Machine Learning · Cross-listed from cs.RO
Conditional Flow Matching for Continuous Anomaly Detection in Autonomous Driving on a Manifold-Aware Spectral Space
Antonio Guillen-Perez
Safety validation for Level 4 autonomous vehicles (AVs) is currently bottlenecked by the inability to scale the detection of rare, high-risk long-tail scenarios using traditional rule-based heuristics. We present Deep-Flow, an unsupervised framework for safety-critical anomaly detection that utilizes Optimal Transport Conditional Flow Matching (OT-CFM) to characterize the continuous probability density of expert human driving behavior. Unlike standard generative approaches that operate in unstable, high-dimensional coordinate spaces, Deep-Flow constrains the generative process to a low-rank spectral
manifold via a Principal Component Analysis (PCA) bottleneck. This ensures kinematic smoothness by design and enables the computation of the exact Jacobian trace for numerically stable, deterministic log-likelihood estimation. To resolve multi-modal ambiguity at complex junctions, we utilize an Early Fusion
Transformer encoder with lane-aware goal conditioning, featuring a direct skip-connection to the flow head to maintain intent-integrity throughout the network. We introduce a kinematic complexity weighting scheme that prioritizes high-energy maneuvers (quantified via path tortuosity and jerk) during the simulation-free training process. Evaluated on the Waymo Open Motion
Dataset (WOMD), our framework achieves an AUC-ROC of 0.766 against a heuristic golden set of safety-critical events. More significantly, our analysis reveals a fundamental distinction between kinematic danger and semantic non-compliance. Deep-Flow identifies a critical predictability gap by surfacing out-of-distribution behaviors, such as lane-boundary violations and non-normative junction maneuvers, that traditional safety filters overlook. This work provides a mathematically rigorous foundation for defining statistical safety gates, enabling objective, data-driven validation for the safe deployment of autonomous fleets.
Read the full abstract ▾
Machine Learning
Canonicalizing Multimodal Contrastive Representation Learning
Sharut Gupta, Sanyam Kansal, Stefanie Jegelka et al. (+2 more)
As models and data scale, independently trained networks often induce analogous notions of similarity. But, matching similarities is weaker than establishing an explicit correspondence between the representation spaces, especially for
multimodal models, where consistency must hold not only within each modality, but also for the learned image-text coupling. We therefore ask: given two independently trained
multimodal contrastive models (with encoders $(f, g)$ and $(\widetilde{f},\widetilde{g})$) – trained on different distributions and with different architectures – does a systematic
geometric relationship exist between their
embedding spaces? If so, what form does it take, and does it hold uniformly across modalities? In this work, we show that across model families such as CLIP, SigLIP, and FLAVA, this
geometric relationship is well approximated by an orthogonal map (up to a global mean shift), i.e., there exists an orthogonal map $Q$ where $Q^\top Q = I$ such that $\widetilde{f}(x)\approx Q f(x)$ for paired images $x$. Strikingly, the same $Q$ simultaneously aligns the text encoders i.e., $\widetilde{g}(y)\approx Q g(y)$ for texts $y$. Theoretically, we prove that if the
multimodal kernel agrees across models on a small anchor set i.e. $\langle f(x), g(y)\rangle \approx \langle \widetilde{f}(x), \widetilde{g}(y)\rangle$, then the two models must be related by a single orthogonal map $Q$ and the same $Q$ maps images and text across models. More broadly, this finding enables backward-compatible model upgrades, avoiding costly re-
embedding, and has implications for the
privacy of learned representations. Our project page: https://canonical-
multimodal.github.io/
Read the full abstract ▾
All authors Sharut Gupta, Sanyam Kansal, Stefanie Jegelka, Phillip Isola, Vikas Garg
Machine Learning · Cross-listed from cs.DS
Simultaneous Blackwell Approachability and Applications to Multiclass Omniprediction
Lunjia Hu, Kevin Tian and Chutong Yang
Omniprediction is a learning problem that requires suboptimality bounds for each of a family of losses $\mathcal{L}$ against a family of comparator predictors $\mathcal{C}$. We initiate the study of omniprediction in a multiclass setting, where the comparator family $\mathcal{C}$ may be infinite. Our main result is an extension of the recent binary omniprediction
algorithm of [OKK25] to the multiclass setting, with sample complexity (in statistical settings) or regret horizon (in online settings) $\approx \varepsilon^{-(k+1)}$, for $\varepsilon$-omniprediction in a $k$-class prediction problem. En route to proving this result, we design a framework of potential broader interest for solving Blackwell approachability problems where multiple sets must simultaneously be approached via coupled actions.
Read the full abstract ▾
Machine Learning
Be Wary of Your Time Series Preprocessing
Sofiane Ennadir, Tianze Wang, Oleg Smirnov et al. (+2 more)
Normalization and scaling are fundamental preprocessing steps in time series modeling, yet their role in
Transformer-based models remains underexplored from a theoretical perspective. In this work, we present the first formal analysis of how different normalization strategies, specifically instance-based and global scaling, impact the expressivity of
Transformer-based architectures for time series representation learning. We propose a novel expressivity framework tailored to time series, which quantifies a model's ability to distinguish between similar and dissimilar inputs in the representation space. Using this framework, we derive theoretical bounds for two widely used normalization methods: Standard and Min-Max scaling. Our analysis reveals that the choice of normalization strategy can significantly influence the model's representational capacity, depending on the task and data characteristics. We complement our theory with empirical validation on classification and
forecasting benchmarks using multiple
Transformer-based models. Our results show that no single normalization method consistently outperforms others, and in some cases, omitting normalization entirely leads to superior performance. These findings highlight the critical role of preprocessing in time series learning and motivate the need for more principled normalization strategies tailored to specific tasks and
datasets.
