Peer Revenue Equity in science

Papers

Fresh from arXiv, typeset properly. Highlighted words trade on the market.

Computer Science Machine Learning

cs.LG Watch Machine Learning (sign in to use)

Sunday 28 December

Machine Learning

Breaking the Memory Wall: Exact Analytical Differentiation via Tiled Operator-Space Evolution

Selective State Space Models (SSMs) achieve linear-time inference, yet their gradient-based sensitivity analysis remains bottlenecked by O(L) memory scaling during backpropagation. This memory constraint precludes genomic-scale modeling (L > 10^5) on consumer-grade hardware. We introduce Phase Gradient Flow (PGF), a framework that computes exact analytical derivatives by operating directly in the state-space manifold, bypassing the need to materialize the intermediate computational graph. By reframing SSM dynamics as Tiled Operator-Space Evolution (TOSE), our method delivers O(1) memory complexity relative to sequence length, yielding a 94% reduction in peak VRAM and a 23x increase in throughput compared to standard Autograd. Unlike parallel prefix scans that exhibit numerical divergence in stiff ODE regimes, PGF ensures stability through invariant error scaling, maintaining near-machine precision across extreme sequences. We demonstrate the utility of PGF on an impulse-response benchmark with 128,000-step sequences - a scale where conventional Autograd encounters prohibitive memory overhead, often leading to out-of-memory (OOM) failures in multi-layered models. Our work enables chromosome-scale sensitivity analysis on a single GPU, bridging the gap between theoretical infinite-context models and practical hardware limitations.
Read the full abstract

All authors Shuhuan Wang, Yuzhen Xie, Jiayi Li, Yinliang Diao

Machine Learning · Cross-listed from cs.AI

The Reward Model Selection Crisis in Personalized Alignment

Personalized alignment from preference data has focused primarily on improving personal reward model (RM) accuracy, with the implicit assumption that better preference ranking translates to better personalized behavior. However, in deployment, computational constraints necessitate inference-time adaptation such as reward-guided decoding (RGD) rather than per-user policy fine-tuning. This creates a critical but overlooked requirement: reward models must not only rank preferences accurately but also effectively guide generation. We demonstrate that standard RM accuracy fails catastrophically as a selection criterion for deployment-ready personalized rewards. We introduce policy accuracy; a metric quantifying whether RGD-adapted LLMs correctly discriminate between preferred and dispreferred responses and show that upstream RM accuracy correlates only weakly with downstream policy accuracy (Kendall's tau = 0.08–0.31). More critically, we introduce Pref-LaMP the first personalized alignment benchmark with ground-truth user completions, enabling direct behavioural evaluation. On Pref-LaMP, we expose a complete decoupling between discriminative ranking and generation metrics: methods with 20-point RM accuracy differences produce almost identical output quality, and methods with high ranking accuracy can fail to generate behaviorally aligned responses. These findings reveal that the field has been optimizing for proxy metrics that do not predict deployment performance, and that current personalized alignment methods fail to operationalize preferences into behavioral adaptation under realistic deployment constraints. In contrast, we find simple in-context learning (ICL) to be highly effective - dominating all reward-guided methods for models $\geq$3B parameters, achieving $\sim$3 point ROUGE-1 gains over the best reward method at 7B scale.
Read the full abstract

All authors Fady Rezk, Yuangang Pan, Chuan-Sheng Foo, Xun Xu, Nancy Chen, Henry Gouk, Timothy Hospedales

Machine Learning

PI-MFM: Physics-informed multimodal foundation model for solving partial differential equations

Partial differential equations (PDEs) govern a wide range of physical systems, and recent multimodal foundation models have shown promise for learning PDE solution operators across diverse equation families. However, existing multi-operator learning approaches are data-hungry and neglect physics during training. Here, we propose a physics-informed multimodal foundation model (PI-MFM) framework that directly enforces governing equations during pretraining and adaptation. PI-MFM takes symbolic representations of PDEs as the input, and automatically assembles PDE residual losses from the input expression via a vectorized derivative computation. These designs enable any PDE-encoding multimodal foundation model to be trained or adapted with unified physics-informed objectives across equation families. On a benchmark of 13 parametric one-dimensional time-dependent PDE families, PI-MFM consistently outperforms purely data-driven counterparts, especially with sparse labeled spatiotemporal points, partially observed time domains, or few labeled function pairs. Physics losses further improve robustness against noise, and simple strategies such as resampling collocation points substantially improve accuracy. We also analyze the accuracy, precision, and computational cost of automatic differentiation and finite differences for derivative computation within PI-MFM. Finally, we demonstrate zero-shot physics-informed fine-tuning to unseen PDE families: starting from a physics-informed pretrained model, adapting using only PDE residuals and initial/boundary conditions, without any labeled solution data, rapidly reduces test errors to around 1% and clearly outperforms physics-only training from scratch. These results show that PI-MFM provides a practical and scalable path toward data-efficient, transferable PDE solvers.
Read the full abstract

