论文检索

Reinforcement Learning · Everything Else

Chris Cundy, Adam Gleave

As AI systems become more capable, deceptive behaviors can undermine evaluation and mislead users at deployment. Recent work has shown that lie detectors can accurately classify deceptive behavior, but they are not typically used in the training pipeline due to concerns around contamination and objective hacking. We examine these concerns by incorporating a lie detector into the labelling step of LLM post-training and evaluating whether the learned policy is genuinely more honest, or instead learns to fool the lie detector while remaining deceptive. Using DolusChat, a novel 65k-example dataset with paired truthful/deceptive responses, we identify three key factors that determine the honesty of learned policies: amount of exploration during preference learning, lie detector accuracy, and KL regularization strength. We find that preference learning with lie detectors and GRPO can lead to policies which evade lie detectors, with deception rates of over 85\%. However, if the lie detector true positive rate (TPR) or KL regularization is sufficiently high, GRPO learns honest policies. In contrast, off-policy algorithms (DPO) consistently lead to deception rates under 25\% for realistic TPRs. Our results illustrate a more complex picture than previously assumed: depending on the context, lie-detector-enhanced training can be a powerful tool for scalable oversight, or a counterproductive method encouraging undetectable misalignment.

General Machine Learning · Everything Else

Matthew Tamayo-Rios, Alexander Schell, Rima Alaifari

The signature kernel is a recent state-of-the-art tool for analyzing high-dimensional sequential data, valued for its theoretical guarantees and strong empirical performance. In this paper, we present a novel method for efficiently computing the signature kernel of long, high-dimensional time series via adaptively truncated recursive local power series expansions. Building on the characterization of the signature kernel as the solution of a Goursat PDE, our approach employs tilewise Neumann‐series expansions to derive rapidly converging power series approximations of the signature kernel that are locally defined on subdomains and propagated iteratively across the entire domain of the Goursat solution by exploiting the geometry of the time series. Algorithmically, this involves solving a system of interdependent Goursat PDEs via adaptively truncated local power series expansions and recursive propagation of boundary conditions along a directed graph in a topological ordering. This method strikes an effective balance between computational cost and accuracy, achieving substantial performance improvements over state-of-the-art approaches for computing the signature kernel. It offers (a) adjustable and superior accuracy, even for time series with very high roughness; (b) drastically reduced memory requirements; and (c) scalability to efficiently handle very long time series (e.g., with up to one million data points or more) on a single GPU. As demonstrated in our benchmarks, these advantages make our method particularly well-suited for rough-path-assisted machine learning, financial modeling, and signal processing applications involving very long and highly volatile sequential data.

Deep Learning · Algorithms

Zihao Cheng, Binrui Wu, Zhiwei Li, Yuesen Liao, Su Zhao, Shuai Chen, Yuan Gao, Weizhong Zhang

Dynamic coreset selection is a promising approach for improving the training efficiency of deep neural networks by periodically selecting a small subset of the most representative or informative samples, thereby avoiding the need to train on the entire dataset. However, it remains inherently challenging due not only to the complex interdependencies among samples and the evolving nature of model training, but also to a critical *coreset representativeness degradation issue* identified and explored in-depth in this paper, that is, the representativeness or information content of the coreset degrades over time as training progresses. Therefore, we argue that, in addition to designing accurate selection rules, it is equally important to endow the algorithms with the ability to assess the quality of the current coreset. Such awareness enables timely re-selection, mitigating the risk of overfitting to stale subsets—a limitation often overlooked by existing methods. To this end, this paper proposes an **E**fficient **R**epresentativeness-**A**ware **C**oreset **S**election method for deep neural networks, a lightweight framework that enables dynamic tracking and maintenance of coreset quality during training. While the ideal criterion—gradient discrepancy between the coreset and the full dataset—is computationally prohibitive, we introduce a scalable surrogate based on the signal-to-noise ratio (SNR) of gradients within the coreset, which is the main technical contribution of this paper and is also supported by our theoretical analysis. Intuitively, a decline in SNR indicates overfitting to the subset and declining representativeness. Leveraging this observation, our method triggers coreset updates without requiring costly Hessian or full-batch gradient computations, maintaining minimal computational overhead. Experiments on multiple datasets confirm the effectiveness of our approach. Notably, compared with existing gradient-based dynamic coreset selection baselines, our method achieves up to a 5.4\% improvement in test accuracy across multiple datasets.