Read the full abstract ▾
All authors Sofiane Ennadir, Tianze Wang, Oleg Smirnov, Sahar Asadi, Lele Cao
Machine Learning · Cross-listed from math.ST
Optimal Unconstrained Self-Distillation in Ridge Regression: Strict Improvements, Precise Asymptotics, and One-Shot Tuning
Hien Dang, Pratik Patil and Alessandro Rinaldo
Self-
distillation (SD) is the process of retraining a student on a mixture of ground-truth labels and the teacher's own predictions using the same architecture and training data. Although SD has been empirically shown to often improve generalization, its formal guarantees remain limited. We study SD for ridge regression in unconstrained setting in which the mixing weight $\xi$ may be outside the unit interval. Conditioned on the training data and without any distributional assumptions, we prove that for any squared prediction risk (including out-of-distribution), the optimally mixed student strictly improves upon the ridge teacher for every regularization level $\lambda > 0$ at which the teacher ridge risk $R(\lambda)$ is nonstationary (i.e., $R'(\lambda) \neq 0$). We obtain a closed-form expression for the optimal mixing weight $\xi^\star(\lambda)$ for any value of $\lambda$ and show that it obeys the sign rule: $\operatorname{sign}(\xi^\star(\lambda))=-\operatorname{sign}(R'(\lambda))$. In particular, $\xi^\star(\lambda)$ can be negative, which is the case in over-regularized regimes. To quantify the risk improvement due to SD, we derive exact deterministic equivalents for the optimal SD risk in the proportional asymptotics regime (where the sample and feature sizes $n$ and $p$ both diverge but their aspect ratio $p/n$ converges) under general anisotropic covariance and deterministic signals. Our asymptotic analysis extends standard second-order ridge deterministic equivalents to their fourth-order analogs using block linearization, which may be of independent interest. From a practical standpoint, we propose a consistent one-shot tuning method to estimate $\xi^\star$ without grid search, sample splitting, or refitting. Experiments on real-world
datasets and pretrained neural network features support our theory and the one-shot tuning method.
Read the full abstract ▾
Machine Learning
Revisiting Weight Regularization for Low-Rank Continual Learning
Yaoyue Zheng, Yin Zhang, Joost van de Weijer et al. (+4 more)
Continual Learning (CL) with large-scale pre-trained models (PTMs) has recently gained wide
attention, shifting the focus from training from scratch to continually adapting PTMs. This has given rise to a promising paradigm: parameter-efficient continual learning (PECL), where task interference is typically mitigated by assigning a task-specific module during training, such as low-rank adapters. However, weight regularization techniques, such as Elastic Weight Consolidation (EWC)-a key strategy in CL-remain underexplored in this new paradigm. In this paper, we revisit weight regularization in low-rank CL as a new perspective for mitigating task interference in PECL. Unlike existing low-rank CL methods, we mitigate task interference by regularizing a shared low-rank update through EWC, thereby keeping the storage requirement and inference costs constant regardless of the number of tasks. Our proposed method EWC-LoRA leverages a low-rank representation to estimate parameter importance over the full-dimensional space. This design offers a practical, computational- and
memory-efficient solution for CL with PTMs, and provides insights that may inform the broader application of regularization techniques within PECL. Extensive experiments on various
benchmarks demonstrate the effectiveness of EWC-LoRA, achieving a stability-plasticity trade-off superior to existing low-rank CL approaches. These results indicate that, even under low-rank parameterizations, weight regularization remains an effective mechanism for mitigating task interference. Code is available at: https://github.com/yaoyz96/low-rank-cl.
Read the full abstract ▾
All authors Yaoyue Zheng, Yin Zhang, Joost van de Weijer, Gido M van de Ven, Shaoyi Du, Xuetao Zhang, Zhiqiang Tian
Machine Learning
A Theoretical Framework for Modular Learning of Robust Generative Models
Corinna Cortes, Mehryar Mohri and Yutao Zhong
Training large-scale generative models is resource-intensive and relies heavily on heuristic
dataset weighting. We address two fundamental questions: Can we train Large
Language Models (LLMs) modularly, combining small, domain-specific experts to match monolithic performance, and can we do so robustly for any data mixture, eliminating heuristic tuning? We present a theoretical framework for modular generative modeling where a set of pre-trained experts are combined via a gating mechanism. We define the space of normalized gating functions $\mathcal{G}_{1}$ and formulate the problem as a minimax game to find a single robust gate that minimizes divergence to the worst-case data mixture. We prove the existence of such a robust gate using Kakutani's fixed-point
theorem and show that modularity acts as a strong regularizer, with generalization bounds scaling with the lightweight gate's complexity. Furthermore, we prove that this modular approach can theoretically outperform models retrained on aggregate data, with the gap characterized by the Jensen-Shannon Divergence. Finally, we introduce a scalable
Stochastic Primal-Dual
algorithm and a Structural
Distillation method for efficient inference. Empirical results on synthetic and real-world
datasets confirm that our modular architecture effectively mitigates gradient conflict and can robustly outperform monolithic baselines.