All authors Min Zhu, Jingmin Sun, Zecheng Zhang, Hayden Schaeffer, Lu Lu

Machine Learning

Mechanistic Evidence for Preserved-but-Misaligned Representations in Non-IID FedAvg

Federated Averaging (FedAvg) often degrades under non-IID client data, but it remains unclear whether this degradation reflects the loss of client-learned representations or a failure to use representations that are still present. We study this question mechanistically in sparse client-trained vision models, using dense-model controls to test whether the observed effects depend on sparsity. Our analysis combines class-specific circuit discovery, linear probing of frozen representations, head-only finetuning, and sparse feature dictionaries. Across CNN and ResNet models on CIFAR-10 and Fashion-MNIST, severe label skew can drive some per-class accuracies near zero even when class-specific internal structure remains recoverable. Linear probes substantially outperform the aggregated classifier, head-only finetuning partially restores accuracy, and USAE transfer reveals a largely shared feature basis between IID and non-IID models. Together, these diagnostics suggest that, in our setting, non-IID FedAvg degradation is not fully explained by representational erasure; it also reflects misalignment between preserved internal structure and the final prediction pathway.
Read the full abstract

All authors Muhammad Haseeb, Salaar Masood, Muhammad Abdullah Sohail, Mohammad Fatim Shoaib, Muhammad Tahir

Machine Learning · Cross-listed from cs.CL

Is Chain-of-Thought Really Not Explainability? Chain-of-Thought Can Be Faithful without Hint Verbalization

Recent work, using the Biasing Features metric, labels a CoT as unfaithful if it omits a prompt-injected hint that affected the prediction. We argue this metric adopts a narrow notion of faithfulness and confuses unfaithfulness with incompleteness, the lossy compression needed to turn distributed transformer computation into a linear natural language narrative. On multi-hop reasoning tasks with instruct-tuned and reasoning models, many CoTs flagged as unfaithful by Biasing Features are judged faithful by other metrics, exceeding 50% in some models. With a new faithful@k metric, we show that larger inference-time budgets greatly increase hint verbalization (up to 90% in some settings), suggesting much apparent unfaithfulness is due to tight token limits. Using Causal Mediation Analysis, we further show that even non-verbalized hints can causally mediate prediction changes through the CoT. We therefore caution against relying solely on hint-based evaluations and advocate a broader interpretability toolkit, including causal mediation and corruption-based metrics. We do not claim all CoTs are faithful, only that the absence of hint words alone does not prove unfaithfulness.
Read the full abstract

Machine Learning

Merge before Forget: A Single LoRA Continual Learning via Continual Merging

Parameter-efficient continual learning has emerged as a promising approach for large language models (LLMs) to mitigate catastrophic forgetting while enabling adaptation to new tasks. Current Low-Rank Adaptation (LoRA) continual learning techniques often retain and freeze previously learned LoRAs or generate data representations to overcome forgetting, typically utilizing these to support new LoRAs learn new tasks. However, these methods not only ignore growing computational memory with tasks and limited storage space but also suffer from potential task interference due to the lack of effective LoRA merging mechanisms. In this paper, we propose a novel continual learning method that orthogonally initializes and sequentially merges LoRAs updates into a single unified LoRA. Our method leverages orthogonal basis extraction from previously learned LoRA to initialize the learning of new tasks, further exploits the intrinsic asymmetry property of LoRA components by using a time-aware scaling mechanism to balance new and old knowledge during continual merging. Our approach maintains constant memory complexity with respect to the number of tasks, minimizes interference between past and new tasks via orthogonal basis initialization, and improves performance over asymmetric LoRA merging via adaptive scaling. We provide theoretical analysis to justify our design and conduct extensive experiments across diverse continual learning benchmarks using various Llama models, demonstrating the effectiveness and efficiency of our method.
Read the full abstract

Machine Learning · Cross-listed from stat.ML

JADAI: Jointly Amortizing Adaptive Design and Bayesian Inference

We consider problems of parameter estimation where design variables can be actively optimized to maximize information gain. To this end, we introduce JADAI, a framework that jointly amortizes Bayesian adaptive design and inference by training a policy, a history network, and an inference network end-to-end. The networks minimize a generic loss that aggregates incremental reductions in posterior error along experimental sequences. Inference networks are instantiated with diffusion-based posterior estimators that can approximate high-dimensional and multimodal posteriors at every experimental step. Across standard adaptive design benchmarks, JADAI achieves superior or competitive performance.
Read the full abstract

All authors Niels Bracher, Lars Kühmichel, Desi R. Ivanova, Xavier Intes, Paul-Christian Bürkner, Stefan T. Radev

