PlainPairs: A Corpus Pairing the Prescriber and Patient Texts of U.S. Prescription Medicine Labels

Obada Kraishan

原文摘要

A prescription medicine in the United States comes with labeling written for prescribers, and many medicines also carry a text written for patients: a Medication Guide or a patient package insert. The two texts describe the same medicine for two readers, yet public resources rarely hold them side by side. I present PlainPairs, an open corpus built from the openFDA drug label files. Version 0.1.0 holds 35,177 prescription labels that carry a patient text, 21,416 distinct patient texts, 250,829 patient-text segments matched to prescriber sections through their regulatory headings, reading-level measures on both sides, and rule-based counts of plain-language wording. A second source, the National Cancer Institute's cancer information summaries, adds 756 summaries in health-professional and patient versions, in English and Spanish. Measured on the structure of the label, the median Medication Guide reads at grade 8.7 by Flesch-Kincaid and at grade 11.1 by SMOG. The same guides measured on the flattened text that the bulk files serve read about two grades higher, a difference that comes from lost list and heading boundaries. Patient texts address the reader in 41 of every 100 sentences, against 0.2 in prescriber sections. A blind hand check of 100 pairs supports the heading-based alignment for seven of ten heading types. The corpus, its code and its documents are public, and the release keeps texts that show a company copyright mark out of the text tables.

cs.CL · cs.CY

版本更新 · v3arXiv:2609.35040

GAZEleak: Passcode Inference Against Eye-tracking XR Devices Through External Observation

Hwanjo Heo, Junhee Lee, Jinwoo Kim

原文摘要

Mixed-reality headsets such as Apple Vision Pro replace the touch screen with gaze-as-pointer interaction: the wearer looks at a target and confirms with an air pinch. Because the display is inside the headset and the eye tracker is walled off from third-party software, such input is widely assumed to be unobservable to bystanders, a built-in defense against the shoulder-surfing that plagues phones and laptops. We present GAZEleak, a side-channel attack that recovers gaze-driven input from external video of head motion alone, under a strictly local, physical-observer threat model: the adversary only films the wearer from across the room and installs no software on the device. The attack exploits the centrally coupled eye-head motor program: gaze shifts recruit small, target-dependent head reorientations that project into sub-degree pose changes recoverable from commodity video. GAZEleak implements a measurement-based, sparse-optical-flow inference pipeline for users' 6-digit device passcodes. On a preliminary front-view dataset from three author-subjects who were aware of the attack hypothesis, GAZEleak places the true code within the top ten guesses for 56% (10 of 18) of test codes under a cross-person protocol with no labeled victim data, and for every code of the most exposed subject. The performance is subject-dependent: no passcode from the least exposed subject reaches the top ten, although its median guessed passcode rank is 12,786 rather than 500,000 expected from an uninformative ordering. These results provide preliminary evidence that gaze-coupled head motion can expose passcode information under controlled conditions, while motivating broader evaluation across users, behaviors, and capture settings.

cs.CR

新论文 · v1arXiv:2611.20986

Calpha: Toward Native Biomolecular Intelligence in a General-Purpose Language Model

Yong Liu, Zhanpeng Shi, Peicong Lin, Zhenyi Zhong, Yizhou Dang, Wendi Wu, Haoling Li, Shuangjia Zheng

原文摘要

Biomolecular foundation models predict the structures of proteins and their complexes accurately, and large language models (LLMs) reason in natural language, but the two remain separate: the abilities of the former are not part of a general-purpose LLM, and the reasoning of the latter does not enter a structure prediction. Calpha aims at one general model that folds proteins and complexes, understands protein sequence and structure, and lets its language reasoning influence the structures it predicts. This report covers the first step, in which an LLM outputs single-chain protein structures. Qwen3.5-9B reads a sequence enriched with ESM2 residue features. It can respond in text or emit a token that invokes a pretrained SimpleFold structure head, which receives information about the protein only through Qwen's residue states. Under a fixed folding request, the head achieves mean TM-scores of 0.808 on the FoldBench-334 development set, 0.635 on CASP16 domains and 0.900 on a CAMEO-style PDB set. We compare these predictions with a baseline in which ESM2-3B encodes the sequence and a fine-tuned SimpleFold-3B head predicts the coordinates. The two systems also differ in trainable weights, training duration and crop length. For chains of 100--299 residues, all three paired confidence intervals include zero. At 300--600 residues, the mean differences are -0.044, -0.160 and -0.025, respectively. In a separate test, four English request types triggered folding on every evaluated example. Protein training reduced accuracy on a 350-item MMLU-Pro subset from 77.4% to 70.3%. An LLM can therefore output single-chain structures from its own states, with lower long-chain scores in the current comparisons and reduced accuracy on the evaluated MMLU-Pro subset. Complex prediction, structure understanding and the effect of language reasoning on the predicted coordinates remain future work.