Read the full abstract ▾
Machine Learning
MASPO: Unifying Gradient Utilization, Probability Mass, and Signal Reliability for Robust and Sample-Efficient LLM Reasoning
Xiaoliang Fu, Jiaye Lin, Yangyi Fang et al. (+7 more)
Existing Reinforcement Learning with Verifiable Rewards (RLVR)
algorithms, such as GRPO, rely on rigid, uniform, and symmetric trust region mechanisms that are fundamentally misaligned with the complex optimization dynamics of Large
Language Models (LLMs). In this paper, we identify three critical challenges in these methods: (1) inefficient gradient utilization caused by the binary cutoff of hard clipping, (2) insensitive probability mass arising from uniform ratio constraints that ignore the
token distribution, and (3) asymmetric signal reliability stemming from the disparate credit assignment ambiguity between positive and negative samples. To bridge these gaps, we propose Mass-Adaptive Soft Policy Optimization (MASPO), a unified framework designed to harmonize these three dimensions. MASPO integrates a differentiable soft Gaussian gating to maximize gradient utility, a mass-adaptive limiter to balance exploration across the probability spectrum, and an asymmetric risk controller to align update magnitudes with signal confidence. Extensive evaluations demonstrate that MASPO serves as a robust, all-in-one RLVR solution, significantly outperforming baselines. Our code is available at: https://github.com/FlyTune/MASPO-RL.
Read the full abstract ▾
All authors Xiaoliang Fu, Jiaye Lin, Yangyi Fang, Binbin Zheng, Chaowen Hu, Zekai Shao, Cong Qin, Lu Pan, Ke Zeng, Xunliang Cai
Machine Learning · Cross-listed from cs.CL
Learning to Stay Safe: Adaptive Regularization Against Safety Degradation during Fine-Tuning
Jyotin Goel, Souvik Maji and Pratik Mazumder
Instruction-following
language models are trained to be helpful and safe, yet their safety behavior can deteriorate under benign fine-tuning and worsen under
adversarial updates. Existing defenses often offer limited protection or force a trade-off between safety and utility. We introduce a training framework that adapts regularization in response to safety risk, enabling models to remain aligned throughout fine-tuning. To estimate safety risk at training time, we explore two distinct approaches: a judge-based Safety Critic that assigns high-level harm scores to training batches, and an activation-based risk predictor built with a lightweight classifier trained on intermediate model activations to estimate harmful intent. Each approach provides a risk signal that is used to constrain updates deemed higher risk to remain close to a safe reference policy, while lower-risk updates proceed with standard training. We empirically verify that harmful intent signals are predictable from pre-generation activations and that judge scores provide effective high-recall safety guidance. Across multiple model families and attack scenarios, adaptive regularization with either risk estimation approach consistently lowers attack success rate compared to standard fine-tuning, preserves downstream performance, and adds no inference-time cost. This work demonstrates a principled mechanism for maintaining safety without sacrificing utility.
Read the full abstract ▾
Machine Learning · Cross-listed from math.OC
Adaptive Decentralized Composite Optimization via Three-Operator Splitting
Xiaokai Chen, Ilya Kuruzov and Gesualdo Scutari
The paper studies decentralized optimization over networks, where
agents minimize a sum of locally smooth (strongly) convex losses and plus a nonsmooth convex extended value term. We propose decentralized methods wherein
agents adaptively adjust their stepsize via local backtracking procedures coupled with lightweight min-consensus protocols. Our design stems from a three-operator splitting factorization applied to an equivalent reformulation of the problem. The reformulation is endowed with a new BCV preconditioning metric (Bertsekas-O'Connor-Vandenberghe), which enables efficient decentralized implementation and local stepsize adjustments. We establish robust convergence guarantees. Under mere convexity, the proposed methods converge with a sublinear rate. Under strong convexity of the sum-function, and assuming the nonsmooth component is partly smooth, we further prove linear convergence. Numerical experiments corroborate the theory and highlight the effectiveness of the proposed adaptive stepsize strategy.
