Machine Learning
Spectral Souping: A Unified Framework for Online Preference Alignment
Yinlam Chow, Guy Tennenholtz, Ted Yun et al. (+4 more)
Reinforcement Learning from Human Feedback (RLHF) effectively aligns Large
Language Models (LLMs) with aggregate human preferences but often fails to address the diverse and conflicting needs of individual users. To overcome this issue, we introduce Spectral Souping, a unified framework for efficient, online preference
alignment. Our contribution is the discovery of a universal spectral representation within LLMs, which is proven to be highly amenable to model merging. This theoretical insight enables a two-phase methodology: we first learn a basis of specialized policies offline, each focused on a distinct, fine-grained preference dimension. An online adaptation
algorithm then efficiently “soups” these policies at inference time, either by merging their outputs or parameters, enabling rapid model adaptation without the need for costly online retraining w.r.t. tailored preference rewards. Experiments on online preference
alignment benchmarks demonstrate that our method achieves significant performance improvements over existing state-of-the-art approaches, presenting a scalable and computationally efficient solution for dynamically adapting LLMs to individual user preferences.
Read the full abstract ▾
All authors Yinlam Chow, Guy Tennenholtz, Ted Yun, James Harrison, Arthur Gretton, Andre Barreto, Bo Dai
Machine Learning
Decomposing MXFP4 quantization error for LLM reinforcement learning: reducible bias, recoverable deadzone, and an irreducible floor
Xiaocan Li, Shiliang Wu and Zheng Shen
MXFP4 arithmetic can dramatically accelerate reinforcement learning (RL) post-training of large
language models (LLMs), yet the quantization error introduces severe accuracy degradation. Existing work treats the quantization error as a monolithic noise term, missing the distinct mechanisms upon interpreting how quantization error damages training. We prove an exact three-way decomposition of quantization error and show how each component dominates a distinct RL training pathway. Our theoretical and empirical analysis decomposes the MXFP4 quantization error into three additive components: "scale bias" from power-of-two rounding, "deadzone truncation" from zeroing small values, and "grid noise" from rounding to the nearest 4-bit grid. Each component dominates a distinct RL failure mode: scale bias accumulates multiplicatively through the backward pass, affecting gradient accuracy; deadzone truncation degrades rollout quality; and grid noise raises the policy's
entropy. We combine corrections that are RL failure mode-targeted but not component-exclusive: Macro-block scaling to reduce scale bias, Outlier Fallback recovers deadzone entries, but also partially reduces scale bias induced error, and Adaptive Quantization Noise (AQN) for controlling the policy
entropy. On Qwen2.5-3B dense and Qwen3-30B-A3B-Base mixture-of-experts model, the targeted corrections recover BF16 accuracy to within 0.7% and exceed BF16 by +1.0% respectively.
Read the full abstract ▾
Machine Learning · Cross-listed from stat.AP
Understanding Deterioration Random Effects for Causal Discovery in Infrastructure Management
Takato Yasuno
Infrastructure deterioration poses significant challenges for asset management, yet existing approaches rely on population-averaged models that overlook equipment-specific heterogeneity. We present a novel framework that combines
Bayesian hierarchical hazard modeling with causal discovery to identify operational patterns that drive heterogeneous deterioration rates in pump equipment. Our approach first estimates pump-specific random effects $u_i$ using GPU-accelerated No-U-Turn Sampling (NUTS), achieving 3–5$\times$ speedup over CPU implementations. We then employ DirectLiNGAM to discover causal relationships between 22 engineered time-series features and deterioration rates, stratified by positive ($u_i > 0$, faster deterioration) versus negative ($u_i \leq 0$, slower deterioration) random effects. Analyzing 112 pumps with 92,861 observations over 650 days, we uncover striking heterogeneity: the negative group exhibits causal effects 400$\times$ larger than the positive group, with standard deviation (std) showing a strong positive causal effect ($+1.515$) on deterioration rates in low-risk equipment. We validate linearity assumptions through NonlinearLiNGAM comparison and demonstrate practical scalability through GPU acceleration. Our findings enable targeted maintenance strategies by revealing that different operational regimes require fundamentally distinct management approaches, advancing predictive maintenance from population-averaged to heterogeneity-aware decision making.
Read the full abstract ▾
Machine Learning
Score-Based Causal Discovery of Latent Variable Causal Models
Ignavier Ng, Xinshuai Dong, Haoyue Dai et al. (+3 more)
Identifying latent variables and the causal structure involving them is essential across various scientific fields. While many existing works fall under the category of constraint-based methods (with e.g. conditional independence or rank deficiency tests), they may face empirical challenges such as testing-order dependency, error propagation, and choosing an appropriate significance level. These issues can potentially be mitigated by properly designed score-based methods, such as Greedy Equivalence Search (GES) (Chickering, 2002) in the specific setting without latent variables. Yet, formulating score-based methods with latent variables is highly challenging. In this work, we develop score-based methods that are capable of identifying causal structures containing causally-related latent variables with identifiability guarantees. Specifically, we show that a properly formulated scoring function can achieve score equivalence and consistency for structure learning of latent variable causal models. We further provide a characterization of the degrees of freedom for the marginal over the observed variables under multiple structural assumptions considered in the literature, and accordingly develop both exact and continuous score-based methods. This offers a unified view of several existing constraint-based methods with different structural assumptions. Experimental results validate the effectiveness of the proposed methods.