q-bio.QM · cs.AI

版本更新 · v2arXiv:2610.26430

Physiology-Aware Graph Attention for Early Ventricular Reentry Prediction

Ma{ë}lis Morier, Jairo Rodr{í}guez Padilla, Mihaela Pop, Patrick Gallinari, Maxime Sermesant

原文摘要

Sudden cardiac death mostly originates from post-infarct ventricular tachycardia. This reentry is sustained by propagation around scar tissue, but its risk remains difficult to predict from infarct imaging alone. While in silico induction studies can reveal these mechanisms, exploring many substrates and pacing locations remains computationally expensive. We propose a graph attention classifier for early prediction of sustained ventricular reentry from simulated cardiac electrophysiology. Each simulation is represented as a graph with early nodal voltage samples and edge attributes encoding geometry, stimulation frequency, tissue labels, amplitude, and diastolic interval. These descriptors are computed only from the first two paced activations, before sustained reentry is observed. On a balanced independent test set of 1000 simulations, the classifier achieves 94.6% accuracy, 95.4% sensitivity, 93.8% specificity, 94.6% F1-score, 0.987 ROC-AUC, and 0.985 PR-AUC. Ablation experiments show that diastolic interval is the most critical descriptor. Qualitative attention inspection suggests that different heads focus on border-zone, stimulation location, and early propagation path. These results support graph attention as a fast early-prediction approach for large-scale in silico ventricular reentry studies.

cs.AI

新论文 · v1arXiv:2611.20841

Vision-Language Model-Based Phase-OTDR Event Classification

Ahmet Bestami Kose, Mehmet Sergen Catal, Muhammet Cagri Yeke, Kivilcim Yuksel, Abdurrahman Gumus

原文摘要

Phase-sensitive optical time-domain reflectometry (Phase-OTDR) enables distributed vibration sensing, but reliable event classification has traditionally relied on fully supervised models that require extensive annotated data. This study examines the adaptability of a Vision-Language Model (VLM) for Phase-OTDR event recognition through few-shot learning. Raw samples are converted into Short-Time Fourier Transform (STFT) dB-magnitude spectrogram mosaics, color-mapped and resized to RGB images, and processed by a Contrastive Language-Image Pre-training (CLIP) image encoder; few-shot adaptation optimizes only the Tip-Adapter-F cache keys, leaving both encoders unchanged. Under a leakage-controlled protocol (train-side support/validation pool of 512 images per class, fixed held-out test of 120 per class, no test-set access during selection, 20 distinct support-sampling trials), the selected configuration (the ViT-L-14 Vision Transformer encoder with OpenAI weights, 128 shots per class) reaches 96.81 +/- 0.42% accuracy and a macro-averaged F1 score of 96.79 +/- 0.42% on the held-out test set; a label-efficient 16-shot setting, that is, 96 labeled support images at adaptation time across the six classes, already attains 90.17 +/- 0.81%. Tip-Adapter-F clearly outperforms both zero-shot CLIP and the standard training-free Tip-Adapter, and prompt and backbone analyses show that performance is stable across Vision Transformer encoders and insensitive to prompt wording at the selected operating point. Overall, CLIP visual features carry a transferable structure that a lightweight support cache turns into a label-efficient Phase-OTDR event classifier. The code and the corresponding image dataset are made publicly available on GitHub to support further research: https://github.com/miralab-ai/vlm-phase-otdr.

cs.CV · cs.LG

新论文 · v1arXiv:2611.20696

Conic Structure of the Cross-Entropy Loss at the Final Layer and Its Absence in Rectifier Backbones

Omer Nezih Gerek

原文摘要

Slices of the loss along lines and planes in parameter space are widely used to visualize deep networks. They often show a single basin with nearly straight walls, which raises the question of whether such a loss is a cone rounded at its apex, or a lower envelope of several cones that meet along creases. We answer this question for softmax cross-entropy at two levels of a classifier. Over the weights of an affine final layer, the loss equals a polyhedral cone plus an offset that decreases monotonically and exponentially along every ray, towards a limit in closed form. The distance at which the asymptotic wall reaches the height of the starting point, which we call the core radius, equals the mean surprisal of the classes that the direction favours, divided by the slope of the wall. Over the parameters of a rectifier backbone, the loss along a line has a kink wherever the activation pattern of a sample changes. Using the dimension theory of semialgebraic sets, we show that, along almost every line, the samples that switch at one point share an irreducible factor of their pre-activation polynomials. Under a coprimality condition, each kink therefore carries the share of one sample among $n$, so creases comparable to the slope do not form. Measurements on ResNet-18 and ResNet-56 trained on CIFAR-10 illustrate both results, with a test that bounds the size of any crease that sampled loss values can hide.