Probabilistic Methods · Everything Else

Yinxu Tang, Stylianos Loukas Vasileiou, Vincent Derkinderen, William Yeoh

In human-AI interaction, effective communication relies on aligning the AI agent’s model with the human user’s mental model -- a process known as model reconciliation. However, existing model reconciliation approaches predominantly assume deterministic models, overlooking the fact that human knowledge is often uncertain or probabilistic. To bridge this gap, we present a probabilistic model reconciliation framework that resolves inconsistencies in MPE outcome probabilities between an agent’s and a user’s models. Our approach is built on probabilistic logic programming (PLP) using ProbLog, where explanations are generated as cost-optimal model updates that reconcile these probabilistic differences. We develop two search algorithms - a generic baseline and an optimized version. The latter is guided by theoretical insights and further extended with greedy and weighted variants to enhance scalability and efficiency. Our approach is validated through a user study on explanation types and computational experiments showing that the optimized version consistently outperforms the generic baseline.

Probabilistic Methods · Bayesian Models and Methods

Guohao Sun, Hang Hua, Jian Wang, Jiebo Luo, Sohail Dianat, MAJID RABBANI, Raghuveer Rao, Zhiqiang Tao

Chain-of-thought (CoT) reasoning is critical for improving the interpretability and reliability of Large Vision-Language Models (LVLMs). However, existing training algorithms such as SFT, PPO, and GRPO may not generalize well across unseen reasoning tasks and heavily rely on a biased reward model. To address this challenge, we reformulate reasoning in LVLMs as posterior inference and propose a scalable training algorithm based on amortized variational inference. By leveraging diversity-seeking reinforcement learning algorithms, we introduce a novel sparse reward function for token-level learning signals that encourage diverse, high-likelihood latent CoT, overcoming deterministic sampling limitations and avoiding reward hacking. Additionally, we implement a Bayesian inference-scaling strategy that replaces costly Best-of-N and Beam Search with a marginal likelihood to efficiently rank optimal rationales and answers. We empirically demonstrate that the proposed method enhances the state-of-the-art LVLMs on four reasoning benchmarks, in terms of effectiveness, generalization, and interpretability.

Theory · Domain Adaptation and Transfer Learning

Christoph Dann, Yishay Mansour, Teodor Vanislavov Marinov, Mehryar Mohri

The rapid proliferation of domain-specialized machine learning models presents a challenge: while individual models excel in specific domains, their performance varies significantly across diverse applications. This makes selecting the optimal model when faced with an unknown mixture of tasks, especially with limited or no data to estimate the mixture, a difficult problem. We address this challenge by formulating it as a multiple-source domain adaptation (MSA) problem. We introduce a novel, scalable algorithm that effectively routes each input to the best-suited model from a pool of available models. Our approach provides a strong performance guarantee: remarkably, for any mixture domain, the accuracy achieved by the best source model is maintained. This guarantee is established through a theoretical bound on the regret for new domains, expressed as a convex combination of the best regrets in the source domains, plus a concentration term that diminishes as the amount of source data increases. While our primary contributions are theoretical and algorithmic, we also present empirical results demonstrating the effectiveness of our approach.

jusheng zhang, Kaitong Cai, Yijia Fan, Ningyuan Liu, Keze Wang

We propose a novel collaborative multi-agent optimization framework for adaptive training in multi-label image classification, fundamentally advancing beyond static decision rules and isolated automation. Our method deploys a set of distributed, task-specific agents, each responsible for dynamically orchestrating critical training components—including data augmentation, optimization methods, learning rate schedules, and loss functions—according to evolving visual-semantic relationships and training states. Each agent employs an advanced non-stationary multi-armed bandit algorithm, integrating both $\epsilon$-greedy and upper confidence bound strategies, to judiciously balance exploration with exploitation throughout the training lifecycle. A hierarchical composite reward mechanism synergizes overall classification accuracy, rare class recognition, and training stability, fostering both independent optimization and implicit collaborative behavior among agents. The framework further leverages refined techniques such as dual-rate exponential moving average smoothing and structured mixed-precision training to enhance robustness and computational efficiency. Extensive experiments across benchmarks including Pascal VOC, COCO, Yeast, and Mediamill demonstrate that our approach achieves superior mean average precision and rare-class F1 scores compared to state-of-the-art methods, while also exhibiting rapid convergence and remarkable cross-domain generalization. Our results indicate that collaborative multi-agent adaptive optimization offers a scalable and principled solution for self-optimizing deep learning in complex multi-label scenarios.

Reinforcement Learning · Deep RL

Seohong Park, Kevin Frans, Deepinder Mann, Benjamin Eysenbach, Aviral Kumar, Sergey Levine

In this work, we study the scalability of offline reinforcement learning (RL) algorithms. In principle, a truly scalable offline RL algorithm should be able to solve any given problem, regardless of its complexity, given sufficient data, compute, and model capacity. We investigate if and how current offline RL algorithms match up to this promise on diverse, challenging, previously unsolved tasks, using datasets up to 1000× larger than typical offline RL datasets. We observe that despite scaling up data, many existing offline RL algorithms exhibit poor scaling behavior, saturating well below the maximum performance. We hypothesize that the horizon is the main cause behind the poor scaling of offline RL. We empirically verify this hypothesis through several analysis experiments, showing that long horizons indeed present a fundamental barrier to scaling up offline RL. We then show that various horizon reduction techniques substantially enhance scalability on challenging tasks. Based on our insights, we also introduce a minimal yet scalable method named SHARSA that effectively reduces the horizon. SHARSA achieves the best asymptotic performance and scaling behavior among our evaluation methods, showing that explicitly reducing the horizon unlocks the scalability of offline RL.