Read the full abstract ▾
All authors Ignavier Ng, Xinshuai Dong, Haoyue Dai, Biwei Huang, Peter Spirtes, Kun Zhang
Machine Learning · Cross-listed from cs.CR
Latent Geometry as a Structural Monitor: Eigenspace Alignment for Anomaly Detection in Anonymity Networks
Vaibhav Chhabra
Traditional anomaly detection marks events when measured signals cross predefined thresholds. This captures the moment of transition but not the structural pressure that precedes it. We propose treating large behavioral populations as
geometric energy landscapes whose deformation can be measured before and during major transitions. The central thesis is that structure precedes geometry: the structural organization of the population is the signal, and
geometric metrics are instruments for measuring it. Applied to the Tor anonymity network across 67 consecutive daily observation windows, the dual-observer pipeline identifies a stable nine-dimensional load-bearing subspace invariant across the observation period and validates this structure by Monte Carlo simulation at 16.8 sigma above the noise floor. Primary detection gates achieve 0.0% false positive rate on 24 confirmed stable windows. Forensic analysis of the February 20, 2026 confirmed infrastructure event formally falsifies the relay-departure hypothesis, identifying connectivity degradation without
topology change as a detectable network failure mode. The result is a candidate structural-monitoring framework for behavioral populations with sufficient telemetry.
Read the full abstract ▾
Machine Learning · Cross-listed from cs.CV
STELLAR: Scaling 3D Perception Large Models for Autonomous Driving
Yingwei Li, Xin Huang, Yang Liu et al. (+13 more)
Model scaling has demonstrated remarkable success through large-scale training on diverse
datasets. It remains an open question whether the same paradigm would apply to autonomous driving perception systems due to unique challenges, such as fusing heterogeneous sensor data and the need for sophisticated 3D spatial understanding. To bridge this gap, we present a comprehensive study on systematically analyzing the impact of scale on these systems. We develop our
STELLAR model based on Sparse Window
Transformer, by extending the input modalities to include
LiDAR, radar, camera, and map prior. We train the model on a large-scale
dataset of 50 million driving examples with up to 500 million parameters. Our large-scale experiments reveal empirical scaling trends that connect model performance to model size, data, and compute. The resulting model establishes a new state-of-the-art on the Waymo Open
Dataset challenge, outperforming prior arts by a large margin. Our work demonstrates that large-scale training is a highly promising path for advancing the capabilities of perception models for autonomous driving.
Read the full abstract ▾
All authors Yingwei Li, Xin Huang, Yang Liu, Yang Fu, Alex Zihao Zhu, Chen Song, Junwen Yao, Anant Subramanian, Hao Xiang, Weijing Shi, Yuliang Zou, Tom Hoddes, Zhaoqi Leng, Govind Thattai, Dragomir Anguelov, Mingxing Tan
Machine Learning
Nonlocal operator learning for fMRI encoding and decoding tasks
Andreas Kramer, Saugat Acharya, Alice Giola et al. (+1 more)
Functional MRI data exhibit high-dimensional spatiotemporal structure, making both prediction and decoding challenging. In this work, we investigate neural integral-operator-based models for encoding and decoding tasks in fMRI, with particular emphasis on the role of nonlocal spatiotemporal context. We implement a latent neural integral operator framework that performs fixed point iterations in an auxiliary space from which classification and stimuli prediction is performed via a decoder. We evaluate our model on two open-source fMRI
datasets. Our experiments examine both decoding of stimuli from fMRI recordings and encoding of fMRI dynamics from stimulus representations. A main focus is the effect of spatiotemporal context: we systematically compare short and long temporal windows, as well as the use of visual cortex vs whole
brain recordings, and analyze their influence on performance and latent-space geometry. Across tasks and
datasets, larger temporal windows generally improve results and produce more structured learned representations. In decoding experiments, the learned latent space often provides clearer class separation than the raw data. In encoding experiments, although absolute performance remains moderate due to the difficulty of the task, longer temporal windows still yield consistent gains. These findings suggest that neural integral operators provide a promising framework for modeling fMRI dynamics and that broader spatiotemporal context can be beneficial for both prediction and representation learning. More broadly, the results indicate that exploiting distributed nonlocal structure in
brain dynamics requires model architectures specifically designed to capture such dependencies.