Read the full abstract ▾
Machine Learning · Cross-listed from stat.ML
genriesz: A Python Package for Automatic Debiased Machine Learning with Generalized Riesz Regression
Masahiro Kato
Efficient estimation of causal and structural parameters can be automated using the Riesz representation
theorem and debiased machine learning (DML). We present genriesz, an open-source Python package that implements automatic DML and generalized Riesz regression, a unified framework for estimating Riesz representers by minimizing empirical Bregman divergences. This framework includes
covariate balancing, nearest-neighbor matching, calibrated estimation, and density ratio estimation as special cases. A key design principle of the package is automatic regressor balancing (ARB): given a Bregman generator $g$ and a representer model class, genriesz} automatically constructs a compatible link function so that the generalized Riesz regression
estimator satisfies balancing (moment-matching) optimality conditions in a user-chosen basis. The package provides a modulr interface for specifying (i) the target linear functional via a black-box evaluation oracle, (ii) the representer model via basis functions (
polynomial, RKHS approximations, random forest leaf encodings, neural
embeddings, and a nearest-neighbor catchment basis), and (iii) the Bregman generator, with optional user-supplied derivatives. It returns regression adjustment (RA), Riesz weighting (RW), augmented Riesz weighting (ARW), and TMLE-style
estimators with cross-fitting, confidence intervals, and $p$-values. We highlight representative workflows for estimation problems such as the average treatment effect (ATE), ATE on treated (ATT), and average marginal effect estimation. The Python package is available at https://github.com/MasaKat0/genriesz and on PyPI.
Read the full abstract ▾
Machine Learning · Cross-listed from cs.RO
IRIS: Learning-Driven Task-Specific Cinema Robot Arm for Visuomotor Motion Control
Qilong Cheng, Matthew Mackay and Ali Bereyhi
Robotic camera systems enable dynamic, repeatable motion beyond human capabilities, yet their adoption remains limited by the high cost and operational complexity of industrial-grade platforms. We present the Intelligent Robotic Imaging System (IRIS), a task-specific 6-DOF manipulator designed for autonomous, learning-driven cinematic motion control. IRIS integrates a lightweight, fully 3D-printed hardware design with a goal-conditioned visuomotor imitation learning framework based on Action Chunking with
Transformers (ACT). The system learns object-aware and perceptually smooth camera trajectories directly from human demonstrations, eliminating the need for explicit
geometric programming. The complete platform costs under $1,000 USD, supports a 1.5 kg payload, and achieves approximately 1 mm repeatability. Real-world experiments demonstrate accurate trajectory tracking, reliable autonomous execution, and generalization across diverse cinematic motions.
Read the full abstract ▾
Machine Learning
Position: Evaluation of ECG Representations Must Be Fixed
Zachary Berger, Daniel Prakah-Asante, John Guttag et al. (+1 more)
This position paper argues that current benchmarking practice in 12-lead ECG representation learning must be fixed to ensure progress is reliable and aligned with clinically meaningful objectives. The field has largely converged on three public multi-label
benchmarks (PTB-XL, CPSC2018, CSN) dominated by arrhythmia and waveform-morphology labels, even though the ECG is known to encode substantially broader clinical information. We argue that downstream evaluation should expand to include an assessment of structural heart disease and patient-level
forecasting, in addition to other evolving ECG-related endpoints, as relevant clinical targets. Next, we outline evaluation best practices for multi-label, imbalanced settings, and show that when they are applied, the literature's current conclusion about which representations perform best is altered. Furthermore, we demonstrate the surprising result that a randomly initialized encoder with linear evaluation matches state-of-the-art pre-training on many tasks. This motivates the use of a random encoder as a reasonable baseline model. We substantiate our observations with an empirical evaluation of five representative ECG pre-training approaches across six evaluation settings: the three standard
benchmarks, a structural disease
dataset, hemodynamic inference, and patient
forecasting.
Read the full abstract ▾
All authors Zachary Berger, Daniel Prakah-Asante, John Guttag, Collin M. Stultz
Machine Learning
Provably Explaining Neural Additive Models
Shahaf Bassan, Yizhak Yisrael Elboher, Tobias Ladner et al. (+4 more)
Despite significant progress in post-hoc explanation methods for neural networks, many remain heuristic and lack provable guarantees. A key approach for obtaining explanations with provable guarantees is by identifying a cardinally-minimal subset of input features which by itself is provably sufficient to determine the prediction. However, for standard neural networks, this task is often computationally infeasible, as it demands a worst-case exponential number of verification queries in the number of input features, each of which is NP-hard. In this work, we show that for Neural Additive Models (NAMs), a recent and more interpretable neural network family, we can efficiently generate explanations with such guarantees. We present a new model-specific
algorithm for NAMs that generates provably cardinally-minimal explanations using only a logarithmic number of verification queries in the number of input features, after a parallelized preprocessing step with logarithmic runtime in the required precision is applied to each small univariate NAM component. Our
algorithm not only makes the task of obtaining cardinally-minimal explanations feasible, but even outperforms existing
algorithms designed to find the relaxed variant of subset-minimal explanations - which may be larger and less informative but easier to compute - despite our
algorithm solving a much more difficult task. Our experiments demonstrate that, compared to previous
algorithms, our approach provides provably smaller explanations than existing works and substantially reduces the computation time. Moreover, we show that our generated provable explanations offer benefits that are unattainable by standard sampling-based techniques typically used to interpret NAMs.