cs.LG · eess.SP

新论文 · v1arXiv:2611.20645

Defect-Aware Masked Autoencoders for Unsupervised Clustering of Steel Surface Defects

Shrey Patel

原文摘要

Reliable recognition of steel surface defects is essential for automated quality inspection, but supervised methods require manually annotated images that are costly to produce. This creates a need for methods that can learn useful defect representations directly from unlabelled surface images. This work presents Defect aware MAE, a label free masked autoencoder for clustering steel surface defects. Each image is divided into 196 patches, with 147 masked and 49 processed by a Vision Transformer encoder. Texture energy based masking preferentially hides structurally informative patches, while a shallow decoder reconstructs HOG descriptors that capture local texture and edge orientation. Strong augmentation and an intermediate layer readout are used to form the final image features. Pretraining uses no class labels, bounding boxes, detection branch, or contrastive loss. On the held out NEU DET split, UMAP followed by Ward clustering achieves 98.6% Hungarian matched accuracy, 0.968 NMI, 0.967 ARI, and 0.838 silhouette score. Under an identical encoder and training schedule, it exceeds a SimCLR contrastive baseline in accuracy, NMI, ARI and silhouette scores across 5 seeds.

cs.CV

新论文 · v1arXiv:2611.20575

PerturbJEPA: Joint-Embedding Predictive Learning of Cellular Perturbation Responses

Jiangshuan Pang, Youzhe He, Yiting Deng, Wangyang Tang, Huinan Xu, Qi Lu, Shiping Liu, Zhenkun Zhuang

原文摘要

Identifying the drug and dose that produced a cellular response requires preserving treatment effects within each cellular background. A joint-embedding predictor can match its targets while ignoring the intervention if its encoder preserves differences across backgrounds but erases differences between treated and control populations within each one. We call this failure interventional contrast collapse. PerturbJEPA encodes the difference in mean gene expression between treated and control populations as an effect vector. It aligns the displacement between their latent centroids with this vector and regularizes the vectors to preserve variation across population pairs. We show that the measured displacements cannot all coincide within a batch when effect-vector variation exceeds the root-mean-square alignment error. Trained with separate drug and dose ranking objectives, the model predicts candidate responses from a common control population and matches them to the observed response without cell pairing or expression reconstruction. We evaluate 3,592 Tahoe-100M queries on held-out combinations of known interventions and cellular backgrounds, each with over 700 drug-dose candidates in a fixed, background-specific catalog. PerturbJEPA achieves 91.46% macro joint drug-dose Recall@1, exceeding STATE by 3.40 percentage points (95% drug-cluster bootstrap interval of 1.72 to 5.07 points). Controlled diagnostics show that treatment-control differences can shrink even when representations remain globally dispersed. With the drug ranking objective fixed, adding effect-vector regularization to alignment raises joint Recall@1 from 18.57% to 87.47%.

cs.CV

新论文 · v1arXiv:2611.20570

OrthoMark: Place Recognition with Point-Anchored Orthographic Height Mark

Haichuan Li

原文摘要

LiDAR place recognition provides the loop closures of Simultaneous Localization and Mapping(SLAM). Perspective from Z-axis, a gravity-aligned scan is an orthographic map of the scene, and a revisit from another position and heading moves this map rigidly in the plane, a motion that a polar Fourier spectrum ignores and phase correlation recovers. Along the vertical, the heights of the returns describe the structures, but which heights are observed depends on the range, because the sensor's vertical field of view cuts nearby structures low; height thus misleads a global key, yet is decisive once two scans are aligned and compared cell by cell where they overlap. We present \method, a training-free method built on one principle: two LiDAR scans can only be compared through features computed from the same geometric surfaces in both scans. \method retrieves candidates with a translation- and rotation-invariant polar spectrum of the bird's-eye view (BEV), 1.9,KB per keyframe, and verifies each candidate by estimating its planar pose through phase correlation and scoring it with \emph{point-anchored height columns}: every BEV cell keeps a 16-bit mask of the heights of its own returns; within a session, the score is weighted by the expected overlap of the two scans' sensing ranges, computed from their distance. We show that this vertical link pays off in verification but hurts in retrieval, where what a height slice contains depends on the range to the structure. For single robot, \method has the highest AP of the training-free methods without 3D registration on the KITTI development sequences; across robots, on all multi-robot sequences, it matches or exceeds RING++ at 1/350 of its descriptor size and closes more inter-robot loops than the learned BEVPlace++ on all tests.

cs.RO · cs.CV

版本更新 · v3arXiv:2607.28591

Change2Task: From Repository Changes to Executable Coding Agent Tasks and Environments