Read the full abstract ▾
All authors Andreas Kramer, Saugat Acharya, Alice Giola, Emanuele Zappala
Machine Learning · Cross-listed from cs.CL
DEL: Digit Entropy Loss for Numerical Learning of Large Language Models
Zhaohui Zheng, Chenhang He, Shihao Wang et al. (+3 more)
Number prediction stands as a fundamental capability of large
language models (LLMs) in mathematical problem-solving and code generation. The widely adopted maximum likelihood estimation (MLE) for LLM training is not tailored to number prediction. Recently, penalty-driven approaches, e.g., Number
Token Loss and Discretized Distance Loss, introduce an inductive bias of numerical distance but induce over-sharpened and over-flattened digit distributions, respectively. In this paper, we make an in-depth analysis on LLM numerical learning, and show that existing numerical learning methods conceptually follow a criterion-distance formulation, where the criterion term represents optimization pattern and the distance term instills
geometric prior. Consequently, we present Digit
Entropy Loss (DEL) for auto-regressive numerical learning, which reformulates the conventional unsupervised
entropy optimization in three key designs: leveraging digit conditional probability and binary cross-
entropy to guide the
entropy optimization into a supervised manner; deprecating the distance term to bypass the issue of numerical distance; and generalizing the integer-based numerical learning to floating-point number optimization, enabling more accurate number prediction. Our DEL formulation can incorporate integers, decimals, and decimal points, expanding the learning objective from a single digit to the floating-point number domain. Experiments conducted on seven mathematical
reasoning benchmarks with four representative LLMs, including CodeLlama, Mistral, DeepSeek, and Qwen-2.5, demonstrate that DEL consistently outperforms its counterparts in both overall prediction accuracy and numerical distance. Source codes are at https://github.com/PolyU-VCLab/DEL
Read the full abstract ▾
All authors Zhaohui Zheng, Chenhang He, Shihao Wang, Yuxuan Li, Ming-Ming Cheng, Lei Zhang
Machine Learning
Consistently Informative Soft-Label Temperature for Knowledge Distillation
Hoang-Chau Luong, Nghia Van Vo, Kaiqi Zhao et al. (+1 more)
Knowledge
distillation (KD) transfers knowledge from a high-capacity teacher to a compact student by matching their predictive distributions, with temperature scaling serving as a central mechanism for smoothing teacher predictions and exposing informative "dark knowledge" beyond the hard label. However, the standard fixed-temperature design is inherently sample-agnostic. Since samples differ in logit scale and learning difficulty, a single global temperature produces teacher soft labels with highly inconsistent
entropy: some predictions remain overly sharp and provide limited inter-class information, whereas others become over-smoothed and lose class-discriminative information. Moreover, sharing the same temperature between teacher and student further imposes rigid logit-scale
alignment despite their capacity mismatch. To address these limitations, we propose CIST (Consistently Informative Soft-label Temperature), which assigns separate sample-wise adaptive temperatures to the teacher and student. This design produces consistently informative teacher soft labels while relaxing rigid teacher–student logit-scale matching. It also reweights the
distillation objective according to teacher confidence and student learning difficulty. Theoretically, we show that teacher-label
entropy is largely governed by the ratio between the maximum teacher logit and the temperature, providing a principled basis for adaptive smoothing. Empirically, CIST mitigates the inconsistency induced by fixed temperature, and experiments on both
vision and
language distillation tasks show consistent improvements over standard KD and strong baselines with negligible computational overhead.
Read the full abstract ▾
All authors Hoang-Chau Luong, Nghia Van Vo, Kaiqi Zhao, Lingwei Chen
Machine Learning · Cross-listed from cs.RO
Proximal State Nudging: Reducing Skill Atrophy from AI Assistance
Megha Srivastava, Jonathan Ouyang, Eric Zhou et al. (+6 more)
Skill atrophy, the gradual decline of human capability under AI assistance, poses a safety risk in shared-control of semi-autonomous systems, where operators may be unable to distinguish their own inputs from autonomous corrections. We propose Proximal State Nudging (PSN), a shared autonomy
algorithm that jointly optimizes for skill development and task performance by nudging users toward states estimated to be most learnable. We first show that PSN outperforms existing shared autonomy baselines in balancing student improvement in unassisted reward with overall shared performance, using simulated students in the classic LunarLander environment. We then present, to the best of our knowledge, the first human subject studies of a planner incorporating learning-compatible shared autonomy: across two driving tasks in the CARLA simulator (High Performance Racing and Parallel Parking, n = 60), PSN produces up to 7x larger gains in unassisted skill than standard blended shared autonomy, while incurring 50% fewer collisions than unassisted self-practice.
Read the full abstract ▾
All authors Megha Srivastava, Jonathan Ouyang, Eric Zhou, Andrew Silva, Emily Sumner, Dorsa Sadigh, Yuchen Cui, Deepak Gopinath, Guy Rosman
Machine Learning
Symmetrization of Loss Functions for Robust Training of Neural Networks in the Presence of Noisy Labels
Alexandre Lemire Paquin, Brahim Chaib-Draa and Philippe Giguère
Labeling a training set is often expensive and susceptible to errors, making the design of robust loss functions for label noise an important problem. The
symmetry condition provides theoretical guarantees for robustness to such noise. In this work, we study a symmetrization method arising from the unique decomposition of any multi-class loss function into a symmetric component and a class-insensitive term. In particular, symmetrizing the cross-
entropy loss leads to a linear multi-class extension of the unhinged loss. Unlike in the binary case, the multi-class version must have specific coefficients in order to satisfy the
symmetry condition. Under suitable assumptions, we show that this multi-class unhinged loss is the unique convex multi-class symmetric loss. We also show that it has a fundamental local role: the linear approximation of any symmetric loss around score vectors with equal components is equivalent to the multi-class unhinged loss. We then introduce SGCE and alpha-MAE, two loss functions that interpolate between the multi-class unhinged loss and the Mean Absolute Error while allowing control of the beta-smoothness of the loss. Experiments on standard noisy-label
benchmarks show competitive performance compared with existing robust loss functions.
Read the full abstract ▾
Machine Learning · Cross-listed from stat.ML
Corrected Integrated Laplace Approximation for Bayesian Inference in Latent Gaussian Models
Jinlin Lai, Charles C. Margossian and Daniel R. Sheldon
Latent Gaussian models (LGMs) are a popular class of
Bayesian hierarchical models that include Gaussian processes, as well as certain spatial models and mixed-effect models. Efficient
Bayesian inference of LGMs often requires marginalizing out the latent variables. For LGMs with a non-Gaussian likelihood, exact marginalization is not possible and a popular approach is to do approximate marginalization with an integrated Laplace approximation (ILA). Using ILA produces an approximate posterior which, in some settings, can differ significantly from the correct posterior, which impacts downstream applications. We propose an importance sampling scheme to correct the error introduced by ILA. By increasing the number of samples in importance sampling, the posterior with ILA converges to the correct posterior. This idea is realized with various techniques, including pseudo-marginalization, quasi-Monte Carlo and randomized quasi-Monte Carlo. We implement our methods in an automatic differentiation framework to support gradient-based
algorithms when doing inference on the hyperparameters. For the latter, we specifically consider the use of
Hamiltonian Monte Carlo. We demonstrate the benefits of reduced error in various applied models.