Machine Learning

Fusion or Confusion? Multimodal Complexity Is Not All You Need

Multimodal learning has become a prominent research area, with the potential of substantial performance gains by combining information across modalities. At the same time, model development has trended toward increasingly complex deep learning architectures, motivated by the assumption that multimodal-specific methods improve performance. We challenge this assumption through a large-scale empirical study by reimplementing 19 high-impact multimodal methods across nine diverse datasets with up to 23 modalities. Under standardized experimental conditions, including hyperparameter tuning, weight initialization, cross-validation, and statistical testing, increased multimodal complexity often yields confusion rather than effective fusion of data modalities. Accordingly, complex multimodal architectures do not reliably outperform unimodal baselines and a Simple Baseline for Multimodal Learning (SimBaMM). Through a focused case study, we further demonstrate concrete methodological shortcomings even in top-tier multimodal learning publications, underscoring the need for standardized evaluation practices. In summary, we argue for a shift in focus for multimodal learning: away from the pursuit of architectural novelty and toward methodological rigor.
Read the full abstract

Machine Learning · Cross-listed from math.OC

Risk-Averse Learning with Varying Risk Levels

In safety-critical decision-making, the environment may evolve over time, and the learner adjusts its risk level accordingly. This work investigates risk-averse online optimization in dynamic environments with varying risk levels, employing Conditional Value-at-Risk (CVaR) as the risk measure. To capture the dynamics of the environment and risk levels, we employ the function variation metric and introduce a novel risk-level variation metric. Two information settings are considered: a first-order scenario, where the learner observes both function values and their gradients; and a zeroth-order scenario, where only function evaluations are available. For both cases, we develop risk-averse learning algorithms with a limited sampling budget and analyze their dynamic regret bounds in terms of function variation, risk-level variation, and the total number of samples. The regret analysis demonstrates the adaptability of the algorithms in non-stationary and risk-sensitive settings. Finally, numerical experiments are presented to demonstrate the efficacy of the methods.
Read the full abstract

Machine Learning

A Context-Aware Temporal Modeling through Unified Multi-Scale Temporal Encoding and Hierarchical Sequence Learning for Single-Channel EEG Sleep Staging

Automatic sleep staging is a critical task in healthcare due to the global prevalence of sleep disorders. This study focuses on single-channel electroencephalography (EEG), a practical and widely available signal for automatic sleep staging. Existing approaches face challenges such as class imbalance, limited receptive-field modeling, and insufficient interpretability. This work proposes a context-aware and interpretable framework for single-channel EEG sleep staging, with particular emphasis on improving detection of the N1 stage. Many prior models operate as black boxes with stacked layers, lacking clearly defined and interpretable feature extraction roles.The proposed model combines compact multi-scale feature extraction with temporal modeling to capture both local and long-range dependencies. To address data imbalance, especially in the N1 stage, classweighted loss functions and data augmentation are applied. EEG signals are segmented into sub-epoch chunks, and final predictions are obtained by averaging softmax probabilities across chunks, enhancing contextual representation and robustness.The proposed framework achieves an overall accuracy of 89.72% and a macro-average F1-score of 85.46%. Notably, it attains an F1- score of 61.7% for the challenging N1 stage, demonstrating a substantial improvement over previous methods on the SleepEDF datasets. These results indicate that the proposed approach effectively improves sleep staging performance while maintaining interpretability and suitability for real-world clinical applications.
Read the full abstract

Machine Learning · Cross-listed from math.OC

Deep Learning for the Multiple Optimal Stopping Problem

This paper presents a novel deep learning framework for solving multiple optimal stopping problems in high dimensions. While deep learning has recently shown promise for single stopping problems, the multiple exercise case involves complex recursive dependencies that remain challenging. We address this by combining the Dynamic Programming Principle with neural network approximation of the value function. Unlike policy-search methods, our algorithm explicitly learns the value surface. We first consider the discrete-time problem and analyze neural network training error. We then turn to continuous problems and analyze the additional error due to the discretization of the underlying stochastic processes. Numerical experiments on high-dimensional American basket options and nonlinear utility maximization demonstrate that our method provides an efficient and scalable method for the multiple optimal stopping problem.
Read the full abstract

Machine Learning

FLOW: A Feedback-Driven Synthetic Longitudinal Dataset of Work and Wellbeing

Access to longitudinal, individual-level data on work-life balance and wellbeing is limited by privacy, ethical, and logistical constraints. This poses challenges for reproducible research, methodological benchmarking, and education in domains such as stress modeling, behavioral analysis, and machine learning. We introduce FLOW, a synthetic longitudinal dataset designed to model daily interactions between workload, lifestyle behaviors, and wellbeing. FLOW is generated using a rule-based, feedback-driven simulation that produces coherent temporal dynamics across variables such as stress, sleep, mood, physical activity, and body weight. The dataset simulates 1{,}000 individuals over a two-year period with daily resolution and is released as a publicly available resource. In addition to the static dataset, we describe a configurable data generation tool that enables reproducible experimentation under adjustable behavioral and contextual assumptions. FLOW is intended as a controlled experimental environment rather than a proxy for observed human populations, supporting exploratory analysis, methodological development, and benchmarking where real-world data are inaccessible.
Read the full abstract