Read the full abstract ▾
All authors Shahaf Bassan, Yizhak Yisrael Elboher, Tobias Ladner, Volkan Şahin, Jan Kretinsky, Matthias Althoff, Guy Katz
Machine Learning
The Anxiety of Influence: Bloom Filters in Transformer Attention Heads
Peter Balogh
Some
transformer attention heads appear to function as membership testers, dedicating themselves to answering the question "has this
token appeared before in the context?" We identify these heads across four
language models (GPT-2 small, medium, and large; Pythia-160M) and show that they form a spectrum of membership-testing strategies. Two heads (L0H1 and L0H5 in GPT-2 small) function as high-precision membership filters with false positive rates of 0-4% even at 180 unique context
tokens – well above the $d_\text{head} = 64$ bit capacity of a classical Bloom filter. A third head (L1H11) shows the classic Bloom filter capacity curve: its false positive rate follows the theoretical formula $p \approx (1 - e^{-kn/m})^k$ with $R^2 = 1.0$ and fitted capacity $m \approx 5$ bits, saturating by $n \approx 20$ unique
tokens. A fourth head initially identified as a Bloom filter (L3H0) was reclassified as a general prefix-
attention head after confound controls revealed its apparent capacity curve was a sequence-length artifact. Together, the three genuine membership-testing heads form a multi-resolution system concentrated in early layers (0-1), taxonomically distinct from induction and previous-
token heads, with false positive rates that decay monotonically with
embedding distance – consistent with distance-sensitive Bloom filters. These heads generalize broadly: they respond to any repeated
token type, not just repeated names, with 43% higher generalization than duplicate-
token-only heads. Ablation reveals these heads contribute to both repeated and novel
token processing, indicating that membership testing coexists with broader computational roles. The reclassification of L3H0 through confound controls strengthens rather than weakens the case: the surviving heads withstand the scrutiny that eliminated a false positive in our own analysis.
Read the full abstract ▾
Machine Learning
Variational inference via radial transport
Luca Ghafourpour, Sinho Chewi, Alessio Figalli et al. (+1 more)
In variational inference (VI), the practitioner approximates a high-dimensional distribution $\pi$ with a simple surrogate one, often a (product) Gaussian distribution. However, in many cases of practical interest, Gaussian distributions might not capture the correct radial profile of $\pi$, resulting in poor coverage. In this work, we approach the VI problem from the perspective of optimizing over these radial profiles. Our
algorithm radVI is a cheap, effective add-on to many existing VI schemes, such as Gaussian (mean-field) VI and Laplace approximation. We provide theoretical convergence guarantees for our
algorithm, owing to recent developments in optimization over the Wasserstein space–the space of probability distributions endowed with the Wasserstein distance–and new regularity properties of radial transport maps in the style of Caffarelli (2000).
Read the full abstract ▾
All authors Luca Ghafourpour, Sinho Chewi, Alessio Figalli, Aram-Alexandre Pooladian
Machine Learning
LORA-CRAFT: Cross-layer Rank Adaptation via Frozen Tucker Decomposition of Pre-trained Attention Weights
Kasun Dewage, Marianna Pensky, Shankadeep Mondal et al. (+1 more)
We introduce LoRA-CRAFT (Cross-layer Rank Adaptation via Frozen Tucker), abbreviated CRAFT throughout, an extremely parameter-efficient fine-tuning (PEFT) method that applies Tucker tensor decomposition to pre-trained
attention weight
matrices stacked across
transformer layers and trains only small square adaptation
matrices on the resulting frozen Tucker factors. Existing tensor-based PEFT methods decompose gradient updates: LoTR applies Tucker decomposition with shared factor
matrices, while SuperLoRA groups and reshapes $\Delta W$ across layers before applying Tucker decomposition. Separately, methods such as PiSSA apply SVD to pre-trained weights but operate independently per layer. CRAFT bridges these two lines of work: it performs full Tucker decomposition via Higher-Order SVD (HOSVD) directly on pre-trained weights organized as cross-layer 3D tensors, freezes all resulting factors, and adapts the model through lightweight trainable transformations applied to each factor
matrix. Experiments on the GLUE
benchmark using RoBERTa-base and RoBERTa-large, as well as commonsense
reasoning benchmarks using LLaMA2-7B and LLaMA3-8B, demonstrate that CRAFT achieves competitive performance with existing methods while requiring only \rev{extremely low Tucker adaptation parameters}. \fixw{On LLaMA3-8B, CRAFT} \rev{exceeds the average accuracy of LoRA} using hundreds of times fewer parameters\fixw{; on LLaMA2-7B the same holds at a $0.252$M budget}. Our results suggest that CRAFT's efficiency advantage grows with model scale, as the frozen Tucker factors better capture the richer cross-layer structure of larger pre-trained models.
Read the full abstract ▾
All authors Kasun Dewage, Marianna Pensky, Shankadeep Mondal, Suranadi De Silva
Machine Learning
Retrospective In-Context Learning for Temporal Credit Assignment with Large Language Models
Wen-Tse Chen, Jiayu Chen, Fahim Tajwar et al. (+4 more)
Learning from self-sampled data and sparse environmental feedback remains a fundamental challenge in training self-evolving
agents. Temporal credit assignment mitigates this issue by transforming sparse feedback into dense supervision signals. However, previous approaches typically depend on learning task-specific value functions for credit assignment, which suffer from poor sample efficiency and limited generalization. In this work, we propose to leverage pretrained knowledge from large
language models (LLMs) to transform sparse rewards into dense training signals (i.e., the advantage function) through retrospective in-context learning (RICL). We further propose an online learning framework, RICOL, which iteratively refines the policy based on the credit assignment results from RICL. We empirically demonstrate that RICL can accurately estimate the advantage function with limited samples and effectively identify critical states in the environment for temporal credit assignment. Extended evaluation on four BabyAI scenarios show that RICOL achieves comparable convergent performance with traditional online RL
algorithms with significantly higher sample efficiency. Our findings highlight the potential of leveraging LLMs for temporal credit assignment, paving the way for more sample-efficient and generalizable RL paradigms.