Read the full abstract ▾
Machine Learning
Causal Unlearning in Collaborative Optimization: Exact and Approximate Influence Reversal under Adversarial Contributions
Ali Mahdavi, Azadeh Zamanifar, Amirfarhad Farhadi et al. (+1 more)
Federated learning systems must support data deletion requests to comply with
privacy regulations, yet retraining from scratch after each deletion is computationally prohibitive. We present HF-KCU, a method that removes a client's contribution by approximating the influence function through conjugate gradient iterations in Krylov subspaces, reducing complexity from O(d^3) to O(kd) where k<<d.A causal weighting mechanism ensures that only clients holding the deleted data receive parameter updates, preventing spurious changes to unaffected clients. Our method is designed to handle bounded
adversarial perturbations to the Hessian and gradient, providing graceful degradation under realistic threat models. We validate HF-KCU across convolutional (ResNet-18, SimpleCNN) and
transformer (ViT-Lite) architectures on CIFAR-10, MNIST, and Fashion-MNIST. On CIFAR-10 under Dirichlet (alpha=0.5) partitioning, HF-KCU achieves 47.75 times speedup over retraining while maintaining test accuracy within 0.60% of the rational baseline(71.16 vs 71.76 %). Membership inference attacks on the forget set yield success rates of 0.499 matching the retrained model and confirming effective
privacy restoration. We provide convergence guarantees showing that the Krylov approximation error decreases as O((k ^1/2-1)/(k^1/2+1)) where k is the Hessian condition number. The causal weighting mechanism ensures
surgical updates, where only clients holding deleted data are modified, preserving model quality for unaffected participants and avoiding the instability of gradient-based approaches in asynchronous
federated settings. This design provides interpretability as each update is directly traceable to the influence of the deleted data. The method's efficiency and precision make it suitable for production
federated systems where deletion requests arrive asynchronously and computational budgets are constrained.
Read the full abstract ▾
All authors Ali Mahdavi, Azadeh Zamanifar, Amirfarhad Farhadi, Omid Kashefi
Machine Learning
Less Data, Faster Training: repeating smaller datasets speeds up learning via sampling biases
Jingwen Liu, Ezra Edelman, Surbhi Goel et al. (+1 more)
This work investigates the “small-vs-large gap”, where repeating on fewer samples can lead to compute saving during training compared to using a larger
dataset. This is observed across algorithmic tasks, architectures and optimizers and cannot be explained using prior theory. We argue that the speedup comes from appropriate layer-wise growth enabled by sampling biases, which is more pronounced when the
dataset size is smaller. We provide both theoretical analysis and empirical evidence from various interventions. Our results suggest that using a smaller
dataset with more repetitions is not just a fallback strategy under data scarcity, but can be proactively leveraged as a favorable inductive biases for optimization, particularly in
reasoning tasks.
Read the full abstract ▾
All authors Jingwen Liu, Ezra Edelman, Surbhi Goel, Bingbin Liu
Machine Learning
WaveGraphNet: Physics-Consistent Guided-Wave Damage Localization through Coupled Inverse-Forward Graph Learning
Vinay Sharma, Aditya Bharade and Olga Fink
Guided-wave structural health monitoring enables damage localization in composite plates using sparse networks of bonded piezoelectric transducers. However, supervised localization remains weakly constrained when measurements are available from only a limited set of damage locations. Because exhaustive spatial coverage is impractical, models trained at observed locations may generalize poorly to unseen regions. We propose WaveGraphNet, an inverse-forward graph-learning framework for guided-wave damage localization in carbon-fiber-reinforced
polymer (CFRP) plates. The sensing network is represented as a graph, with transducers as nodes and measured pitch-catch paths as edges. The inverse branch uses order-invariant message passing and a geometry-constrained decoder to map path-wise energy deviations to a damage coordinate. An independently trained forward branch predicts the energy-deviation pattern associated with a candidate coordinate. At test time, gradients propagated through the forward branch update only the inverse-predicted coordinate to reduce the discrepancy between the measured and predicted response patterns. The framework is evaluated on the OGW-1
benchmark using three spatial hold-out splits in which complete damage regions are excluded from training. Comparisons include non-graph and graph-learning baselines together with the training-free Reconstruction
Algorithm for Probabilistic Inspection of Damage (RAPID). Under the validation-selected settings, refined WaveGraphNet achieves the lowest reported test mean absolute error on all three splits. Refinement reduces the standalone localization error by 23.6 percent-67.9 percent, with an average relative reduction of 50.4 percent. These results demonstrate that learned forward-response consistency can improve guided-wave localization under limited spatial training coverage.
