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

+1680新增论文
+1333其中 AI 相关
+187已有论文修订
+59新增 AI 导读

arXiv 最近完成扫描:2026-11-18 04:28 · 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 06:03 — 2026-11-18 06:03(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:2026-11-18 04:28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 652 篇论文 · 按版本更新时间排序上轮扫描起点 2026-11-18 04:27(北京时间)
新论文 · v1arXiv:2611.21584

CertiSum:跨平台与并行布局的逐位一致 MXFP4 训练

CertiSum: Bitwise-Identical MXFP4 Training Across Platforms and Parallelism Layouts

Steve Rhyner, Yubo Gao, Stefan de Lasa, Sankeerth Durvasula, Nandita Vijaykumar

AI 速读 基于摘要

CertiSum 利用 FP32 累加在多数求和顺序下已精确的特性,通过量化统计认证与精确重算,实现 MXFP4 训练在运行、并行布局和 GPU 厂商间的逐位一致。

cs.LG · cs.DC

查看原文摘要

Training with MXFP4, a 4-bit microscaling format, is non-deterministic: identical inputs can produce different results across parallelism layouts and GPU vendors. These differences complicate debugging and training audits and can cause training and inference engines to disagree. We present CertiSum to make MXFP4 training bitwise invariant across runs, parallelism layouts, and GPU vendors. Its key insight is that native FP32 accumulation of MXFP4 products is often already exact for every summation order. CertiSum collects statistics during quantization to certify native matrix multiplication results, recomputes uncertified outputs exactly, and uses exact kernels directly where certification rarely succeeds. CertiSum keeps quantization blocks unchanged across layouts and merges partial sums without rounding, including in attention and weight-gradient computation. In 200-step Qwen3-8B-Base experiments, CertiSum preserves training states bit for bit across AMD MI350X and NVIDIA GB200 and across changes in data, tensor, and context parallelism on MI350X. Across three 8B--14B models on FineWeb-Edu using MI350X, it adds 11.8--60.5% to training-step latency relative to its non-exact MXFP4 counterpart at the larger evaluated input shapes, with speedups of up to $1.24\times$ over BF16 and $1.70\times$ over deterministic BF16.

版本更新 · v2arXiv:2602.00092

通过宪法解释与控制模型行为:基于原子概念编辑

Interpreting and Controlling Model Behavior via Constitutions for Atomic Concept Edits

Neha Kalibhat, Zi Wang, Prasoon Bajpai, Drew Proud, Wenjun Zeng, Been Kim, Mani Malek

AI 速读 基于摘要

提出一种黑盒可解释性框架,通过原子概念编辑学习可验证的自然语言“宪法”,以解释和控制模型在多种任务中的行为。

cs.LG · cs.AI · cs.CL · cs.CV

查看原文摘要

We introduce a black-box interpretability framework that learns a verifiable constitution: a natural language summary of how changes to a prompt affect a model's specific behavior, such as its alignment, correctness, or adherence to constraints. Our method leverages atomic concept edits (ACEs), which are targeted operations that add, remove, or replace an interpretable concept in the input prompt. By systematically applying ACEs and observing the resulting effects on model behavior across various tasks, our framework learns a causal mapping from edits to predictable outcomes. This learned constitution provides deep, generalizable insights into the model. Empirically, we validate our approach across diverse tasks, including mathematical reasoning and text-to-image alignment, for controlling and understanding model behavior. We found that for text-to-image generation, GPT-Image tends to focus on grammatical adherence, while Imagen 4 prioritizes atmospheric coherence. In mathematical reasoning, distractor variables confuse GPT-5 but leave Gemini 2.5 models and o4-mini largely unaffected. Moreover, our results show that the learned constitutions are highly effective for controlling model behavior, achieving an average of 1.86 times boost in success rate over methods that do not use constitutions. Our open-sourced code can be found at https://github.com/google-deepmind/atomic_concept_edits

新论文 · v1arXiv:2611.21465

用少量雅可比样本热启动神经动力学模型

Warm Starting Neural Dynamics Models From Few Jacobian Samples

Maxime Rochkoulets, Joppe De Jonghe, Mariya Ishteva

AI 速读 基于摘要

提出 TDec-Init,利用少量转移雅可比样本通过张量分解解耦出权重与非线性,从而热启动神经网络动力学模型。

cs.LG · cs.AI · eess.SY

查看原文摘要

Model-based reinforcement learning (MBRL) leverages a model of the environment, used for planning and generating samples without interacting with the real world. Motivated by evidence that informed initialization significantly improves the efficiency of MBRL, we propose TDec-Init, an algorithm to warm-start neural dynamics models using transition Jacobians. Using a decoupling algorithm based on the decomposition of a tensor of stacked Jacobian samples, we recover weights and flexible nonlinearities, and use them to initialize a feedforward neural network. In the low-data regime, we show on several benchmark environments that our approach outperforms random initialization and value training, and either matches or outperforms Sobolev training, while requiring fewer training steps between episodes.