Read the full abstract ▾
All authors Wen-Tse Chen, Jiayu Chen, Fahim Tajwar, Hao Zhu, Xintong Duan, Ruslan Salakhutdinov, Jeff Schneider
Machine Learning
Learning with Boolean threshold functions
Veit Elser and Manish Krishan Lal
We develop a method for training neural networks on Boolean data in which the values at all nodes are strictly $\pm 1$, and the resulting models are typically equivalent to networks whose nonzero weights are also $\pm 1$. The method replaces loss minimization with a nonconvex constraint formulation. Each node implements a Boolean threshold function (BTF), and training is expressed through a divide-and-concur decomposition into two complementary constraints: one enforces local BTF consistency between inputs, weights, and output; the other imposes architectural concurrence, equating
neuron outputs with downstream inputs and enforcing weight equality across training-data instantiations of the network. The reflect-reflect-relax (RRR) projection
algorithm is used to reconcile these constraints. Each BTF constraint includes a lower bound on the margin. When this bound is sufficiently large, the learned representations are provably sparse and equivalent to networks composed of simple logical gates with $\pm 1$ weights. Across a range of tasks – including multiplier-
circuit discovery, binary autoencoding, logic-network inference, and cellular automata learning – the method achieves exact solutions or strong generalization in regimes where standard gradient-based methods struggle. These results demonstrate that projection-based constraint satisfaction provides a viable and conceptually distinct foundation for learning in discrete neural systems, with implications for interpretability and efficient inference.
Read the full abstract ▾
Machine Learning
Linear Convergence in Games with Delayed Feedback via Extra Prediction
Yuma Fujimoto, Kenshi Abe and Kaito Ariu
Feedback delays are inevitable in real-world multi-
agent learning. They are known to severely degrade performance, and the convergence rate under delayed feedback is still unclear, even for bilinear games. This paper derives the rate of linear convergence of Weighted Optimistic Gradient Descent-Ascent (WOGDA), which predicts future rewards with extra optimism, in unconstrained bilinear games. To analyze the
algorithm, we interpret it as an approximation of the Extra Proximal Point (EPP), which is updated based on farther future rewards than the classical Proximal Point (PP). Our
theorems show that standard optimism (predicting the next-step reward) achieves linear convergence to the
equilibrium at a rate $\exp(-\Theta(t/m^{5}))$ after $t$ iterations for delay $m$. Moreover, employing extra optimism (predicting farther future reward) tolerates a larger step size and significantly accelerates the rate to $\exp(-\Theta(t/(m^{2}\log m)))$. Our experiments also show accelerated convergence driven by the extra optimism and are qualitatively consistent with our
theorems. In summary, this paper validates that extra optimism is a promising countermeasure against performance degradation caused by feedback delays.
Read the full abstract ▾
Machine Learning
Variational Grey-Box Dynamics Matching
Gurjeet Sangra Singh, Frantzeska Lavda, Giangiacomo Mercatali et al. (+1 more)
Deep generative models such as flow matching and
diffusion models have shown great potential in learning complex distributions and dynamical systems, but often act as black-boxes, neglecting underlying physics. In contrast, physics-based simulation models described by ODEs/PDEs remain interpretable, but may have missing or unknown terms, unable to fully describe real-world observations. We bridge this gap with a novel grey-box method that integrates incomplete physics models directly into generative models. Our approach learns dynamics from observational trajectories alone, without ground-truth physics parameters, in a simulation-free manner that avoids scalability and stability issues of Neural ODEs. The core of our method lies in modelling a structured variational distribution within the flow matching framework, by using two latent encodings: one to model the missing stochasticity and multi-modal velocity, and a second to encode physics parameters as a latent variable with a physics-informed prior. Furthermore, we present an adaptation of the framework to handle second-order dynamics. Our experiments on representative ODE/PDE problems and real-world weather
forecasting demonstrate that our method performs on par with or superior to fully data-driven approaches and previous grey-box baselines, while preserving the interpretability of the physics model. Our code is available at https://github.com/DMML-Geneva/VGB-DM.