Read the full abstract ▾
Machine Learning · Cross-listed from cs.CV
SDM: A Powerful Tool for Evaluating Model Robustness
Xinlei Liu, Tao Hu, Jichao Xie et al. (+3 more)
Gradient-based attacks are important methods for evaluating model robustness. However, since the proposal of APGD, it has been difficult for such methods to achieve significant breakthroughs. To achieve such an effect, we first analyze the issue of "high-loss non-
adversarial examples" that degrades attack performance in previous methods, and prove that this issue arises from inappropriate objectives for
adversarial example generation. Subsequently, we reconstruct the objective as "maximizing the difference between the non-ground-truth label probability upper bound and the ground-truth label probability", and proposes a novel and powerful gradient-based attack method named Sequential Difference Maximization (SDM). SDM establishes a three-layer optimization framework of "cycle-stage-step". It adopts the negative probability loss function and the Directional Probability Difference Ratio (DPDR) loss function in the initial and subsequent optimization stages, respectively, and approaches the ideal objective of
adversarial example generation via stage-wise sequential optimization. Experiments demonstrate that compared with previous state-of-the-art methods, SDM not only achieves stronger attack performance but also exhibits superior cost-effectiveness. The code is available at https://github.com/X-L-Liu/ICML-SDM.
Read the full abstract ▾
All authors Xinlei Liu, Tao Hu, Jichao Xie, Peng Yi, Hailong Ma, Baolin Li
Machine Learning · Cross-listed from cs.CV
WildRoadBench: A Wild Aerial Road-Damage Grounding Benchmark for Vision-Language Models and Autonomous Agents
Bingnan Liu, Chenhang Cui, Rui Huang et al. (+7 more)
We introduce WildRoadBench, a wild aerial road-damage grounding
benchmark that couples direct visual grounding by
vision-
language models with autonomous research-and-engineering by LLM-driven
agents on a single professionally annotated
UAV corpus. The same image set and the same per-class AP_50 metric are evaluated under two protocols. The VLM Track measures whether a fixed VLM can localise domain-specific damage from one image and one short
prompt under a unified prompting, decoding and parsing pipeline. The
Agent Track measures whether an autonomous
agent, given only a written task brief, a small exploratory slice and a fixed interaction budget, can search the public web, adapt pretrained components, write training and inference code, and submit predictions through a scalar-feedback oracle on a hidden holdout. We
benchmark a broad pool of closed-source frontier models and open-source VLMs together with several frontier LLM-driven
agents. Both routes remain far from reliable performance in this wild setting: closed-source frontier models lead the VLM leaderboard but still leave more than half of the metric on the table; open-source grounders plateau well below them, and newer generations or
reasoning-style variants do not consistently improve grounding; small targets collapse for every open-source model;
agents lag the strongest VLM despite richer affordances, and several fail to land a valid submission within the budget. We release the code and data at https://anonymous.4open.science/r/wildroadbench-0607 to support reproducible follow-up research.
Read the full abstract ▾
All authors Bingnan Liu, Chenhang Cui, Rui Huang, Jiani Luo, Zhirong Shen, Tinghao Wang, Xiande Huang, Lingbei Meng, Fei Shen, An Zhang
Machine Learning
AirfoilGen: A valid-by-construction and performance-aware latent diffusion model for airfoil generation
Zhijie Yang, Min Tang, Peng Du et al. (+1 more)
Airfoil shape design is a fundamental task in aerospace engineering, with a direct impact on flight stability and fuel consumption. Deep learning has recently emerged as a promising tool for this task, but existing deep generative approaches remain limited in both
geometric validity and physical controllability. They offer little control over the generated shapes, yielding invalid geometries, and they typically do not condition effectively on aerodynamic performance. To address these issues, this paper proposes AirfoilGen, a valid-by-construction and performance-aware latent
diffusion model for airfoil. It first introduces a novel airfoil representation scheme, the circle sweeping representation, to constrain the generative process so that output shapes respect essential airfoil characteristics. It then enables explicit control over aerodynamic performance (e.g., lift and drag coefficients) by operating in a learned latent space: a
transformer model encodes airfoil shapes into vector
embeddings, and a conditional
diffusion model denoises Gaussian noise into these latent
embeddings while incorporating target aerodynamic performance. In addition, this paper presents a new
dataset of over 200,000 airfoils, which is substantially larger than the widely used UIUC airfoil
dataset (1,650 airfoils) and more suitable for training modern deep generative models. Experiments demonstrate that AirfoilGen enables airfoil generation with far greater
geometric validity and aerodynamic performance controllability than previously achievable, with an average performance-conditioning accuracy of 98.41%.
Read the full abstract ▾
All authors Zhijie Yang, Min Tang, Peng Du, Qiang Zou
Machine Learning
Neural Collapse by Design: Learning Class Prototypes on the Hypersphere
Panagiotis Koromilas, Theodoros Giannakopoulos, Mihalis A. Nicolaou et al. (+1 more)
Supervised classification has a theoretical optimum, Neural Collapse (NC), yet neither of its two dominant paradigms reaches it in practice. Cross
entropy (CE) leaves radial degrees of freedom unconstrained and converges to a degenerate geometry, while supervised contrastive learning (SCL) drives features toward NC during pretraining but discards this structure in a post hoc linear probing phase. We show that both paradigms are different appearances of the same method that contrasts prototypes on the unit hypersphere, and that closing the gap requires fixing each at its point of failure. From the CE side, we propose NTCE and NONL, two normalized losses that import contrastive optimization's missing ingredients into classifier learning: a large effective negative set and decoupled
alignment and uniformity terms. From the SCL side, we prove that SCL's objective already optimizes throughout training for a principled classifier whose weights are the class mean
embeddings, making linear probing both redundant and harmful. Empirically, on four
benchmarks including ImageNet-1K, NTCE and NONL surpass CE accuracy, closely approximate NC ($\geq 95\%$), and match CE's converged NC on 4/5 metrics in under $7.5\%$ of its iterations, while SCL with fixed prototypes matches linear probing without the hours-long classifier training phase. The learned geometry yields $+5.5\%$ mean relative improvement in transfer learning, up to $+8.7\%$ under severe class imbalance, and improved robustness to corruptions on ImageNet-C. Our work recasts supervised learning as prototype learning on the hypersphere, with NC reached by design.