新论文 · v1arXiv:2611.21454

Comments on "Deep Fréchet Regression"

Jianbin Tan, Anru R. Zhang

原文摘要

We discuss the Journal of the American Statistical Association (JASA) discussion paper "Deep Fréchet Regression" and explore an alternative perspective based on conditional flow matching. We propose learning covariate-dependent transport directly within the response space, avoiding separate manifold learning and reverse mapping. This perspective offers a potentially more flexible framework for incorporating geometric structure and modeling full conditional distributions.

stat.ML · cs.LG

版本更新 · v2arXiv:2607.07386

Sparse Delta Memory: Scaling the State of Linear RNNs through Sparsity

Loïc Cabannes, Pierre-Emmanuel Mazaré, Ilze Amanda Auzina, Théo Le Pendeven, Gergely Szilvasy, Matthijs Douze, Maria Lomeli, Justin Carpentier 等 10 位作者

原文摘要

Linear attention models allow a fixed state size and a fixed amount of compute per token. However, due to their limited state size, linear attention models fall behind in long-context recall compared to softmax-attention-based transformer architectures. Increasing the state size of linear attention improves recall performance but at the cost of higher FLOPs. In this work, we introduce Sparse Delta Memory (SDM), an architecture that scales the hidden state of gated linear RNNs to orders of magnitude higher capacity using a sparse addressing scheme. SDM extends the Gated DeltaNet architecture by replacing the dense key-value outer product with sparse reads and writes to a large explicit memory. We show that, under an isoFLOP constraint and with an identical number of parameters, a higher state memory capacity significantly improves performance on in-context learning and long-context retrieval tasks. Moreover, by learning the initial state of the SDM memory and therefore using it as a parametric memory, we show that the model further improves on a wide range of common-knowledge and reasoning tasks.

cs.LG

新论文 · v1arXiv:2611.21295

Deep Learning for Physicists: From Linear Regression to Reasoning Models

Masanori Hanada

原文摘要

These lecture notes are based on an introductory course on neural networks and deep learning that the author taught in the School of Mathematical Sciences at Queen Mary University of London in 2023/24 and 2024/25. The author, however, is a physicist, and it shows: wherever possible, ideas are explained with physical intuition rather than heavy formalism. Readers with a background in physics or mathematics who want a self-contained, introductory route into modern machine learning may therefore find these notes particularly friendly. The notes start from maximum likelihood, least squares, linear regression and (stochastic) gradient descent. They then cover neural networks, backpropagation, the vanishing gradient problem, batch and layer normalization, overfitting and regularization, adversarial examples, and the kernel method. The second half turns to natural language processing: word embeddings, recurrent networks and LSTMs, the attention mechanism and Transformers, and large language models, including scaling laws, the history of GPT, instruction tuning, reinforcement learning from human feedback, hallucination, alignment faking, chain-of-thought reasoning and test-time scaling. The emphasis on large language models reflects requests from the students. The notes are aimed at third- and fourth-year students in the UK system. Because the course was also taken by students from other departments, very little is assumed: even basic material such as matrix multiplication is explained from scratch, so readers familiar with it can safely skip such parts. No originality is claimed; the aim is a coherent, intuitive path through the material. Developments after spring 2025 are not covered, but the basic ideas emphasized here are unlikely to go out of date soon.

cs.LG · hep-lat · hep-th · physics.comp-ph

版本更新 · v4arXiv:2505.07675

Simple yet Effective Semi-supervised Knowledge Distillation from Vision-Language Models via Dual-Head Optimization

Seongjae Kang, Dong Bok Lee, Hyungjoon Jang, Sung Ju Hwang

原文摘要

Semi-supervised learning (SSL) has emerged as a practical solution for addressing data scarcity challenges by leveraging unlabeled data. Recently, vision-language models (VLMs), pre-trained on massive image-text pairs, have demonstrated remarkable zero-/few-shot performance that often surpasses SSL approaches due to their exceptional generalization capabilities. This gap motivates us to question: how can we effectively harness the powerful generalization capabilities of VLMs into task-specific models? Knowledge distillation (KD) offers a natural framework for transferring VLM capabilities, but we identify that it suffers from gradient conflicts between supervised and distillation losses. To address this challenge, we propose Dual-Head Optimization (DHO), which introduces dual prediction heads for each distinct signal. We observe that DHO mitigates gradient conflicts, enabling improved feature learning compared to single-head KD baselines, with practical benefits of minimal computational overhead and test-time hyperparameter tuning without retraining. Extensive experiments across 15 datasets show that DHO consistently outperforms KD baselines, often outperforming teacher models with smaller student models. DHO also achieves new state-of-the-art performance on both in-distribution ImageNet semi-supervised learning and out-of-distribution generalization across ImageNet variants. Code and model checkpoints are publicly available at https://github.com/erjui/DHO/tree/main.

cs.LG · cs.AI · cs.CV

新论文 · v1arXiv:2611.21196

Determining Concentration from Raman Spectroscopy with Manifold Learning and Y-shaped Conformal Autoencoders