Haomin Qi, Xingliang Wang, Xuanqi Gao, Baihui Sang, Xin Zhang, Minghua Ma, Pengfei Gao, Yu Kang 等 12 位作者

原文摘要

Coding agents learn from and are measured on executable tasks, but building such tasks from repository history usually means rebuilding an environment at an old revision for every task, which is costly and brittle. We observe that a merged pull request (PR) can instead be undone on today's code: removing the effect of a bug fix or feature from a maintained revision recreates the same task inside an environment that already builds and runs. We present Change2Task, which converts merged PRs into verified tasks on an up-to-date revision of the same repository. It reconstructs the pre-change state by patch reversal, code mapping, or agent-based rewriting, and keeps a task only if its target tests pass, fail, and pass again across the healthy, task, and restored states. From 1,130 PRs spanning bug fixing, feature addition, test generation, API migration, and security repair, Change2Task builds 900 verified tasks, recovering 29% more than SWE-smith's PR Mirror on matched bug-fix candidates. The reconstructed tasks preserve the evaluation signal: in aggregate, four coding agents solve 41.1% of tasks on both the original and reconstructed versions, with 89.7% per-task agreement. Sharing modern environments across tasks further reduces setup time by 58% and storage by 71%. On 11 private industrial repositories, Change2Task yields 1,000 tasks on which the strongest model reaches only 14.6% Pass@3.

cs.SE · cs.CL · cs.LG

新论文 · v1arXiv:2611.20472

Model Immunization Beyond Condition Numbers: The Importance of Plateau Regions

Quoc Nguyen, Trung Le, Dinh Phung, Mehrtash Harandi

原文摘要

Harmful fine-tuning can rapidly compromise safety alignment in Large Language Models, underscoring the vulnerability of deployed models to post-release modification. While prior work motivates model immunization by inducing ill-conditioned curvature to slow harmful optimization, this approach is not scalable due to the high computational cost of Hessian estimation. We refine this paradigm by extending the theoretical framework to prioritize plateau regions alongside curvature. We further introduce an immunization algorithm based on efficient Hessian estimation and bounds on the extremal singular values of the Hessian. We demonstrate the proposed immunization framework in three applications: (i) Fine-tuning-as-a-service settings where benign user data may be mixed with poisoned harmful data, (ii) open-weight settings where an adversary can fine-tune on harmful-only data with varied training dynamics, and (iii) robust unlearning, where immunized models better resist relearning attacks after unlearning. Extensive experiments confirm that our approach achieves robust defense against adversarial fine-tuning while largely maintaining general utility and downstream trainability. Our code is available at https://github.com/minhquoc0712/BECON.

cs.LG · cs.AI

版本更新 · v4arXiv:2609.36413

One from Infinity: Actualizing Futures from Pretrained World Models into Robot Actions

Bang Du, Yichen Xie, Shuqi Zhao, Yuxin Chen, Menglin Wu, Masayoshi Tomizuka

原文摘要

A pretrained video world model admits many plausible futures for a scene, but a robot must realize the exact task-conditioned one. To turn world models into executable robot policies, existing methods fine-tune the heavy world model backbone using large-scale robot data and computational resources. Challenging this status quo, we argue that the expensive part has already been paid in the world model pretraining since the representation space of a video world model encoder lays out the diverse potential futures. In this case, what remains is to determine the future that accomplishes the task and to read out the actions that realize it. We formalize this task as actualization, which learns a task-conditioned realization on top of a prior supplied by a frozen world model encoder. This can be solved by a tiny actualizer model. We implement RoboActualizer with as few as 60M parameters on top of a frozen world model encoder. The actualizer is composed of two lightweight DiT experts that jointly predict future latents and actions by flow matching. The model can be trained entirely on a single GPU with 32 GB peak memory. With up to 100x fewer trainable parameters than existing WAMs and VLAs, RoboActualizer reaches great performance on simulation benchmarks including LIBERO, LIBERO-Plus, RoboTwin 2.0 and five tasks on two real-world platforms, with a low latency of 39 ms that allows real-time control.

cs.RO

新论文 · v1arXiv:2611.20437

On the Convergence of Continuous-Time AdamW

Bangjie Wang, Ning Ning

原文摘要

Continuous-time AdamW provides a framework for understanding whether an optimizer widely used to train large-scale neural networks settles to an equilibrium or sustains persistent motion. We study this long-time behavior while retaining bias correction and fixed positive weight decay. Under suitable assumptions, we establish that the dynamics are well defined for all time and remain bounded when the decay is sufficiently small. Our main result connects their limiting behavior to the geometry of the objective: trajectories approach a locally invariant manifold whose dimension equals the Hessian nullity along the associated critical component. When critical points are isolated and have at most one such direction, the trajectory converges to a single equilibrium. With two degenerate directions, however, bounded persistent motion can occur even for a smooth, strictly convex objective with a unique minimum. We also quantify how close the long-time trajectory remains to the critical set as weight decay decreases, without assuming convergence to a point. These results identify objective geometry as a key to understanding when continuous-time AdamW converges and when persistent motion is possible.