Machine Learning

APO: Alpha-Divergence Preference Optimization

Two divergence regimes dominate modern alignment practice. Supervised fine-tuning and many distillation-style objectives implicitly minimize the forward KL divergence KL(q || pi_theta), yielding stable mode-covering updates but often under-exploiting high-reward modes. In contrast, PPO-style online reinforcement learning from human feedback behaves closer to reverse KL divergence KL(pi_theta || q), enabling mode-seeking improvements but risking mode collapse. Recent anchored methods, such as ADPO, show that performing the projection in anchored coordinates can substantially improve stability, yet they typically commit to a single divergence. We introduce Alpha-Divergence Preference Optimization (APO), an anchored framework that uses Csiszar alpha-divergence to continuously interpolate between forward and reverse KL behavior within the same anchored geometry. We derive unified gradient dynamics parameterized by alpha, analyze gradient variance properties, and propose a practical reward-and-confidence-guarded alpha schedule that transitions from coverage to exploitation only when the policy is both improving and confidently calibrated. Experiments on Qwen3-1.7B with math-level3 demonstrate that APO achieves competitive performance with GRPO and GSPO baselines while maintaining training stability.
Read the full abstract

Machine Learning

Multiple Token Divergence: Measuring and Steering In-Context Computation Density

Measuring the in-context computational effort of language models is a key challenge, as metrics like next-token loss fail to capture reasoning complexity. Prior methods based on latent state compressibility can be invasive and unstable. We propose Multiple Token Divergence (MTD), a simple measure of computational effort defined as the KL divergence between a model's full output distribution and that of a shallow, auxiliary prediction head. MTD can be computed directly from pre-trained models with multiple prediction heads, requiring no additional training. Building on this, we introduce Divergence Steering, a novel decoding method to control the computational character of generated text. We empirically show that MTD is more effective than prior methods at distinguishing complex tasks from simple ones. On mathematical reasoning benchmarks, MTD correlates positively with problem difficulty. Lower MTD is associated with more accurate reasoning. MTD provides a practical, lightweight tool for analyzing and steering the computational dynamics of language models.
Read the full abstract

All authors Vincent Herrmann, Eric Alcaide, Michael Wand, Jürgen Schmidhuber

Machine Learning · Cross-listed from cs.AI

Geometric Structural Knowledge Graph Foundation Model

Structural knowledge graph foundation models aim to generalize reasoning to completely new graphs with unseen entities and relations. A key limitation of existing approaches like Ultra is their reliance on a single relational transformation (e.g., element-wise multiplication) in message passing, which can constrain expressiveness and fail to capture diverse relational and structural patterns exhibited on diverse graphs. In this paper, we propose Gamma, a novel foundation model that introduces multi-head geometric attention to knowledge graph reasoning. Gamma replaces the single relational transformation with multiple parallel ones, including real, complex, split-complex, and dual number based transformations, each designed to model different relational structures. A relational conditioned attention fusion mechanism then adaptively fuses them at link level via a lightweight gating with entropy regularization, allowing the model to robustly emphasize the most appropriate relational bias for each triple pattern. We present a full formalization of these algebraic message functions and discuss how their combination increases expressiveness beyond any single space. Comprehensive experiments on 56 diverse knowledge graphs demonstrate that Gamma consistently outperforms Ultra in zero-shot inductive link prediction, with a 5.5% improvement in mean reciprocal rank on the inductive benchmarks and a 4.4% improvement across all benchmarks, highlighting benefits from complementary geometric representations.
Read the full abstract

All authors Ling Xin, Mojtaba Nayyeri, Zahra Makki Nayeri, Steffen Staab

Machine Learning

Sat-EnQ: Satisficing Ensembles of Weak Q-Learners for Reliable and Compute-Efficient Reinforcement Learning

Deep Q-learning algorithms remain notoriously unstable, especially during early training when the maximization operator amplifies estimation errors. Inspired by bounded rationality theory and developmental learning, we introduce Sat-EnQ, a two-phase framework that first learns to be “good enough” before optimizing aggressively. In Phase 1, we train an ensemble of lightweight Q-networks under a satisficing objective that limits early value growth using a dynamic baseline, producing diverse, low-variance estimates while avoiding catastrophic overestimation. In Phase 2, the ensemble is distilled into a larger network and fine-tuned with standard Double DQN. We prove theoretically that satisficing induces bounded updates and cannot increase target variance, with a corollary quantifying conditions for substantial reduction. Empirically, Sat-EnQ achieves 3.8x variance reduction, eliminates catastrophic failures (0% vs 50% for DQN), maintains 79% performance under environmental noise}, and requires 2.5x less compute than bootstrapped ensembles. Our results highlight a principled path toward robust reinforcement learning by embracing satisficing before optimization.
Read the full abstract