Anna Taka, Eleni D. Koronaki, Ioannis G. Kevrekidis, Eike Cramer, Joern Viell, Alexander Mitsos

原文摘要

Monitoring of chemical processes is crucial in chemical engineering, motivating sophisticated process analytical technology (PAT). In-line analytics are advantageous, with Raman spectroscopy being an established and reliable tool in various applications, from which many properties can be determined. We recently [Koronaki et al., AIChEJ 2024] proposed determining polymer particle size from Raman spectroscopic data, by combining Diffusion Maps (a manifold learning method) with a Y-shaped Conformal Autoencoder Neural Network. Herein, we extend this workflow to predict liquid concentrations, allowing also for multiple target predictions from the same model. We apply the methods on two datasets consisting of Raman measurements and concentration measurements: one from microgel polymerization and one from citric acid dissociation. The single-target concentration prediction results, where one target is determined from each model at a time, are accurate. Comparing against dedicated methods such as Indirect Hard Modeling (IHM) is not straightforward, due to the different validation strategies and the considered spectra range; nevertheless, it is fair to say that IHM performs slightly better than the proposed workflow. We further compare this approach with other regression methods, such as artificial neural networks, extreme gradient boosting and partial least squares, and the results show that the Conformal Autoencoders achieve higher predictive performance, demonstrating that the workflow can also be used for concentration determination. Finally we find that single-target concentration predictions are more accurate than the multiple target predictions for both case studies.

cs.LG

版本更新 · v2arXiv:2603.18657

Enhancing Multi-Corpus Training in SSL-Based Anti-Spoofing Models: Domain-Invariant Feature Extraction

Anh-Tuan Dao, Driss Matrouf, Mickael Rouvier, Nicholas Evans

原文摘要

The performance of speech spoofing detection often varies across different training and evaluation corpora. Leveraging multiple corpora typically enhances robustness and performance in fields like speaker recognition and speech recognition. However, our spoofing detection experiments show that multi-corpus training does not consistently improve performance and may even degrade it. We hypothesize that dataset-specific biases impair generalization, leading to performance instability. To address this, we propose an Invariant Domain Feature Extraction (IDFE) framework, employing multi-task learning and a gradient reversal layer to minimize corpus-specific information in learned embeddings. The IDFE framework reduces the average equal error rate by 20% compared to the baseline, assessed across four varied datasets.

cs.LG

新论文 · v1arXiv:2611.21020

SWE-OPD: On-Policy Distillation for Agentic Coding

Yuhang He, Yanyu Chen, Chuanzhe Guo, Jie Zhu, Zichen Tian, Hongyu Ge, Zixuan Ren, Huang He 等 14 位作者

原文摘要

In this work, we study on-policy distillation (OPD) for transferring agentic coding capabilities from a large teacher model to a smaller student model. Prior work has focused primarily on reasoning and short-horizon agentic tasks. We present SWE-OPD, a minimal OPD study for software engineering, where the student interacts with executable repositories and the teacher supervises the prefixes visited by the student. Across three teacher--student pairs, two training sources, and three SWE benchmarks, we find that on-policy supervision transfers coding capability: supervision from Qwen3.6-27B increases the SWE-bench Pro resolve rate of Qwen3.5-9B from 35.9% to 39.4%. Refreshing trajectories is not strictly required either, as an offline variant with cached teacher scores still improves over the student baselines while cutting end-to-end GPU-hours by 59.4%--72.6% and halving peak GPU allocation, at benchmark-dependent accuracy. Furthermore, we extend the offline variant to corss-tokenizer scenarios: distilling GLM-4.5-Air from DeepSeek-V4-Flash-0731 raises its SWE-bench Pro and Multilingual resolve rate from 35.8% to 39.7% and from 41.8% to 46.2%, around 4% and 4.6% improvement, respectively. We also find that a decreasing loss, however, does not necessarily indicate improved agent performance: under top-$K$ reverse-KL loss, one student exhibits severe repetition despite a decreasing loss. We identify a support-mass gradient term as a potential mechanism underlying this failure.

cs.AI · cs.LG

新论文 · v1arXiv:2611.21009

Learning and Verifying Stabilizing Controllers with Structured Neural Control Lyapunov Functions

Jose Luis Ledesma Martin, Patricia Pauli

原文摘要

This work investigates the joint learning of neural controllers and control Lyapunov functions (CLFs) for nonlinear discrete-time systems. We exploit the structural properties of Input Convex Neural Networks and Polyak-Łojasiewicz Networks to guarantee positive definiteness of CLF candidates by construction, reducing the conditions requiring formal verification. We develop a mixed-integer programming framework to formally verify the remaining stability conditions and introduce an iterative region of attraction (ROA) expansion procedure to enlarge the certified ROA without retraining. Experiments on an inverted pendulum and a path-tracking system demonstrate the advantages of structured neural CLFs, particularly PLNets, and illustrate how their structural bounds influence the geometry of the certified regions. Our code is available at: https://github.com/JLedesMart/Structured-Neural-Control-Lyapunov-Functions.