Read the full abstract ▾
All authors Panagiotis Koromilas, Theodoros Giannakopoulos, Mihalis A. Nicolaou, Yannis Panagakis
Machine Learning
Robust Subspace-Constrained Quadratic Models for Low-Dimensional Structure Learning
Zheng Zhai and Xiaohui Li
In this paper, we propose a robust subspace-constrained quadratic model (SCQM) for learning low-dimensional structure from high-dimensional data. Building upon the subspace-constrained quadratic
matrix factorization (SQMF) framework, the proposed model accommodates a broad class of noise distributions, including generalized Gaussian and radial Laplace models. This generalization enables reliable performance under both heavy-tailed and light-tailed noise, thereby substantially enhancing robustness across diverse data regimes. To efficiently address the resulting nonconvex optimization problem, we develop a gradient-based
algorithm equipped with a backtracking line-search strategy that ensures stable and efficient convergence. In addition, we present a sensitivity analysis of the $\ell_p^p$ and $\ell_2$ loss functions, elucidating their distinct behaviors under varying noise characteristics. Extensive numerical experiments corroborate the theoretical analysis and demonstrate that the proposed approach consistently outperforms existing methods in terms of robustness and reconstruction accuracy.
Read the full abstract ▾
Machine Learning
Mechanisms of Misgeneralization in Physical Sequence Modeling
Kento Nishi, Raphael Tang, Karun Kumar et al. (+2 more)
Generative sequence models are often trained to plan motion in physical domains, from robotics to mechanical simulations. When constructing a
dataset to train such a model, engineers may curate demonstrations to specify how trajectories should be distributed over a physical quantity like travel distance or mechanical energy. For example, a roboticist building a maze
navigation agent might choose demonstrations whose travel distances cover a fixed range uniformly, hoping to constrain the
agent's expected power usage. We find that standard deep learning can violate this intent: each generated trajectory can seem plausible on its own, but the aggregate distribution over the physical quantity is wrong. We call this failure physical misgeneralization, and develop an account of its mechanism. Using controlled synthetic tasks, we show that physical misgeneralization arises when local errors typical of the model class propagate through the physical measurement to shift the recovered distribution. We estimate these errors with a data deviation kernel, and we use it to predict which physical quantities gain or lose mass in both our synthetic and more applied maze
navigation and double-pendulum motion tasks. Finally, our mechanistic interpretation helps identify which mitigation strategies are structurally promising, and we use it to propose a kernel-informed intervention.
Read the full abstract ▾
All authors Kento Nishi, Raphael Tang, Karun Kumar, Core Francisco Park, Hidenori Tanaka
Machine Learning · Cross-listed from cs.CV
MedCRP-CL: Continual Medical Image Segmentation via Bayesian Nonparametric Semantic Modality Discovery
Ziyuan Gao
Medical image
segmentation faces a fundamental challenge in continual learning: data arrives sequentially from heterogeneous sources, yet effective continual learning requires discovering which tasks share sufficient structure to benefit from joint learning. Existing methods either apply uniform constraints across all tasks, causing catastrophic forgetting when tasks conflict, or require predefined task groupings that cannot anticipate future task diversity. We introduce MedCRP-CL, a framework that performs online task structure discovery and structure-aware continual learning. Leveraging the Chinese Restaurant Process (CRP), our method dynamically infers task groupings from clinical text
prompts as tasks arrive, without requiring predefined cluster counts or access to future tasks. We term these discovered groupings semantic modalities, as they capture finer-grained structure than physical imaging modalities by integrating anatomical region and pathological context. Guided by this discovered structure, we maintain semantic modality-specific LoRA adapters regularized by intra-modality EWC, ensuring parameter isolation across dissimilar task groups while facilitating knowledge transfer within similar ones. The framework is also replay-free, storing only aggregate statistics rather than raw patient data. Experiments on 16 medical
segmentation tasks across four imaging modalities demonstrate that MedCRP-CL achieves 73.3% Dice score with only 4.1% forgetting, outperforming the best baseline by 8.0% while requiring 6$\times$ fewer parameters. Code is available at https://github.com/zygao930/MedCRP-CL.
Read the full abstract ▾
Machine Learning
Spectral Unforgetting: Post-Hoc Recovery of Damaged Capabilities Without Retraining
Aarash Abro and Muhammad Tahir
Fine-tuning a
language model for a target task routinely degrades capabilities the training data never explicitly threatened. We study this phenomenon, known as catastrophic forgetting, and propose a post-hoc repair solution that uses only the pretrained checkpoint $W_{\mathrm{base}}$ and its fine-tuned descendant $W_{\mathrm{ft}}$. The goal is not merely to revert the model toward the base checkpoint, but to recover capabilities damaged by fine-tuning while preserving both the target-task gains and any beneficial held-out improvements. We introduce DG-Hard, a checkpoint-only spectral repair method for the fine-tuning update $\Delta = W_{\mathrm{ft}} - W_{\mathrm{base}}$. DG-Hard treats $\Delta$ as a low-rank task-aligned signal embedded in an IID-like noise residual that gradient descent has no incentive to remove, and applies the Donoho-Gavish hard singular-value threshold to each weight-delta
matrix, keeping the structured high-energy part of the update and removing the spectral bulk. This reduces repair to a closed-form SVD filtering step requiring no data-dependent tuning. A central difficulty is evaluation: average accuracy hides per-
benchmark failures, while naive recovery scores reward models that simply revert toward the base. We therefore introduce a partition-conditional metric that separately tracks healing, preservation, non-damage, and target-task retention. Across $14$ (model, task) settings and nine cross-domain held-out
benchmarks, DG-Hard achieves the strongest balanced repair among post-hoc baselines. DG-Hard also restores safety
alignment degraded by benign fine-tuning on three independent safety axes, despite using no
alignment data. These results suggest that part of fine-tuning-induced capability loss is not an unavoidable consequence of specialization, but a removable spectral residue in the weight update itself.