Read the full abstract ▾
All authors Gurjeet Sangra Singh, Frantzeska Lavda, Giangiacomo Mercatali, Alexandros Kalousis
Machine Learning · Cross-listed from cs.CL
ABCD: All Biases Come Disguised
Mateusz Nowak, Xavier Cadet and Peter Chin
Multiple-choice question (MCQ)
benchmarks have been a standard evaluation practice for measuring LLMs' ability to reason and answer knowledge-based questions. Through a synthetic NonsenseQA
benchmark, we observe that different LLMs exhibit varying degrees of label-position-few-shot-
prompt bias, where the model either uses the answer position, the label in front of the answer, the distributions of correct answers present in the few-shot
prompt, or a combination of all to answer each MCQ question. We propose a simple bias-reduced evaluation protocol that replaces the labels of each question with uniform, unordered labels and
prompts the LLM to use the whole answer presented. With a simple sentence similarity model, we demonstrate improved robustness and lower standard deviation between different permutations of answers with a minimal drop in LLM's performance, exposing the LLM's capabilities under reduced evaluation artifacts, without any help from the
prompt examples or the option labels. Across multiple
benchmarks and models, this protocol substantially improves the robustness to answer permutations, reducing mean accuracy variance $3\times$ with only a minimal decrease in the mean model's performance. Through ablation studies on various
embedding models and similarity functions, we show that the method is more robust than the standard ones.
Read the full abstract ▾
Machine Learning · Cross-listed from cs.CL
Fine-Grained Uncertainty Quantification for Long-Form Language Model Outputs: A Comparative Study
Dylan Bouchard, Mohit Singh Chauhan, Viren Bajaj et al. (+1 more)
Uncertainty quantification has emerged as an effective approach to closed-book
hallucination detection for LLMs, but existing methods are largely designed for short-form outputs and do not generalize well to long-form generation. We introduce a taxonomy for fine-grained uncertainty quantification in long-form LLM outputs that distinguishes methods by design choices at three stages: response decomposition, unit-level scoring, and response-level aggregation. We formalize several families of consistency-based black-box scorers, providing generalizations and extensions of existing methods. We also introduce FactScore-STEM-Geo, a new 400-question long-form QA
dataset spanning four categories across STEM and Geography. In our experiments across multiple LLMs and
datasets, we find 1) claim-response entailment consistently performs better or on par with more complex claim-level scorers, 2) claim-level scoring generally yields better results than sentence-level scoring, and 3) uncertainty-aware decoding is highly effective for improving the factuality of long-form outputs. Our framework clarifies relationships between prior methods, enables apples-to-apples comparisons, and provides practical guidance for selecting components for fine-grained UQ.
Read the full abstract ▾
All authors Dylan Bouchard, Mohit Singh Chauhan, Viren Bajaj, David Skarbrevik
Machine Learning
Convergence Analysis of Two-Layer Neural Networks under Gaussian Input Masking
Afroditi Kolomvaki, Fangshuo Liao, Evan Dramko et al. (+2 more)
We investigate the convergence guarantee of two-layer neural network training with Gaussian randomly masked inputs. This scenario corresponds to Gaussian dropout at the input level, or noisy input training common in sensor networks,
privacy-preserving training, and
federated learning, where each user may have access to partial or corrupted features. Using a Neural Tangent Kernel (NTK) analysis, we demonstrate that training a two-layer ReLU network with Gaussian randomly masked inputs achieves linear convergence up to an error region proportional to the mask's variance. A key technical contribution is resolving the randomness within the non-linear activation, a problem of independent interest.
Read the full abstract ▾
All authors Afroditi Kolomvaki, Fangshuo Liao, Evan Dramko, Ziyun Guang, Anastasios Kyrillidis
Machine Learning · Cross-listed from cs.CV
SpectralGCD: Spectral Concept Selection and Cross-modal Representation Learning for Generalized Category Discovery
Lorenzo Caselli, Marco Mistretta, Simone Magistri et al. (+1 more)
Generalized Category Discovery (GCD) aims to identify novel categories in unlabeled data while leveraging a small labeled subset of known classes. Training a parametric classifier solely on image features often leads to overfitting to old classes, and recent
multimodal approaches improve performance by incorporating textual information. However, they treat modalities independently and incur high computational cost. We propose SpectralGCD, an efficient and effective
multimodal approach to GCD that uses CLIP cross-modal image-concept similarities as a unified cross-modal representation. Each image is expressed as a mixture over semantic concepts from a large task-agnostic dictionary, which anchors learning to explicit semantics and reduces reliance on spurious visual cues. To maintain the semantic quality of representations learned by an efficient student, we introduce Spectral Filtering which exploits a cross-modal covariance
matrix over the softmaxed similarities measured by a strong teacher model to automatically retain only relevant concepts from the dictionary. Forward and reverse knowledge
distillation from the same teacher ensures that the cross-modal representations of the student remain both semantically sufficient and well-aligned. Across six
benchmarks, SpectralGCD delivers accuracy comparable to or significantly superior to state-of-the-art methods at a fraction of the computational cost. The code is publicly available at: https://github.com/miccunifi/SpectralGCD.
Read the full abstract ▾
All authors Lorenzo Caselli, Marco Mistretta, Simone Magistri, Andrew D. Bagdanov
Machine Learning
MDP Planning as Policy Inference
David Tolpin
We formulate episodic Markov decision process (MDP) planning as
Bayesian inference over policies. The primary contribution is conceptual: the policy itself is treated as the latent variable, and expected return defines an unnormalized posterior density over policies. This preserves the standard expected-return objective, in contrast to trajectory-centric planning-as-inference formulations that introduce auxiliary optimality variables and to
entropy-regularized policy optimization methods that solve a different objective. In the exact formulation, the posterior over deterministic policies induces what we define here as an optimal
stochastic policy under preference uncertainty, namely the
stochastic policy induced by that posterior. For discrete MDPs with
stochastic transitions, we study variational sequential Monte Carlo (VSMC) as one approximate inference method for this posterior, introducing policy consistency under state revisitation and coupled transition randomness across particles. Experiments on grid worlds, Blackjack, Triangle Tireworld, and Academic Advising examine the consequences of inference over policies and compare its induced behavior with
entropy-regularized policy optimization. The results support the view that MDP planning can be naturally cast as
Bayesian inference over policies.