eess.SY · cs.LG

新论文 · v1arXiv:2611.20900

Joint Learning of Control Lyapunov Functions and Controllers via Stable-by-Design Surrogates

Patricia Pauli, Matthieu Barreau

原文摘要

Training and verifying neural network-based control Lyapunov functions (CLFs) can be computationally demanding. To address this challenge, we propose a framework that jointly learns a CLF and a stabilizing controller using structured neural networks and a projection-based parameterization. Specifically, we employ Polyak-Lojasiewicz networks (PLNets), which parameterize nonconvex, positive-definite CLF candidates, and construct a stable-by-design surrogate of the closed-loop dynamics that satisfies the Lyapunov decrease condition by construction. The controller combines linear state feedback with a higher-order correction term resembling Sontag's universal formula and is trained to align the resulting closed-loop dynamics with the stable-by-design surrogate. Unlike classical approaches based on Sontag's formula, our framework does not require a known CLF; instead, it learns the CLF and controller simultaneously. We establish local and global stability guarantees for the learned controller and demonstrate the effectiveness of the proposed approach through numerical examples.

eess.SY · cs.LG

新论文 · v1arXiv:2611.20889

Comfort as a Constraint or as a Cost in Reinforcement Learning and Model Predictive Building Control

Joram Liebeskind, Valentin Bhend, Varsha N. Behrunani, Alisa Rupenyan

原文摘要

A building controller that keeps a zone within comfort bounds at minimum energy can treat the bounds as a soft constraint, by adding the weighted violation to its cost, or as a hard constraint. We analyze the effect of the hard and the soft formulation of the same problem and isolate the effect of the formulation from the effect of the algorithm. To this end, we measure the difference in energy use between the two formulations on the BOPTEST bestest_hydronic_heat_pump case, in model predictive control (MPC) and in reinforcement learning (RL). The soft formulations are a weighted MPC and weighted SAC and DDPG learners, each at twelve weights. The hard formulations are an MPC with an exact penalty and the Lagrangian learners SAC-Lagrangian and primal-dual DDPG. We find that the hard formulations use more energy than the soft formulations at the largest weight, but violate the comfort bounds less. For a large weight, the soft formulation approaches the behavior of the hard formulation. The frontiers of the energy-comfort trade-off of the three weighted controllers lie close to each other, and the weight selects an operating point on this frontier. Under the soft formulation, the three algorithms reach very similar operating points, although at somewhat different weights. For this HVAC problem, the formulation and the weight thus affect the energy use more than the choice of algorithm.

eess.SY · cs.LG

版本更新 · v2arXiv:2611.00921

Solvable dynamical model of in-context learning

Ruoran Bai, Haiping Huang

原文摘要

Pretrained Transformer architectures display a surprising in-context learning ability without additional training. How the learning process guides the network to achieve this intriguing capability remains unsolved. In this study, we focus on a single-layer single-head linear attention model, where an empirical risk minimization is defined for learning. Given a finite number of training examples, we determine the exact solution of network-parameter dynamics, which is linked to the training error and generalization error, characterizing how memorization and generalization emerge from learning and how these transitions are related to curvatures of the empirical risk landscape. Furthermore, we calculate a data-dependent free energy, from which the training and generalization errors in the long time limit of learning dynamics can be well predicted. Our work adds insights toward understanding the mechanism of in-context learning especially from a learning dynamics perspective.

cs.LG · cond-mat.dis-nn

新论文 · v1arXiv:2611.20771

Nemotron 3.5 Bolt E4B 4GB: An Open and Efficient Small Language Model for Fast Conversational Agents

NVIDIA, :, Adi Margolin, Aditya Malte, Aishwarya Chhabra, Alex Hodson, Ameya Sunil Mahabaleshwarkar, Ambrish Dantrey 等 113 位作者

原文摘要

We introduce Nemotron 3.5 Bolt E4B 4GB, an open and efficient small language model further post-trained from Gemma-4-E4B-it and optimized for speed and intelligence per GB of memory used for fast conversational agents. The model carries forward Gemma's strong conversational capabilities and further improves multi-turn tool use, instruction following, and input grounding for fast, local voice/device assistants, interactive personas, and embedded/Edge interfaces. The model combines natural conversation with multi-turn agentic capabilities to power voice/device assistants, interactive characters and robotic/IoT interfaces running on GeForce RTX, Jetson, and RTX/DGX Spark. Experiments on 21 benchmarks show that Nemotron 3.5 Bolt E4B 4GB achieves the highest average accuracy while generating the fewest output tokens in both reasoning and non-reasoning modes, compared with Gemma-4-E4B-it and Qwen3.5-4B. The model is released with a DFlash drafter. With DFlash, Bolt E4B achieves $2.74\times$ the per-user decode speed of inference without speculative decoding at concurrency~1, and $1.21\times$ the generation throughput of Gemma-4-E4B-it with MTP at batch size 1.

cs.CL · cs.AI · cs.LG

新论文 · v1arXiv:2611.20687

Rethinking Supervision Scaling in On-Policy Distillation: From Dense Supervision to Stochastic Horizons