math.OC · cs.LG

新论文 · v1arXiv:2611.20417

When the Cheapest Feasible Router Fails: Auditing Deployment Commitments in Cost-Aware LLM Routing

Minxuan Hu, Qishi Zhan, Ziheng Chen, Zhengtao Yao, Zihan Dong, Jiayu Yi, Wenxi Sun

原文摘要

Cost-aware LLM routing often selects the cheapest policy meeting empirical risk and coverage constraints. Yet a router that passes an audit can violate its fixed service contract on new requests, even when candidate policies are constructed independently. We study this deployment-commitment decision with disjoint construction, audit, and test samples. Across 19 RouterBench task families, the cheapest audit-feasible router deploys in 91.82% of opportunities but violates the held-out contract in 32.22% of deployments. We introduce Menu-Aware Routing Commitment (MARC), which qualifies policies using audit evidence and a serving-cost tolerance to retain the apparent winner, deploy a supported alternative, or defer. Its practical policywise variant (MARC-P) reduces observed conditional failure to 0.73% at 85.79% deployment, with higher serving cost; fixed-margin and formal risk-control procedures provide competitive alternatives. Nested audits show how added evidence changes deployable policy sets, and learned semantic routing exhibits the same commitment failure. These results establish deployment commitment as a distinct operational decision connecting router evaluation, service reliability, and inference cost.

cs.LG · cs.CL

新论文 · v1arXiv:2611.20384

JointQ: Joint Scale-and-Integer Quantization for Large Language Models

Keiji Kimura, Yuma Ichikawa

原文摘要

GPTQ is the de facto standard solver for layer-wise post-training quantization (PTQ) of large language models, yet the solver itself has received less attention than weight preprocessing and cross-layer objectives. We target a bottleneck of GPTQ: its greedy procedure optimizes continuous group scales separately from the assignment of discrete integers. Evaluating each integer candidate with its closed-form optimal scale has been used for per-channel quantization; JointQ extends it to group-wise quantization, where a row's scales are coupled through the Hessian, and applies it to refine the GPTQ solution. Each local-search move changes one integer and recomputes its group's optimal scale in closed form, accounting for the other groups through the full Hessian; all scales are periodically re-solved jointly. The search starts from a row-wise selection among scale-refitted candidates including the GPTQ solution, so unregularized JointQ never has a larger layer-wise calibration error than GPTQ. JointQ preserves the GPTQ output format and is thus a drop-in replacement. Across four models and four settings, JointQ improves six-task average zero-shot accuracy over vanilla GPTQ in 14 of 16 comparisons and perplexity in 12 of 16. The mean accuracy gain increases from 0.0044 in W4G128 to 0.1559 in W3 per-channel, and JointQ also improves average accuracy over GPTQ with MSE-based scale search and activation ordering in all eight per-channel comparisons, though rarely in G128. An ablation shows that JointQ's diagonal Tikhonov regularization mitigates severe W3 per-channel perplexity degradation, so the largest gains are not due to the local search alone. Overall, JointQ's primary empirical benefit is downstream accuracy preservation in challenging low-bit and large-group regimes. Code is available in the OneComp library at https://github.com/FujitsuResearch/OneCompression.

cs.LG · cs.CL

新论文 · v1arXiv:2611.20164

Cross-Subject Priors for Maximum a Posteriori Estimation of Latent Structure Influence Models in Motor-Imagery EEG With Few Trials

Sagnik De, Anil Kumar Gupta

原文摘要

Coupled hidden Markov models such as the latent structure influence model (LSIM) describe how the sensorimotor rhythms of the two hemispheres evolve and interact during motor imagery, but subject-specific maximum likelihood (ML) estimates are unstable when few trials are available. This paper derives a maximum a posteriori (MAP) estimator for the LSIM whose conjugate prior equals the expected statistics of $λ$ pseudo-trials from a population model fitted to other subjects. The M-step remains in closed form, expectation maximization keeps its monotone ascent, and ML ($λ=0$) and the population model ($λ\to\infty$) are limiting cases. The prior strength was selected on 35 subjects of the PhysioNet EEG Motor Movement/Imagery Dataset and applied unchanged to 70 held-out subjects and to BCI Competition IV-2a. With $λ=1$ the held-out log-likelihood exceeded that of ML in 61 of 70 subjects (Holm-adjusted $p=9.1\times10^{-8}$) and in all 70 subjects with five training trials per class. A prior learned only from the PhysioNet data improved the held-out likelihood of all nine BCI IV-2a subjects at five trials per class and reduced the error of the joint transition matrix by 61%. Decoding accuracy did not improve ($p=0.17$). Cross-subject priors therefore stabilize few-trial dynamic models of motor-imagery EEG, but they do not improve decoding, which in these data relies mainly on structure shared across subjects.