Machine Learning · Cross-listed from math.OC

A first-order method for nonconvex-strongly-concave constrained minimax optimization

In this paper we study a nonconvex-strongly-concave constrained minimax problem. Specifically, we propose a first-order augmented Lagrangian method for solving it, whose subproblems are nonconvex-strongly-concave unconstrained minimax problems and suitably solved by a first-order method developed in this paper that leverages the strong concavity structure. Under suitable assumptions, the proposed method achieves an operation complexity of $O(\varepsilon^{-3.5}\log\varepsilon^{-1})$, measured in terms of its fundamental operations, for finding an $\varepsilon$-KKT solution of the constrained minimax problem, which improves the previous best-known operation complexity by a factor of $\varepsilon^{-0.5}$.
Read the full abstract

Machine Learning

MetaCD: A Meta Learning Framework for Cognitive Diagnosis based on Continual Learning

Cognitive diagnosis is an essential research topic in intelligent education, aimed at assessing the level of mastery of different skills by students. So far, many research works have used deep learning models to explore the complex interactions between students, questions, and skills. However, the performance of existing method is frequently limited by the long-tailed distribution and dynamic changes in the data. To address these challenges, we propose a meta-learning framework for cognitive diagnosis based on continual learning (MetaCD). This framework can alleviate the long-tailed problem by utilizing meta-learning to learn the optimal initialization state, enabling the model to achieve good accuracy on new tasks with only a small amount of data. In addition, we utilize a continual learning method named parameter protection mechanism to give MetaCD the ability to adapt to new skills or new tasks, in order to adapt to dynamic changes in data. MetaCD can not only improve the plasticity of our model on a single task, but also ensure the stability and generalization of the model on sequential tasks. Comprehensive experiments on five real-world datasets show that MetaCD outperforms other baselines in both accuracy and generalization.
Read the full abstract

Machine Learning

Debugging Tabular Log as Dynamic Graphs

Tabular log abstracts objects and events in the real-world system and reports their updates to reflect the change of the system, where one can detect real-world inconsistencies efficiently by debugging corresponding log entries. However, recent advances in processing text-enriched tabular log data overly depend on large language models (LLMs) and other heavy-load models, thus suffering from limited flexibility and scalability. This paper proposes a new framework, GraphLogDebugger, to debug tabular log based on dynamic graphs. By constructing heterogeneous nodes for objects and events and connecting node-wise edges, the framework recovers the system behind the tabular log as an evolving dynamic graph. With the help of our dynamic graph modeling, a simple dynamic Graph Neural Network (GNN) is representative enough to outperform LLMs in debugging tabular log, which is validated by experimental results on real-world log datasets of computer systems and academic papers.
Read the full abstract

All authors Chumeng Liang, Zhanyang Jin, Zahaib Akhtar, Mona Pereira, Haofei Yu, Jiaxuan You

Machine Learning · Cross-listed from eess.SY

A Neural Network-Based Real-time Casing Collar Recognition System for Downhole Instruments

Casing collar locator (CCL) measurements are widely used as reliable depth markers for positioning downhole instruments in cased-hole operations, enabling accurate depth control for operations such as perforation. However, autonomous collar recognition in downhole environments remains challenging because CCL signals are often corrupted by toolstring- or casing-induced magnetic interference, while stringent size and power budgets limit the use of computationally intensive algorithms and specific operations require real-time, in-situ processing. To address these constraints, we propose Collar Recognition Nets (CRNs), a family of domain-specific lightweight 1-D convolutional neural networks for collar signature recognition from streaming CCL waveforms. With depthwise separable convolutions and input pooling, CRNs optimize efficiency without sacrificing accuracy. Our most compact model achieves an F1-score of 0.972 on field data with only 1,985~parameters and 8,208~MACs, and deployed on an ARM Cortex-M7 based embedded system using TensorFlow Lite for Microcontrollers (TFLM) library, the model demonstrates a throughput of 1,000 inference per second and 343.2 {\mu}s latency, confirming the feasibility of robust, autonomous, and real-time collar recognition under stringent downhole constraints.
Read the full abstract

All authors Si-Yu Xiao, Xin-Di Zhao, Xiang-Zhan Wang, Tian-Hao Mao, Ying-Kai Liao, Xing-Yu Liao, Yu-Qiao Chen, Jun-Jie Wang, Shuang Liu, Tu-Pei Chen, Yang Liu