Xinhao Yao, Zhenxing Dou, Yong Liu

原文摘要

On-policy distillation (OPD) provides dense token-level supervision along student-generated trajectories, yet the learning benefits of longer trajectories may not scale proportionally with their computational cost. We investigate how supervision utility evolves with trajectory length. Controlled experiments reveal diminishing returns from additional supervised tokens, while showing that later positions remain learnable and exhibit substantial cross-position supervision transfer. Gradient analysis further identifies shared update components across response positions and diverse position-specific residuals whose cross-rollout consistency decreases at greater depths. Motivated by these findings, we propose Stochastic Horizons On-Policy Distillation (SH-OPD), which reduces redundant supervision by randomly varying rollout horizons while preserving access to deeper response states. SH-OPD samples a shared horizon for each training batch, enabling efficient parallel generation without per-sequence horizon scheduling overhead. We characterize the resulting reweighting bias and derive a compute-reweighting trade-off under standard optimization assumptions. Experiments on mathematical reasoning and code generation demonstrate that SH-OPD achieves slightly higher mathematical reasoning accuracy at approximately 45% of the per-update cost of standard OPD, while maintaining competitive code generation performance at around 60% of the baseline cost. These results highlight the benefits of efficient supervision allocation across response depths.

cs.LG

新论文 · v1arXiv:2611.20621

GPromptShield: Elevating Resilience in Graph Prompt Tuning against Adversarial Attacks

Shuhan Song, Ping Li, Ming Dun, Maolei Huang, Huawei Cao, Xiaochun Ye

原文摘要

The paradigm of ``pre-training and prompt-tuning", with its effectiveness and lightweight characteristics, has rapidly spread from the language field to the graph field. Several pioneering studies have designed specialized prompt functions for diverse downstream graph tasks based on various graph pre-training strategies. These prompts concentrate on the compatibility between the pre-training pretext and downstream graph tasks, aiming to bridge the gap between them. However, designing prompts blindly to adapt to downstream tasks based on this concept neglects crucial security issues. By conducting covert attacks on downstream graph data, we find that even when the downstream task data closely matches that of the pre-training tasks, it is still feasible to generate highly misleading prompts using simple deceptive techniques. In this paper, we shift the primary focus of graph prompts from compatibility to vulnerability issues in adversarial attack scenarios. We design a highly extensible shield defense system for the prompts, which enhances their robustness from two perspectives: Direct Handling and Indirect Amplification. When downstream graph data contains unreliable biases, the former directly combats invalid information by incorporating hybrid multi-defense prompts to the input graph's feature space, while the latter adopts a training strategy to bypass the invalid components and amplifies valid part. We provide a theoretical derivation that proves their feasibility, indicating that unbiased prompts exist under certain conditions on unreliable data. Extensive experiments across various scenarios of adversarial attacks indicate that the prompts within our defense system exhibit enhanced resilience and superiority. This paper explores a new perspective in graph prompt learning, offering a novel option for robust prompt tuning in downstream tasks.

cs.LG

版本更新 · v2arXiv:2611.06704

Where It Breaks: Diagnosing Perception, Memory, and Decision Failures of a Vision-Language Agent in Pac-Man

Xinglu Wang, Edward Wang, Yong Zhang, Zhenan Fan

原文摘要

Spatial agents must interpret observations, maintain consistent spatial memory and make decisions over long interactions. We use Pac-Man as a testbed to study the training challenges behind these capabilities in a Qwen3.5-9B-based agent. Our research trajectory progresses from primitive-action RL on text-based mazes to visual inputs with an internal-state-assisted harness, then toward observation-driven memory and joint policy--harness improvement. The main analysis focuses on text-based maze RL; supplementary sections report visual-adaptation and harness experiments and discuss subsequent directions. Text-based maze RL requires curriculum and training-recipe design to overcome cold starts; normal-start clear rates on eight held-out mazes rise from 0% to a peak of 81.25%. In the second step, model adaptation and harness improvements raise the clear rate from 26.6% to 96.9% on one fixed map with three extra lives. These results highlight three development lessons: use curricula to address cold starts, establish visual grounding before language adaptation, and treat harness design as a determinant of agent performance. They provide practical guidance for subsequent VLM training and spatial-agent development.

cs.LG

新论文 · v1arXiv:2611.20430

TrainExplorer: Autonomous LLM Post-Training via Plan-and-Decouple Exploration

Xuannan Liu, Tian Lan, Zekun Li, Xing Cui, Zhengfeng Yu, Feihu Jiang, Qianghuai Jia, Longyue Wang 等 12 位作者

原文摘要

LLM-based agents can automate model post-training, accelerating progress toward recursive self-improvement. Despite their promise, existing approaches remain under-explored and face two main challenges: (1) strategy lock-in that limits exploration beyond the initial strategy; and (2) coupled execution that obscures the effects of individual modifications. To address these challanges, we propose TrainExplorer, a multi-agent framework for autonomous post-training, that incorporates strategy planning and execution decoupling, to achieve broaden strategy exploration. First, strategy planning draws on scientific literature to construct a candidate search space of training strategies. Then, decoupled execution enables single-component interventions and attribution to guide strategy exploration. Extensive experiments on PostTrainBench demonstrate that TrainExplorer achieves a weighted aggregate score of 39.02, outperforming Claude Code using the same LLM by 7.32 and surpassing GPT-5.6 Sol-powered Codex, which scores 36.23. Code will be available.