eess.SP · cs.LG

版本更新 · v3arXiv:2512.16700

CLARiTy: A Vision Transformer for Multi-Label Classification and Weakly-Supervised Localization of Chest X-ray Pathologies

John M. Statheros, Hairong Wang, Richard Klein

原文摘要

Chest X-ray interpretation requires multi-label classification and spatial localization, yet region-level annotations are scarce, so weakly supervised localization involves training on image-level labels to produce class heatmaps. The step that converts heatmaps into boxes is rarely calibrated, that is, fitted on validation images, and localization results seldom leave the training dataset. We present CLARiTy (Class Localizing and Attention Refining Image Transformer), a vision transformer whose class tokens yield class-specific attention maps without gradient class activation mapping. A SegmentCAM head trained on anatomical region masks, not pathology masks, suppresses background. OneBox then converts the heatmaps into boxes, with parameters fitted per finding and resolution. On the official NIH split, CLARiTy outperforms the compared methods at matched resolution in localization and remains competitive in classification. Against published ThoraX-PriorNet values at matched 224 px resolution, the relative gain in macro intersection-over-union (IoU) accuracy at IoU 0.5 is 50.7%. The largest relative gain is on nodules: at that threshold the 224-pixel model reaches 0.154, where no compared method exceeds 0.013. Without training any weights on RSNA Pneumonia or VinDr-CXR, we refit only the OneBox parameters on target validation images, leaving the model and heatmaps unchanged. Localization is competitive with RAD-DINO and ahead of TorchXRayVision DenseNet121-NIH at every threshold. Fitting the box extractor on validation images, freezing it before testing and reporting localization across thresholds gives a protocol that carries to new corpora without retraining the model. Code is available at https://github.com/jstat17/clarity-cxr.

cs.LG

新论文 · v1arXiv:2611.19985

When Zero Is Competitive: A Descriptive Pilot of Transformer and Gate-Graph Surrogates on Random-Angle Quantum Circuits

Mahesh Reddy Pagadala

原文摘要

We examine supervised classical prediction of all one-qubit Z and two-qubit ZZ expectation values on noiseless, layered 4-10-qubit RX/RY/RZ-CNOT circuits. In a 63-cell stratified workload with disjoint circuit-template tests, we evaluate one transformer encoder and gate-dependency graph neural networks against literal-zero and training-derived constant predictors. On the Arm K training labels, 44.89% of scalar targets have absolute magnitude below 0.05. The held-out-template zero predictor has equal-instance mean absolute error 0.193399, and none of the evaluated transformer checkpoints or final graph-network endpoints achieves a lower score. A separate product-state positive control achieves MAE 0.007755 versus 0.514811 for zero, but does not explain the failure on the entangled workload. The difference between zero and a training-fitted mean constant under training MSE is only 0.14%, a descriptive point estimate without uncertainty analysis. The transformer's 815,233-parameter source identity is locally verified; detailed packed-input semantics have not been independently audited. These results constitute a descriptive pilot, not a causal explanation, formal equivalence test, or impossibility claim. Comparable circuits can be regenerated from the workload specification, but exact frozen datasets and model-run evidence remain private.

cs.LG

新论文 · v1arXiv:2611.19958

Multimodal AI in Healthcare: Review of Vision-Language Foundation Models for Real-World Medical Applications

Taha Razzaq, Murtaza Taj, Asim Iqbal

原文摘要

The emergence of foundation models has marked a transformative shift in AI, enabling robust generalization across diverse downstream tasks through putative zero-shot learning. Large Language Models and Vision Language Models have demonstrated strong capabilities in tasks such as image interpretation, report generation, and question answering by effectively learning from multimodal data-images paired with associated text often with minimal supervision. In the healthcare domain, this ability to align visual and textual information reduces the reliance on extensive manual annotations, as models can leverage existing clinical reports and imaging data to learn meaningful representations. This integration holds promise for improving diagnostic support, treatment planning, and overall patient care, even in data constrained settings. In this review, we provide a definitive taxonomy of the medical VLM landscape, tracing the evolution from early Contrastive Alignment and Generative MLLMs to the cutting edge frontiers of Dense Pixel Grounding, Sparse Mixture-of-Experts (MoE), and Reasoning Incentivized (RL) architectures. We critically examine the "medical bottleneck" identifying the persistent challenges of data scarcity, the "hallucination" risks in generative diagnostics, the computational strain of 3D volumetric processing, and the lack of standardized, clinically grounded evaluation metrics.