Machine Learning

Federated Multi-Task Clustering

Spectral clustering has emerged as one of the most effective clustering algorithms due to its superior performance. However, most existing models are designed for centralized settings, rendering them inapplicable in modern decentralized environments. Moreover, current federated learning approaches often suffer from poor generalization performance due to reliance on unreliable pseudo-labels, and fail to capture the latent correlations amongst heterogeneous clients. To tackle these limitations, this paper proposes a novel framework named Federated Multi-Task Clustering (i.e.,FMTC), which intends to learn personalized clustering models for heterogeneous clients while collaboratively leveraging their shared underlying structure in a privacy-preserving manner. More specifically, the FMTC framework is composed of two main components: client-side personalized clustering module, which learns a parameterized mapping model to support robust out-of-sample inference, bypassing the need for unreliable pseudo-labels; and server-side tensorial correlation module, which explicitly captures the shared knowledge across all clients. This is achieved by organizing all client models into a unified tensor and applying a low-rank regularization to discover their common subspace. To solve this joint optimization problem, we derive an efficient, privacy-preserving distributed algorithm based on the Alternating Direction Method of Multipliers, which decomposes the global problem into parallel local updates on clients and an aggregation step on the server. To the end, several extensive experiments on multiple real-world datasets demonstrate that our proposed FMTC framework significantly outperforms various baseline and state-of-the-art federated clustering algorithms.
Read the full abstract

All authors Suyan Dai, Gan Sun, Fazeng Li, Xu Tang, Qianqian Wang, Yang Cong

Machine Learning

Theory and Algorithms for Learning with Multi-Class Abstention and Multi-Expert Deferral

Large language models (LLMs) have achieved remarkable performance but face critical challenges: hallucinations and high inference costs. Leveraging multiple experts offers a solution: deferring uncertain inputs to more capable experts improves reliability, while routing simpler queries to smaller, distilled models enhances efficiency. This motivates the problem of learning with multiple-expert deferral. This thesis presents a comprehensive study of this problem and the related problem of learning with abstention, supported by strong consistency guarantees. First, for learning with abstention (a special case of deferral), we analyze score-based and predictor-rejector formulations in multi-class classification. We introduce new families of surrogate losses and prove strong non-asymptotic, hypothesis set-specific consistency guarantees, resolving two existing open questions. We analyze both single-stage and practical two-stage settings, with experiments on CIFAR-10, CIFAR-100, and SVHN demonstrating the superior performance of our algorithms. Second, we address general multi-expert deferral in classification. We design new surrogate losses for both single-stage and two-stage scenarios and prove they benefit from strong $H$-consistency bounds. For the two-stage scenario, we show that our surrogate losses are realizable $H$-consistent for constant cost functions, leading to effective new algorithms. Finally, we introduce a novel framework for regression with deferral to address continuous label spaces. Our versatile framework accommodates multiple experts and various cost structures, supporting both single-stage and two-stage methods. It subsumes recent work on regression with abstention. We propose new surrogate losses with proven $H$-consistency and demonstrate the empirical effectiveness of the resulting algorithms.
Read the full abstract

Machine Learning

Fundamental Novel Consistency Theory: $H$-Consistency Bounds

In machine learning, the loss functions optimized during training often differ from the target loss that defines task performance due to computational intractability or lack of differentiability. We present an in-depth study of the target loss estimation error relative to the surrogate loss estimation error. Our analysis leads to $H$-consistency bounds, which are guarantees accounting for the hypothesis set $H$. These bounds offer stronger guarantees than Bayes-consistency or $H$-calibration and are more informative than excess error bounds. We begin with binary classification, establishing tight distribution-dependent and -independent bounds. We provide explicit bounds for convex surrogates (including linear models and neural networks) and analyze the adversarial setting for surrogates like $\rho$-margin and sigmoid loss. Extending to multi-class classification, we present the first $H$-consistency bounds for max, sum, and constrained losses, covering both non-adversarial and adversarial scenarios. We demonstrate that in some cases, non-trivial $H$-consistency bounds are unattainable. We also investigate comp-sum losses (e.g., cross-entropy, MAE), deriving their first $H$-consistency bounds and introducing smooth adversarial variants that yield robust learning algorithms. We develop a comprehensive framework for deriving these bounds across various surrogates, introducing new characterizations for constrained and comp-sum losses. Finally, we examine the growth rates of $H$-consistency bounds, establishing a universal square-root growth rate for smooth surrogates in binary and multi-class tasks, and analyze minimizability gaps to guide surrogate selection.
Read the full abstract

Machine Learning · Cross-listed from cs.MA

Reinforcement Networks: novel framework for collaborative Multi-Agent Reinforcement Learning tasks