Read the full abstract ▾
Machine Learning
A feature-stable and explainable machine learning framework for trustworthy decision-making under incomplete clinical data
Justyna Andrys-Olek, Paulina Tworek, Luca Gherardini et al. (+4 more)
Machine learning models are increasingly applied to biomedical data, yet their adoption in high stakes domains remains limited by poor robustness, limited interpretability, and instability of learned features under realistic data perturbations, such as missingness. In particular, models that achieve high predictive performance may still fail to inspire trust if their key features fluctuate when data completeness changes, undermining reproducibility and downstream decision-making. Here, we present CACTUS (Comprehensive Abstraction and Classification Tool for Uncovering Structures), an explainable machine learning framework explicitly designed to address these challenges in small, heterogeneous, and incomplete clinical
datasets. CACTUS integrates feature abstraction, interpretable classification, and systematic feature stability analysis to quantify how consistently informative features are preserved as data quality degrades. Using a real-world haematuria cohort comprising 568 patients evaluated for bladder cancer, we
benchmark CACTUS against widely used machine learning approaches, including random forests and gradient boosting methods, under controlled levels of randomly introduced missing data. We demonstrate that CACTUS achieves competitive or superior predictive performance while maintaining markedly higher stability of top-ranked features as missingness increases, including in sex-stratified analyses. Our results show that feature stability provides information complementary to conventional performance metrics and is essential for assessing the trustworthiness of machine learning models applied to biomedical data. By explicitly quantifying robustness to missing data and prioritising interpretable, stable features, CACTUS offers a generalizable framework for trustworthy data-driven decision support.
Read the full abstract ▾
All authors Justyna Andrys-Olek, Paulina Tworek, Luca Gherardini, Mark W. Ruddock, Mary Jo Kurt, Peter Fitzgerald, Jose Sousa
Machine Learning
2Mamba2Furious: Linear in Complexity, Competitive in Accuracy
Gabriel Mongaras and Eric C. Larson
Linear
attention transformers have become a strong alternative to softmax
attention due to their efficiency. However, linear
attention tends to be less expressive and results in reduced accuracy compared to softmax
attention. To bridge the accuracy gap between softmax
attention and linear
attention, we manipulate Mamba-2, a very strong linear
attention variant. We first simplify Mamba-2 down to its most fundamental and important components, evaluating which specific choices make it most accurate. From this simplified Mamba variant (Mamba-2S), we improve the A-mask and increase the order of the hidden state, resulting in a method, which we call 2Mamba, that is nearly as accurate as softmax
attention, yet much more
memory efficient for long context lengths. We also investigate elements to Mamba-2 that help surpass softmax
attention accuracy. Code is provided for all our experiments.
Read the full abstract ▾
Machine Learning
Shortcut learning in geometric knot classification
Djordje Mihajlovic and Davide Michieletto
Classifying the
topology of closed curves is a central problem in low dimensional
topology with applications beyond mathematics spanning
protein folding,
polymer physics and even magnetohydrodynamics. The central problem is how to determine whether two
embeddings of a closed arc are equivalent under ambient isotopy. Given the striking ability of neural networks to solve complex classification tasks, it is therefore natural to ask if the
knot classification problem can be tackled using Machine Learning (ML). In this paper, we investigate generic shortcut methods employed by ML to solve the
knot classification challenge and specifically discover hidden non-topological features in training data generated through Molecular Dynamics simulations of polygonal
knots that are used by ML to arrive to positive classifications results. We then provide a rigorous foundation for future attempts to tackle the
knot classification challenge using ML by developing a publicly-available (i)
dataset, that aims to remove the potential of non-topological feature classification and (ii) code, that can generate
knot embeddings that faithfully explore chosen
geometric state space with fixed
knot topology. We expect that our work will accelerate the development of ML models that can solve complex
geometric knot classification challenges.
Read the full abstract ▾
Machine Learning · Cross-listed from cs.DM
Partial Optimality in the Preordering Problem
David Stein, Jannik Irmai and Bjoern Andres
Preordering is a generalization of clustering and partial ordering with applications in bioinformatics and social network analysis. Given a finite set $V$ and a value $c_{ab} \in \mathbb{R}$ for every ordered pair $ab$ of elements of $V$, the preordering problem asks for a preorder $\lesssim$ on $V$ that maximizes the sum of the values of those pairs $ab$ for which $a \lesssim b$. Building on the state of the art in solving this NP-hard problem partially, we contribute new partial optimality conditions and efficient
algorithms for deciding these conditions. In experiments with real and synthetic data, these new conditions increase, in particular, the fraction of pairs $ab$ for which it is decided efficiently that $a \not\lesssim b$ in an optimal preorder.
Read the full abstract ▾