cs.CV

版本更新 · v2arXiv:2605.06819

A Theory of Online Learning with Autoregressive Chain-of-Thought Reasoning

Ilan Doron-Arad, Idan Mehalel, Elchanan Mossel

原文摘要

Autoregressive generation lies at the heart of the mechanism of large language models. It can be viewed as the repeated application of a next-token generator: starting from an input string (prompt), the generator is applied for $M$ steps, and the last generated token is taken as the final output. [Joshi et al., 2025] proposed a PAC model for studying the learnability of the input-output maps arising from this process. We develop an online analogue of this framework, focusing on the mistake bound of learning the final output induced by an unknown next-token generator. We distinguish between two forms of feedback. In the End-to-End model, after each round the learner observes only the final token produced after $M$ autoregressive steps. In the Chain-of-Thought model, the learner is additionally shown the entire $M$-step trajectory. Our goal is to understand how the optimal mistake bound depends on the generation horizon $M$, and to what extent observing intermediate tokens can reduce this dependence. Our main results show that the online theory of autoregressive learning exhibits a qualitative picture analogous to the statistical one found by [Hanneke et al., 2026], but with a different scale of dependence on the generation horizon. In the End-to-End model, we prove a taxonomy of possible mistake-bound growth rates in the generation horizon $M$: essentially any rate between constant and logarithmic can arise. We further show that this logarithmic ceiling is unavoidable. In the Chain-of-Thought model, we show that access to the full generated trajectory eliminates the dependence on $M$ altogether. We also analyze autoregressive linear threshold classes, and prove optimal mistake bounds, as well as a new lower bound for the statistical setting. Along the way, our results resolve several questions left open by [Joshi et al., 2025].

cs.LG

新论文 · v1arXiv:2611.19820

Hermitian Positive Definite Neural Networks for Matrix-CFAR Geometric Signal Detection in Nonhomogeneous Clutter

Xiyao Chen, Takumi Sato, Linyu Peng

原文摘要

Matrix information geometry (MIG) detectors exploit the Riemannian geometry of Hermitian positive definite (HPD) matrices to achieve robust object detection in nonhomogeneous clutter. The matrix constant false alarm rate (matrix-CFAR) method is a representative MIG detector that relies on the geometric structures with respect to specific metrics. In most cases, computing representative HPD covariance matrices requires numerical methods, typically involving computationally intensive iterative optimization algorithms. Although explicit formulas are available for certain cases, such as the log-Euclidean (LE) mean, they are generally unavailable for many other matrix means and medians. To address this challenge, we propose integrating the Hermitian positive definite neural network (HPDNet) into the matrix-CFAR framework, replacing conventional numerical optimization procedures with a learned, geometry-preserving mapping that directly estimates representative HPD covariance matrices from data. In the simulations, the framework is evaluated for MIG detectors using both means and medians associated with the affine-invariant Riemannian metric, the LE metric, and the Bures--Wasserstein metric, on both simulated compound-Gaussian clutter and real IPIX sea clutter. In both cases, the HPDNet is trained solely on simulated data to assess its applicability to the study of real data. The results show a significant reduction of computational cost in testing, while maintaining detection performance and robustness.

eess.SP · cs.LG · stat.AP

新论文 · v1arXiv:2611.19754

Input-Space versus Feature-Level Conditioning for Robust Facial Expression Recognition

Weihang Yuan

原文摘要

Facial expression recognition (FER) degrades sharply when illumination, exposure or resolution departs from the training distribution. Two remedies dominate practice: enriching the input signal, or conditioning intermediate features on an estimated image quality. The two are rarely compared under matched backbones, training budgets and condition annotations, and their cost is seldom reported. We present a controlled comparison on RAF-DB. Input-space conditioning uses four deterministic, label-free views (logarithmic stretching, inverse-logarithmic stretching, down-up resampling and identity), each processed by a shared backbone with a branch-specific BatchNorm and linear head, with the branch logits averaged. Feature-space conditioning is implemented as FiLM and Conditional BatchNorm under an identical protocol. Across six backbone architectures the input-space variant improves top-1 accuracy on every one (mean +0.94 percentage points; bootstrap interval [0.51, 1.63]; sign test p = 0.016), with a seed-controlled estimate of +0.42 percentage points on ResNet-18, whereas feature-level conditioning yields +0.10 and +0.03 percentage points. We attribute the gap to the information content of the heuristic condition label rather than to the conditioning mechanism. The gain costs 4.00x FLOPs, 4.18x latency and 3.79x training time, i.e. 0.26 percentage points per additional GFLOP, and we state an explicit rule for when the trade is worth making.

cs.CV

版本更新 · v2arXiv:2609.31117

Comparative Evaluation of an XR Pen-based Control Interface for Semi-Autonomous Mobile Robot Navigation in Service Environments