Modern AI systems often comprise multiple learnable components that can be naturally organized as graphs. A central challenge is the end-to-end training of such systems without restrictive architectural or training assumptions. Such tasks fit the theory and approaches of the collaborative Multi-Agent Reinforcement Learning (MARL) field. We introduce Reinforcement Networks, a general framework for MARL that organizes agents as vertices in a directed acyclic graph (DAG). This structure extends hierarchical RL to arbitrary DAGs, enabling flexible credit assignment and scalable coordination while avoiding strict topologies, fully centralized training, and other limitations of current approaches. We formalize training and inference methods for the Reinforcement Networks framework and connect it to the LevelEnv concept to support reproducible construction, training, and evaluation. We demonstrate the effectiveness of our approach on several collaborative MARL setups by developing several Reinforcement Networks models that achieve improved performance over standard MARL baselines. Beyond empirical gains, Reinforcement Networks unify hierarchical, modular, and graph-structured views of MARL, opening a principled path toward designing and training complex multi-agent systems. We conclude with theoretical and practical directions - richer graph morphologies, compositional curricula, and graph-aware exploration. That positions Reinforcement Networks as a foundation for a new line of research in scalable, structured MARL.
Read the full abstract

All authors Maksim Kryzhanovskiy, Svetlana Glazyrina, Roman Ischenko, Konstantin Vorontsov

Machine Learning · Cross-listed from cs.CR

Adaptive Trust Consensus for Blockchain IoT: Comparing RL, DRL, and MARL Against Naive, Collusive, Adaptive, Byzantine, and Sleeper Attacks

Securing blockchain-enabled IoT networks against sophisticated adversarial attacks remains a critical challenge. This paper presents a trust-based delegated consensus framework integrating Fully Homomorphic Encryption (FHE) with Attribute-Based Access Control (ABAC) for privacy-preserving policy evaluation, combined with learning-based defense mechanisms. We systematically compare three reinforcement learning approaches – tabular Q-learning (RL), Deep RL with Dueling Double DQN (DRL), and Multi-Agent RL (MARL) – against five distinct attack families: Naive Malicious Attack (NMA), Collusive Rumor Attack (CRA), Adaptive Adversarial Attack (AAA), Byzantine Fault Injection (BFI), and Time-Delayed Poisoning (TDP). Experimental results on a 16-node simulated IoT network reveal significant performance variations: MARL achieves superior detection under collusive attacks (F1=0.85 vs. DRL's 0.68 and RL's 0.50), while DRL and MARL both attain perfect detection (F1=1.00) against adaptive attacks where RL fails (F1=0.50). All agents successfully defend against Byzantine attacks (F1=1.00). Most critically, the Time-Delayed Poisoning attack proves catastrophic for all agents, with F1 scores dropping to 0.11-0.16 after sleeper activation, demonstrating the severe threat posed by trust-building adversaries. Our findings indicate that coordinated multi-agent learning provides measurable advantages for defending against sophisticated trust manipulation attacks in blockchain IoT environments.
Read the full abstract

Machine Learning · Cross-listed from cs.CV

ByteLoom: Weaving Geometry-Consistent Human-Object Interactions through Progressive Curriculum Learning

Human-object interaction (HOI) video generation has garnered increasing attention due to its promising applications in digital humans, e-commerce, advertising, and robotics imitation learning. However, existing methods face two critical limitations: (1) a lack of effective mechanisms to inject multi-view information of the object into the model, leading to poor cross-view consistency, and (2) heavy reliance on fine-grained hand mesh annotations for modeling interaction occlusions. To address these challenges, we introduce ByteLoom, a Diffusion Transformer (DiT)-based framework that generates realistic HOI videos with geometrically consistent object illustration, using simplified human conditioning and 3D object inputs. We first propose an RCM-cache mechanism that leverages Relative Coordinate Maps (RCM) as a universal representation to maintain object's geometry consistency and precisely control 6-DoF object transformations in the meantime. To compensate HOI dataset scarcity and leverage existing datasets, we further design a training curriculum that enhances model capabilities in a progressive style and relaxes the demand of hand mesh. Extensive experiments demonstrate that our method faithfully preserves human identity and the object's multi-view geometry, while maintaining smooth motion and object manipulation.
Read the full abstract

All authors Bangya Liu, Xinyu Gong, Zelin Zhao, Ziyang Song, Yulei Lu, Suhui Wu, Jun Zhang, Suman Banerjee, Hao Zhang