cs.LG · cs.CL

新论文 · v1arXiv:2611.20427

Fuzzy Explainable Continual Hierarchical Reinforcement Learning for Robotic Navigation

Peng Qin, Tao Zhao

原文摘要

Continual robot navigation requires reinforcement learning policies to continuously adapt to new environments while retaining previously acquired knowledge. Meanwhile, explainable reinforcement learning aims to learn meaningful behavioral representations through policy interaction. However, these objectives are inherently conflicting, as continual learning emphasizes policy stability whereas explainable behavior learning requires adaptive exploration. Existing methods rarely address both simultaneously. To resolve this issue, this paper proposes a Fuzzy Explainable Continual Hierarchical Reinforcement Learning (FECHRL) method. A two-stage behavior cloning strategy is introduced to balance rapid policy adaptation and explainable behavior evolution by progressively transitioning from strong imitation guidance to autonomous reinforcement learning. Furthermore, a fuzzy critic based on a Takagi--Sugeno fuzzy inference system is proposed to alleviate value extrapolation errors during the transition between imitation-guided learning and autonomous exploration, thereby improving continual training stability. Experimental results on sequential multi-scenario navigation tasks demonstrate that the proposed method effectively unifies continual learning capability and policy explainability within a single reinforcement learning framework, achieving stable knowledge retention and meaningful behavioral explanation.

cs.LG

版本更新 · v4arXiv:2502.01920

Anomaly Detection via Autoencoder Composite Features and NCE

Yalin Liao, Austin J. Brockmeier

原文摘要

Unsupervised anomaly detection is a challenging task. Autoencoders (AEs) or generative models are often employed to model the data distribution of normal inputs and subsequently identify anomalous, out-of-distribution inputs by high reconstruction error or low likelihood, respectively. However, AEs may generalize and achieve small reconstruction errors on abnormal inputs. We propose a decoupled training approach for anomaly detection that both an AE and a likelihood model trained with noise contrastive estimation (NCE). After training the AE, NCE estimates a probability density function, to serve as the anomaly score, on the joint space of the AE's latent representation combined with features of the reconstruction quality. To further reduce the false negative rate in NCE we systematically varying the reconstruction features to augment the training and optimize the contrastive Gaussian noise distribution. Experimental assessments on multiple benchmark datasets demonstrate that the proposed approach matches the performance of prevalent state-of-the-art anomaly detection algorithms.

cs.LG

新论文 · v1arXiv:2611.20360

Flow Matching for Sampled-Data Control of Angular Distributions

Xin Mao, Anqi Dong, Hongteng Xu, Kenan Zhang

原文摘要

Synchronization and clustering in oscillator networks depend on the joint distribution of the phases. Flow Matching can steer this distribution by learning an angular velocity field from shortest arcs between paired configurations. Each arc is traversed at constant speed, yet the resulting field can vary along its own trajectories. In this article, we study how to control angular transport on the circle and product torus when each controller evaluation selects a rate to hold until the next update. Using the same local velocity samples, Flow Matching learns the instantaneous angular rate and MeanFlow learns its average along the reference trajectory over the control interval. Under direct angular actuation, holding the exact average gives the reference displacement. We relate the displacement error of holding the instantaneous rate and the training residual of MeanFlow to the conditional covariance of the teacher velocities. Experiments with 100 phases show that endpoint pairing affects the comparison. MeanFlow gives smaller one-step errors under minibatch pairing, while permutation alignment reverses the ranking for exchangeable phase clusters.

cs.LG · eess.SY

新论文 · v1arXiv:2611.20325

Measuring Coherence in Predictive Models

Jamie Reason, Vik Shirvaikar, Mark van der Wilk, Chris Holmes

原文摘要

Many sequential decision-making procedures, such as active learning, implicitly assume that a model's current predictions agree with those it expects to make after updating on new data, a property we call coherence. Most models in use do not update by exact conditioning and are not guaranteed to be coherent, which can lead to worse decisions. Yet there is currently no general way to measure incoherence or to tell when it matters. We develop tools to measure incoherence and characterise when it affects decisions. We introduce the incoherence ratio, an MMD-based measure of how far any updating procedure departs from coherence, computed only from the procedure's own predictions and updates, and accompany it with an anytime-valid test of coherence. In Bayesian experimental design, acquisition objectives that are equivalent for coherent models choose different data under every approximate model we study, and the incoherence ratio tracks how far incoherence distorts them. We also introduce action coherence, which captures only the incoherence that changes a decision. What a resulting disagreement costs, however, is determined primarily by predictive accuracy. Sequential decisions depend on how a model updates as well as on what it predicts, so coherence, like accuracy, should be checked rather than assumed.