General Machine Learning · Causality

Yushu Pan, Elias Bareinboim

The process of editing an image can be naturally modeled as evaluating a counterfactual query: “What would an image look like if a particular feature had changed?” While recent advances in text-guided image editing leverage powerful pre-trained models to produce visually appealing images, they often lack counterfactual consistency -- ignoring how features are causally related and how changing one may affect others. In contrast, existing causal-based editing approaches offer solid theoretical foundations and perform well in specific settings, but remain limited in scalability and often rely on labeled data. In this work, we aim to bridge the gap between causal editing and large-scale text-to-image generation through two main contributions. First, we introduce Backdoor Disentangled Causal Latent Space (BD-CLS), a new class of latent spaces that allows for the encoding of causal inductive biases. One desirable property of this latent space is that, even under weak supervision, it can be shown to exhibit counterfactual consistency. Second, and building on this result, we develop BD-CLS-Edit, an algorithm capable of learning a BD-CLS from a (non-causal) pre-trained Stable Diffusion model. This enables counterfactual image editing without retraining. Our method ensures that edits respect the causal relationships among features, even when some features are unlabeled or unprompted and the original latent space is oblivious to the environment’s underlying cause-and-effect relationships.

Optimization · Neural Network Optimization

Zeju Qiu, Simon Buchholz, Tim Xiao, Maximilian Dax, Bernhard Schölkopf, Weiyang Liu

While large language models (LLMs) are driving the rapid advancement of artificial intelligence, effectively and reliably training these large models remains one of the field's most significant challenges. To address this challenge, we propose POET, a novel reParameterized training algorithm that uses Orthogonal Equivalence Transformation to optimize neurons. Specifically, POET reparameterizes each neuron with two learnable orthogonal matrices and a fixed random weight matrix. Because of its provable preservation of spectral properties of weight matrices, POET can stably optimize the objective function with improved generalization. We further develop efficient approximations that make POET flexible and scalable for training large-scale neural networks. Extensive experiments validate the effectiveness and scalability of POET in training LLMs.

Kevin Wang, Ishaan Javali, Michał Bortkiewicz, Tomasz Trzcinski, Benjamin Eysenbach

Scaling up self-supervised learning has driven breakthroughs in language and vision, yet comparable progress has remained elusive in reinforcement learning (RL). In this paper, we study building blocks for self-supervised RL that unlock substantial improvements in scalability, with network depth serving as a critical factor. Whereas most RL papers in recent years have relied on shallow architectures (around 2 -- 5 layers), we demonstrate that increasing the depth up to 1024 layers can significantly boost performance. Our experiments are conducted in an unsupervised goal-conditioned setting, where no demonstrations or rewards are provided, so an agent must explore (from scratch) and learn how to maximize the likelihood of reaching commanded goals. Evaluated on simulated locomotion and manipulation tasks, our approach increases performance on the self-supervised contrastive RL algorithm by $2\times$ -- $50\times$, outperforming other goal-conditioned baselines. Increasing the model depth not only increases success rates but also qualitatively changes the behaviors learned.

Guillaume Pourcel, Maxence Ernoult

Deep State Space Models (SSMs) reignite physics-grounded compute paradigms, as RNNs could natively be embodied into dynamical systems. This calls for dedicated learning algorithms obeying to core physical principles, with efficient techniques to simulate these systems and guide their design. We propose \emph{Recurrent Hamiltonian Echo Learning} (RHEL), an algorithm which provably computes loss gradients as finite differences of physical trajectories of non-dissipative, \emph{Hamiltonian systems}. In ML terms, RHEL only requires three ``forward passes'' irrespective of model size, without explicit Jacobian computation, nor incurring any variance in the gradient estimation. Motivated by the potential to implement our algorithm in non-digital physical systems, we first introduce RHEL in \emph{continuous time} and demonstrate its formal equivalence with the continuous adjoint state method. To facilitate the simulation of Hamiltonian systems trained by RHEL, we propose a \emph{discrete-time} version of RHEL which is equivalent to Backpropagation Through Time (BPTT) when applied to a class of recurrent modules which we call \emph{Hamiltonian Recurrent Units} (HRUs). This setting allows us to demonstrate the scalability of RHEL by generalizing these results to hierarchies of HRUs, which we call \emph{Hamiltonian SSMs} (HSSMs). We apply RHEL to train HSSMs with linear and nonlinear dynamics on a variety of time-series tasks ranging from mid-range to long-range classification and regression with sequence length reaching $\sim 50k$. We show that RHEL consistently matches the performance of BPTT across all models and tasks. This work opens new doors for the design of scalable, energy-efficient physical systems endowed with self-learning capabilities for sequence modelling.