Machine Learning · Cross-listed from econ.EM

Causal-Policy Forest for End-to-End Policy Learning

This study proposes an end-to-end algorithm for policy learning in causal inference. We observe data consisting of covariates, treatment assignments, and outcomes, where only the outcome corresponding to the assigned treatment is observed. The goal of policy learning is to train a policy from the observed data, where a policy is a function that recommends an optimal treatment for each individual, to maximize the policy value. In this study, we first show that maximizing the policy value is equivalent to minimizing the mean squared error for the conditional average treatment effect (CATE) under $\{-1, 1\}$ restricted regression models. Based on this finding, we modify the causal forest, an end-to-end CATE estimation algorithm, for policy learning. We refer to our algorithm as the causal-policy forest. Our algorithm has three advantages. First, it is a simple modification of an existing, widely used CATE estimation method, therefore, it helps bridge the gap between policy learning and CATE estimation in practice. Second, while existing studies typically estimate nuisance parameters for policy learning as a separate task, our algorithm trains the policy in a more end-to-end manner. Third, as in standard decision trees and random forests, we train the models efficiently, avoiding computational intractability.
Read the full abstract

Machine Learning

TEACH: Temporal Variance-Driven Curriculum for Reinforcement Learning

Reinforcement Learning (RL) has achieved significant success in solving single-goal tasks. However, uniform goal selection often results in sample inefficiency in multi-goal settings where agents must learn a universal goal-conditioned policy. Inspired by the adaptive and structured learning processes observed in biological systems, we propose a novel Student-Teacher learning paradigm with a Temporal Variance-Driven Curriculum to accelerate Goal-Conditioned RL. In this framework, the teacher module dynamically prioritizes goals with the highest temporal variance in the policy's confidence score, parameterized by the state-action value (Q) function. The teacher provides an adaptive and focused learning signal by targeting these high-uncertainty goals, fostering continual and efficient progress. We establish a theoretical connection between the temporal variance of Q-values and the evolution of the policy, providing insights into the method's underlying principles. Our approach is algorithm-agnostic and integrates seamlessly with existing RL frameworks. We demonstrate this through evaluation across 11 diverse robotic manipulation and maze navigation tasks. The results show consistent and notable improvements over state-of-the-art curriculum learning and goal-selection methods.
Read the full abstract

Machine Learning

Long-Range Distillation: Distilling 10,000 Years of Simulated Climate into Long Timestep AI Weather Models

Accurate long-range weather forecasting remains a major challenge for AI models, both because errors accumulate over autoregressive rollouts and because reanalysis training datasets provide limited sampling of the slow modes of variability underpinning predictability. Most AI and all numerical weather prediction models are autoregressive, repeatedly applying short-lead forecasts to reach subseasonal-to-seasonal (S2S) or seasonal lead times, often causing instability and miscalibration. Long-timestep probabilistic models that generate long-range forecasts in a single step offer an alternative, but training on the 40-year reanalysis record leads to overfitting, suggesting orders of magnitude more data are required. We introduce long-range distillation, whereby a long-timestep probabilistic “student” model is trained to forecast directly at long-range using a huge synthetic dataset generated by an autoregressive “teacher” model. Using the Deep Learning Earth System Model (DLESyM) as teacher, we generate over 10,000 years of simulated climate to train student models across a range of timescales. In perfect-model experiments, the distilled models outperform climatology and approach their teacher's skill while replacing hundreds of autoregressive steps with a single timestep. In the real world, both a diffusion model student fine-tuned on ERA5 and a linear inverse model student with mixed DLESyM-ERA5 training achieve S2S forecast skill competitive with the ECMWF IFS ensemble, with synthetic pre-training providing improvements beyond ERA5-only training. This demonstrates the benefit of synthetic pre-training for both deep learning and classical data-driven methods, and represents the first demonstration that real-world skill can be improved by training on AI-generated synthetic data in volumes far exceeding the observational record.
Read the full abstract

All authors Scott A. Martin, Noah Brenowitz, Dale Durran, Michael Pritchard

Machine Learning

MoR: Mixture Of Representations For Mixed-Precision Training

Mixed-precision training is a crucial technique for scaling deep learning models, but successful mixedprecision training requires identifying and applying the right combination of training methods. This paper presents our preliminary study on Mixture-of-Representations (MoR), a novel, per-tensor and sub-tensor level quantization framework that dynamically analyzes a tensor's numerical properties to select between a variety of different representations. Based on the framework, we have proposed and experimented concrete algorithms that choose dynamically between FP8 and BF16 representations for both per-tensor and sub-tensor level granularities. Our universal approach is designed to preserve model quality across various quantization partition strategies and datasets. Our initial findings show that this approach can achieve state-of-the-art results with 98.38% of tensors quantized to the FP8 format. This work highlights the potential of dynamic, property-aware quantization while preserving model quality. We believe this approach can generally improve the robustness of low precision training, as demonstrated by achieving FP8 accuracies that are on par with existing approaches without the need for fine-grain partitioning, or can be used in combination with other training methods to improve the leverage of even lower precision number formats such as NVFP4.
Read the full abstract

All authors Bor-Yiing Su, Peter Dykas, Mike Chrzanowski, Jatin Chhugani