Read the full abstract ▾
Machine Learning
Quant.npu: Enabling Efficient Mobile NPU Inference for on-device LLMs via Fully Static Quantization
Jinghe Zhang, Daliang Xu, Chenghua Wang et al. (+5 more)
Large
language models (LLMs) are increasingly deployed on mobile devices, where Neural Processing Units (NPUs) necessitate fully static quantization for optimal inference efficiency. However, existing post-training quantization (PTQ) methods predominantly rely on dynamic activation quantization, rendering them incompatible with NPU hardware constraints. To bridge the gap between high-fidelity PTQ and NPU-constrained inference, we propose Quant.npu, a integer-only fully static quantization framework. It incorporates learnable quantization parameters and rotation
matrices, enabling low-bit activation-weight quantization without runtime quantization parameters re-computation. Crucially, we identify that initialization and selective optimization of quantization parameters is pivotal for optimization stability, as improper initialization and naive joint optimization induce gradient instability that disrupts the optimization of rotation
matrices. To address this, we propose a rotation-and-bit-width-aware initialization tailored to diverse activation profiles and a distribution-aware selective optimization (two-stage quantization pipeline) tailored to rotated and unrotated tensors. Furthermore, we introduce a sensitivity-guided adaptive mixed-precision scheme to balance accuracy with inference efficiency. Extensive experiments on real-world mobile NPUs demonstrate that Quant.npu achieves comparable accuracy to state-of-the-art methods, while reducing inference latency by up to 15.1%.
Read the full abstract ▾
All authors Jinghe Zhang, Daliang Xu, Chenghua Wang, Weikai Xie, Tao Qi, Yun Ma, Mengwei Xu, Gang Huang
Machine Learning
Closed-form predictive coding via hierarchical Gaussian filters
Aleksandrs Baskakovs, Sylvain Estebe, Kenneth Enevoldsen et al. (+3 more)
Predictive coding (PC) offers a local and biologically grounded alternative to backpropagation in the training of artificial neural networks, yet to date, it remains slower, and performance degrades sharply as network depth increases. We trace both problems to a single simplification: current PC networks fix the precision
matrix to the identity, discarding precision-weighted prediction errors that the variational derivation requires to be fast, local, and
Bayesian. We close this gap by expressing predictive coding networks as deep hierarchical Gaussian filters (HGFs) and restore precision-weighted message passing, yielding dynamic uncertainty estimates and Hebbian-compatible update rules at every layer. The resulting networks can simultaneously learn activations, weights, and precisions under a single free-energy objective, with no global error signal, and resolve inference without requiring iterations or automatic differentiation. On FashionMNIST, our solution approaches backpropagation in epoch-level wall-clock cost while converging in fewer epochs, and outperforms it on online, data efficiency, and concept-drift tasks. We thus establish that closed-form variational inference with online precision learning provides a tractable foundation for deep predictive coding networks, retaining biological and interpretative advantages, without requiring iterative relaxation or global error signals.
Read the full abstract ▾
All authors Aleksandrs Baskakovs, Sylvain Estebe, Kenneth Enevoldsen, Kristoffer Nielbo, Chris Mathys, Nicolas Legrand
Machine Learning
Learning to Select Source-Traceable Evidence for Language-Model Prediction from Irregular Clinical Time Series
Kwanhyung Lee, Juhwan Choi, Jongheon Kim et al. (+5 more)
Numerical time-series models effectively process irregular electronic health record (EHR) trajectories, but do not expose which temporal patterns support each prediction as readable evidence. Existing text-based interfaces either serialize observations, preserving source traceability but offering limited clinical interpretation, or generate patient-level summaries that improve readability but can obscure links to source measurements. We introduce STEP-CTS (Source-Traceable Evidence for Prediction from Clinical Time-Series), which learns to select source-traceable text evidence for a
language-model predictor. Multi-scale window statistics of each trajectory are verbalized as sets of deterministic threshold predicates, each set linked to its source window (e.g., "[5-7 h]: last temperature at least 38C"). An offline LLM, run once per unique predicate set without access to patient records, the prediction task, or outcome labels, attaches clinical concepts such as "fever" to their supporting predicates and abstains when none applies. Each predicate set, together with its concepts, forms an evidence unit. A learned Evidence Selector selects a fixed-size subset of the evidence units, which a pretrained clinical
language-model encoder reads to make the prediction, with no patient-level text generation. Across three ICU
benchmarks, STEP-CTS outperforms evaluated text-based baselines, improving AUPRC over the strongest by 5.1, 3.7, and 11.7 percentage points on P2012, MIMIC-III, and P2019, respectively, and is competitive with dedicated numerical time-series models. Ablations show that clinical concepts and learned evidence selection each contribute to predictive performance. In a blinded clinician study, the selected evidence is rated as traceable as deterministic serialization, near the ceiling of the scale, and above generated summaries on clinical interpretation.