stat.ML · cs.LG

版本更新 · v2arXiv:2605.19403

TIDE: Asymmetric Neural Circuits for Stabilized Temporal Inhibitory-Excitatory Dynamics

Alexander Kyuroson, Denis Kleyko, Marcus Liwicki

原文摘要

Recent Continuous Thought Machine (CTM) architecture decouples internal computation from external inputs via neural dynamics, but relies on unconstrained multi-layer perceptrons for its recurrent computations. We propose to model neural dynamics using asymmetric Excitatory-Inhibitory (E-I) networks, which can be stabilized via principles from network theory, while obeying Dale's principle and can be expressed as energy-based systems optimized through a game-theoretic loss. Building on this perspective, we introduce Temporal Inhibitory-Excitatory Dynamic Engine (TIDE), a neuro-inspired architecture that computes internal representations through neural dynamics stabilized by incorporating the Wilson-Cowan dynamics and lateral inhibition. TIDE balances biological realism by, for instance, using Hierarchical Receptive Fields and enforcing Dale's principle to ensure a realistic $80:20$ E-I balance ratio with an end-to-end trainable architecture. The aim of this paper is to introduce a new architecture that brings neuro-inspired learning and architectural principles to the forefront. We present proofs of convergence, stability, and complexity bounds, along with empirical ablation studies. Overall, TIDE achieves higher $\texttt{top-1}$ accuracy than CTM on the majority of the investigated datasets while requiring a smaller training budget. Furthermore, across varying training budgets, TIDE improves $\texttt{top-1}$ accuracy on ImageNet-C by an average of $+1.65$% to $+6.42$% under different perturbation severity levels.

cs.LG

新论文 · v1arXiv:2611.20268

GNN-Based Deep Reinforcement Learning for Coastal Emergency Response under Extreme Water Levels

Marcello Sano, Samuele Casagrande, Davide Ferrario, Sebastiano Vascon, Silvia Torresan, Andrea Critto

原文摘要

When a storm surge approaches a coastal town, the emergency plan must be put into action quickly using water-level forecasts, flood extent, road access, vulnerable residents and shelters. Bringing this information together under time pressure is slow and depends on the duty officer's expertise. This paper is one step in a wider research line from data to safe decisions: machine learning to understand the hazard, reinforcement learning and planning to choose actions, large language models (LLMs) to explain and cross-check them, and an assurance layer that keeps people in charge. Here we address the action step on a real place. We ask whether a policy can learn the timing of emergency actions from decades of real events. Lido di Venezia is represented as a heterogeneous graph built from public elevation, road, building and census data and driven by hourly tide-gauge records from 1983 to 2024, with seven actions taken from the municipal emergency plan. A PPO policy is trained on this graph and compared, on storm windows unseen in training, with a rule-based system simulating the officer's discretionary decisions. On 18 historical test scenarios, PPO achieves a similar reduction in population exposure to the rule. With uncertain forecasts and limited protection it again performs similarly but does not outperform the tuned rule, and removing graph message passing does not substantially reduce performance. PPO therefore learns an effective protective response from an outcome-based reward alone: it is given the plan's actions and when each is allowed, but not the thresholds that tell the comparator when to act. Next steps test harder edge cases, move closer to a digital twin of the territory, and replace the simulated officer with assessment by emergency-management stakeholders, supported by LLM-based interpretation of each recommendation.

cs.LG · cs.AI

版本更新 · v2arXiv:2609.14735

HELENA for 5G NR LEO NTN Channel Estimation: A Comparative Evaluation

Miguel Camelo Botero, Nina Slamnik-Kriještorac, Johann Marquez-Barja

原文摘要

Deep Learning (DL)-based channel estimation has shown high accuracy and low latency in terrestrial 5G NR, but Low Earth Orbit (LEO) Non-Terrestrial Networks (NTNs) introduce Doppler and synchronization impairments that may require NTN-specific architectures. We investigate whether HELENA, a dual-attention neural channel estimator originally designed for terrestrial channels, remains effective after NTN retraining and suitable across high-performance and power-constrained inference platforms. Its unchanged architecture is trained on paired receiver-compensated and residual-impaired datasets and evaluated alongside terrestrial-origin, NTN-specific, and classical estimators. HELENA's SNR-averaged Normalized Mean Squared Error (NMSE), computed by arithmetic averaging in linear scale, remains within 0.29 and 0.72 dB of the most accurate learned estimator, AttRNet, in the two conditions, respectively. It outperforms global Linear Minimum Mean Squared Error (LMMSE) under compensation and all evaluated LMMSE variants under residual impairments, although profile-aware filtering achieves lower error under compensation. On an RTX PRO 4500, FP16 TensorRT inference reaches 0.0573 ms 99th-Percentile (P99) latency, below the adopted 0.5 ms estimator budget. On a 10 W Jetson Orin NX, HELENA consumes 66.0% less energy per sample than AttRNet, but its 0.9620 ms P99 exceeds the budget; only the less accurate LSiDNN-48 meets it. Thus, HELENA remains effective without NTN-specific redesign for the evaluated task, while accurate embedded inference remains a latency challenge.