Alicia Torck, Carl Tornberg, Eric Piette, Renaud Ronsse, Benoit Macq, Gustavo Alfonso Garcia Ricardez, Lotfi El Hafi, Tadahiro Taniguchi

原文摘要

Service robots remain difficult to deploy in domestic environments, partly because fully autonomous operation is not yet reliable in unpredictable surroundings, and partly because conventional control methods remain inaccessible to novice users. Extended Reality (XR) enables operators to visualize robot information overlaid onto the real world and to interact with augmented elements. Yet, common XR control methods, such as motion controllers and hand gestures, are still perceived as unintuitive. This paper presents a control interface that uses a commercial XR pen to command a semi-autonomous mobile robot in Augmented Reality (AR): the operator points at a position in the room, selects it, and drags an augmented arrow to set the desired orientation of the robot at this destination. Two additional interfaces, based on the XR motion controllers and hand gestures, were developed within the same framework. To assess the performance and users' perception of these interfaces, and of the XR pen in particular, a study with 10 participants compared four control methods, i.e., the XR pen, the XR motion controllers, hand gestures, and a computer-based baseline RViz, in navigation tasks performed in a home-like environment. Repeated-measures analyses show that the control method significantly affects the perceived workload and usability: the XR motion controllers obtain the best mean scores, comparable to those of the computer-based baseline, whereas hand gestures lead to a significantly higher workload than both. The XR pen yields the shortest mean task selection time with the smallest variability, but this difference does not reach statistical significance, and a tracking issue of the XR pen, caused by the headset switching to hand tracking, held back its user experience.

cs.RO · cs.HC

版本更新 · v3arXiv:2503.14712

Near-Optimal Techniques for Distributed Entanglement Generation and Purification

Yukun Yang, Xiaojie Fan, C. R. Ramakrishnan, Himanshu Gupta

原文摘要

We develop a purification-aware optimization framework for distributing high-fidelity entangled states over quantum networks. In repeater-based architectures, the framework jointly selects entanglement swapping, fusion or graph-state growth, and purification operations to maximize generation rate under a fidelity threshold. Different purification models can be incorporated through their fidelity-update, success-probability, and latency rules. We introduce purification-augmented generation trees and dynamic programming (DP) algorithms for entangled-pair (EP) distribution. Hypergraph-flow linear programming (LP) formulations support level-based structures and concurrent generation across multiple trees and demands. We extend the framework to GHZ states and arbitrary graph states constructed from remote EP resources, introduce graph-state construction primitives, and derive closed-form fidelity-vector updates for GHZ fusion and purification and graph-state growth and purification. We provide optimal formulations under the adopted models, discretization, and purification bounds, together with efficient DP and LP heuristics. We bound fidelity- and rate-discretization errors and establish convergence for fixed constructions under stated regularity conditions. NetSquid simulations of EP generation using our LP approach show an average 160% increase in generation rate over the compared prior approaches, with gains up to 330% in some scenarios. GHZ and graph-state simulations demonstrate the heuristics' effectiveness in balancing generation rate, output fidelity, and computational overhead.

quant-ph · cs.NI

新论文 · v1arXiv:2611.19692

MathaMAGICal: Validating Trajectory Data Attribution During Capability Acquisition

Jason Lee, Dikshanta Pokhrel, Phonn Nay Khant, Victor Lee, Moritz Weckbecker, Sebastian Lapuschkin, Wojciech Samek, Mark Riedl 等 9 位作者

原文摘要

Influence-function-based methods like MAGIC provide exact first-order scores for trajectory data attribution, yet only retraining indicates whether those scores accurately predict data removal. We evaluate MAGIC on $86$M- and $152$M-parameter GPT-2-style models learning multistep arithmetic from scratch on filtered web text and four synthetic math families. Full-run attribution at $86$M suffers from numerical overflow, and even with metagradient clipping, scores correlate near chance with retraining (linear datamodeling score, LDS, $0.17$--$0.22$). In contrast, short checkpoint windows yield stable, accurate predictions at both scales (LDS $0.86$--$0.99$ across four $100$-step windows at $152$M), whereas a single-checkpoint gradient-dot baseline shows near-zero correlation. Rankings of individual training blocks predict retraining only within that single run and do not transfer across seeds or hyperparameters ($ρ\le 0.09$), whereas format-level patterns replicate across runs. Two data families, worked examples and primitive facts, occupy $2.1$--$2.6\times$ their corpus share in both the helpful and harmful $5\%$ score tails, reflecting strong bidirectional local sensitivity distinct from net benefit under multi-epoch family deletion. These results show that trajectory attribution is reliable within short, validated windows and yields family-level insights that generalize across runs.

cs.LG · cs.AI · cs.CL