Read the full abstract ▾
All authors Kwanhyung Lee, Juhwan Choi, Jongheon Kim, Joohyung Lee, Hyeongwon Jang, Jeonguk Lee, Jisoo Jung, Eunho Yang
Machine Learning
Weasel: Out-of-Domain Generalization for Web Agents via Importance-Diversity Data Selection
Fatemeh Pesaran Zadeh, Seyeon Choi, Xing Han Lù et al. (+2 more)
Large
language models (LLMs) have enabled web
agents that follow natural
language goals through multi-step browser interactions. However,
agents fine-tuned on specific trajectories and domain often struggle to generalize out of domain, and offline training can be compute-inefficient due to noisy, redundant trajectories and long accessibility-tree (AXTree) states. To address both issues, we propose Weasel, a trajectory selection method for offline training of web
agents. Weasel selects a fixed-budget subset of trajectory steps by optimizing an objective that balances unary importance with pairwise diversity over states, websites, and interaction patterns, solving efficiently with a greedy
algorithm. We further improve efficiency with target-centered AXTree
pruning that keeps only content around the ground-truth action target, and we mitigate style mismatch for
reasoning-native models by replacing expert traces with model-generated, style-consistent rationales. Across AgentTrek and NNetNav training
datasets, evaluations in WebArena, WorkArena, and MiniWob, and experiments with Qwen2.5-7B, Gemma3-4B, and Qwen3-8B, Weasel improves out-of-domain performance while reducing training cost, producing roughly 9.7-12.5$\times$ training speedups over standard fine-tuning. We make the code available at https://github.com/fatemehpesaran310/weasel.
Read the full abstract ▾
All authors Fatemeh Pesaran Zadeh, Seyeon Choi, Xing Han Lù, Siva Reddy, Gunhee Kim
Machine Learning
Plug-and-Play Spiking Operators: Breaking the Nonlinearity Bottleneck in Spiking Transformers
Xinzhe Yuan, Xiang Peng, Bin Gu et al. (+1 more)
ANN-to-SNN conversion offers a practical, training-free route to spiking large
language models. However, current pipelines primarily focus on spike-driven realizations for
Transformer linear-
algebra operations, while providing limited support for key nonlinear operators. This gap limits compatibility with neuromorphic-style execution constraints, where such nonlinearities typically require division, exponentiation, or norm computations that are not naturally supported by standard leaky integrate-and-fire dynamics. To solve this problem, we propose a plug-and-play framework that implements spike-friendly approximations for
Transformer nonlinearities and integrates into existing ANN-to-SNN pipelines. Our method decomposes these nonlinear computations into three recurring primitives – division, exponentiation, and $\ell_2$ norms – and realizes them via population computation using LIF
neuron groups, combined with lightweight bit-shift scaling to avoid floating-point arithmetic. By composing these primitives as modular operator blocks, our framework supports common
Transformer nonlinearities (e.g., Softmax, SiLU, and normalization) without any fine-tuning. Experiments on a range of LLMs
Transformers show that selectively replacing the targeted nonlinear operators incurs less than a $1\%$ accuracy drop across all evaluated tasks.
Read the full abstract ▾
All authors Xinzhe Yuan, Xiang Peng, Bin Gu, Huan Xiong
Machine Learning
FusionCell: Cross-Attentive Fusion of Layout Geometry and Netlist Topology for Standard-Cell Performance Prediction
Haoyi Zhang, Kairong Guo, Bojie Zhang et al. (+2 more)
Standard
cells form the building blocks of digital
circuits, so their delay and power critically influence chip-level performance; yet characterization still relies on slow simulation sweeps, and many fast predictors ignore layout geometry, missing coupling and layout-dependent effects. The challenge is to jointly represent layout geometry and netlist
topology so models capture fine-grained spatial details together with structural connectivity for accurate performance prediction. We introduce FusionCell, a dual-modality predictor that treats routed layout geometry and netlist
topology as inputs and fuses them explicitly in a unified model. A DeiT encoder processes three-layer routed layouts, while a graph
transformer models heterogeneous device/net graphs. The modalities are integrated through a
topology-guided mechanism, where the netlist acts as a structural "map" to actively query relevant physical regions in the layout for joint
geometric and topological
reasoning. We build a 7nm
dataset based on the ASAP7 PDK with over 19.5k
cells spanning 149 types using automatic tools, targeting six metrics: signal rise/fall delay, transition, and power. Experimental results demonstrate that FusionCell reduces regression error, with an average MAPE of 0.92 percent, and improves Spearman/Kendall ranking over baselines, while accelerating the characterization process by orders of magnitude compared to
circuit simulation.
Read the full abstract ▾
All authors Haoyi Zhang, Kairong Guo, Bojie Zhang, Yibo Lin, Runsheng Wang
Machine Learning · Cross-listed from cs.CR
Adaptive Probe-based Steering for Robust LLM Jailbreaking
Junxi Chen, Junhao Dong and Xiaohua Xie
Recent work has demonstrated the potential of contrastive steering for jailbreaking Large
Language Models (LLMs). However, existing methods rely on limited and inherently biased contrastive
prompts and require laborious manual tuning of steering strength, limiting their robustness and effectiveness. In this paper, we leverage the idea of model extraction to guide the learned steering vectors to approximate the ideal one and propose tuning the steering strength adaptively based on contrastive activations' statistics. Experiments demonstrate that our method notably improves the effectiveness and robustness of probe-based steering, without any extra contrastive
prompts or laborious manual tuning. Being an attack paper, this paper focuses on revealing the breakdown of fortified LLMs, raising the average harmfulness score from 6% to 70%. Our code is available at https://github.com/fhdnskfbeuv/adaptiveSteering.
Read the full abstract ▾