eess.SP · cs.LG · cs.NI · cs.PF

版本更新 · v2arXiv:2611.15917

Owned Edits to a Shared Decision Rule: Learning Private Exceptions as Typed, Gated, Withdrawable Weight Edits

John J. O'Hare

原文摘要

Organisations that apply the same published decision rule often hold private evidence that the rule is wrong for their own cases, and will not pool it. We describe a learning mechanism for that setting whose unit of change is an owned edit. The published rule is compiled exactly into a small ReLU network, a structured prior. Each party's learner searches a small typed space of declarations, a few named values such as subject, scope, form and size, and proposes one only when its own records pass a selection-aware evidence gate. The learning is that gated search around the prior. In two pre-registered analyses, on synthetic data whose 20 hidden truths were sampled from that typed space, the learner at 30 records per party was non-inferior, within 2 points, to all seven tuned baselines (a confined gradient learner, Ditto, FedRep, federated averaging, two pooled networks and pooled boosted trees), scoring 96.6% against 91.7% for the unchanged rule; a descriptive greedy ablation with no prior, over a separate space, scored 84.0%. In a third registered test, typed edits over a network trained on shared records beat that network alone by +5.4 points and the same edits over the published rule by +3.1; oracle-tuned fine-tuning and Ditto matched it at 30 records. An acceptance body recomputes each edit from its declaration and writes it into units reserved for that party, so by construction an edit changes no decision about any other subject, withdraws exactly, discloses at most log2 of the typed space's size in bits, and the artefact can hide nothing beyond its declared values; timing is not covered. We report where it loses: truths shared by all parties, relaxations, ten records, a pre-stated gate criterion that failed in both runs, and a language model given the records that decided about as accurately on one draw. On a rule-only base, a table of owned exceptions makes the same decisions.

cs.LG · cs.CY

新论文 · v1arXiv:2611.20187

Cyber-Secure Wearable, Ultra-Low-Power Networked Sensors for Soldier Health Monitoring: Insights from the CAPSARII project

Luciano Bozzi, Christian Celidonio, Umberto Nuzzi, Massimo Biagini, Stefano Cherubin, Asbjørn Djupdal, Tor Andre Haugdahl, Andrea Aliverti 等 22 位作者

原文摘要

The CAPSARII project advances a cyber-secure, ultra-low-power wearable platform and Internet of Battlefield Things (IoBT) framework to support real-time monitoring of military personnel and data-driven decision support. The architecture integrates smart-textile sensing, a body/squad area mesh, edge-deployed AI on secure RISC-V microcontrollers, and cloud/federated analytics for cross-mission learning and comparative studies. Two representative use cases--pilot fatigue and warfighters operating in harsh environments--guide requirements, validation, and interface design (including HUD and haptic feedback). As an initial step toward embedded models, we benchmark stress classification on a public wearable dataset: tree-ensemble baselines achieve state-of-the-art performance for accuracy and F1 (e.g., XGBoost: $\mathrm{Accuracy}=0.901\!\pm\!0.023$, $\mathrm{F1}=0.920\!\pm\!0.019$), while a feature-based Transformer attains the highest AUC ($0.963\!\pm\!0.025$), informing feature-engineering and compression choices for edge deployment. Building on these results, CAPSARII employs HW/SW co-design, model compression (pruning, quantization, distillation), and precision tuning to meet stringent energy and memory budgets; it further investigates Processing-in-Memory accelerators for future efficiency gains. A structured gap analysis highlights the remaining challenges: establishing validation standards for physiological signal quality and textile durability; achieving low-latency, scalable BAN/SAN communications under mobility and interference; and deploying hybrid lightweight/post-quantum cryptography with robust device identity and side-channel resilience on constrained endpoints. By consolidating these elements into a coherent, secure IoBT stack, CAPSARII delineates a practical roadmap from laboratory baselines to mission-ready soldier monitoring in both training and operational scenarios.

cs.ET · cs.AI · cs.AR · cs.CR · cs.HC · cs.LG

版本更新 · v2arXiv:2505.11621

Benign Overfitting of Two-Layer ReLU Networks in Fixed Dimension through Localized Initialization

Junhyung Park, Patrick Bloebaum, Shiva Prasad Kasiviswanathan

原文摘要

Benign overfitting occurs when a model fits noisy training data exactly and still generalizes well. We establish this behavior for finite-width two-layer networks with the standard ReLU activation, trained by gradient flow in fixed dimension. Our data-independent initialization combines Gaussian weights with neurons initially active only on shrinking spherical caps. For uniform spherical inputs and any fixed bounded conditional response law, training and excess risks vanish in probability uniformly over all sufficiently late times. The parameters also converge, with probability tending to one, to an interpolating limit whose excess risk vanishes in probability. We use one schedule depending only on sample size and dimension. Our proof connects the network to a kernel flow whose limit combines ridge regression with localized spikes that fit the training residuals. We give explicit sample-size, width and training-time bounds.

cs.LG · cs.AI · stat.ML

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