Deep Learning · Algorithms

Kunjun Li, Zigeng Chen, Cheng-Yen Yang, Jenq-Neng Hwang

Visual Autoregressive (VAR) modeling has garnered significant attention for its innovative next-scale prediction approach, which yields substantial improvements in efficiency, scalability, and zero-shot generalization. Nevertheless, the coarse-to-fine methodology inherent in VAR results in exponential growth of the KV cache during inference, causing considerable memory consumption and computational redundancy. To address these bottlenecks, we introduce ScaleKV, a novel KV cache compression framework tailored for VAR architectures. ScaleKV leverages two critical observations: varying cache demands across transformer layers and distinct attention patterns at different scales. Based on these insights, ScaleKV categorizes transformer layers into two functional groups: drafters and refiners. Drafters exhibit dispersed attention across multiple scales, thereby requiring greater cache capacity. Conversely, refiners focus attention on the current token map to process local details, consequently necessitating substantially reduced cache capacity. ScaleKV optimizes the multi-scale inference pipeline by identifying scale-specific drafters and refiners, facilitating differentiated cache management tailored to each scale. Evaluation on the state-of-the-art text-to-image VAR model family, Infinity, demonstrates that our approach effectively reduces the required KV cache memory to 10% while preserving pixel-level fidelity.

Reinforcement Learning · Deep RL

Kevin Wang, Ishaan Javali, Michał Bortkiewicz, Tomasz Trzcinski, Benjamin Eysenbach

Scaling up self-supervised learning has driven breakthroughs in language and vision, yet comparable progress has remained elusive in reinforcement learning (RL). In this paper, we study building blocks for self-supervised RL that unlock substantial improvements in scalability, with network depth serving as a critical factor. Whereas most RL papers in recent years have relied on shallow architectures (around 2 -- 5 layers), we demonstrate that increasing the depth up to 1024 layers can significantly boost performance. Our experiments are conducted in an unsupervised goal-conditioned setting, where no demonstrations or rewards are provided, so an agent must explore (from scratch) and learn how to maximize the likelihood of reaching commanded goals. Evaluated on simulated locomotion and manipulation tasks, our approach increases performance on the self-supervised contrastive RL algorithm by $2\times$ -- $50\times$, outperforming other goal-conditioned baselines. Increasing the model depth not only increases success rates but also qualitatively changes the behaviors learned.

Jiatao Gu, Tianrong Chen, David Berthelot, Huangjie Zheng, Yuyang Wang, Ruixiang Zhang, Laurent Dinh, Miguel Angel Bautista, Joshua Susskind, Shuangfei Zhai

We present STARFlow, a scalable generative model based on normalizing flows that achieves strong performance on high-resolution image synthesis. STARFlow's main building block is Transformer Autoregressive Flow (TARFlow), which combines normalizing flows with Autoregressive Transformer architectures and has recently achieved impressive results in image modeling. In this work, we first establish the theoretical universality of TARFlow for modeling continuous distributions. Building on this foundation, we introduce a set of architectural and algorithmic innovations that significantly enhance the scalability: (1) a deep-shallow design where a deep Transformer block captures most of the model’s capacity, followed by a few shallow Transformer blocks that are computationally cheap yet contribute non-negligibly, (2) learning in the latent space of pretrained autoencoders, which proves far more effective than modeling pixels directly, and (3) a novel guidance algorithm that substantially improves sample quality. Crucially, our model remains a single, end-to-end normalizing flow, allowing exact maximum likelihood training in continuous space without discretization. STARFlow achieves competitive results in both class- and text-conditional image generation, with sample quality approaching that of state-of-the-art diffusion models. To our knowledge, this is the **first** successful demonstration of normalizing flows at this scale and resolution. Code and weights available at https://github.com/apple/ml-starflow.

Xueliang Zhao, Wei Wu, Jian Guan, Qintong Li, Lingpeng Kong

In modern sequential decision-making systems, the construction of an optimal candidate action space is critical to efficient inference. However, existing approaches either rely on manually defined action spaces that lack scalability or utilize unstructured spaces that render exhaustive search computationally prohibitive. In this paper, we propose a novel framework named \textsc{DynaAct} for automatically constructing a compact action space to enhance sequential reasoning in complex problem-solving scenarios. Our method first estimates a proxy for the complete action space by extracting general sketches observed in a corpus covering diverse complex reasoning problems using large language models. We then formulate a submodular function that jointly evaluates candidate actions based on their utility to the current state and their diversity, and employ a greedy algorithm to select an optimal candidate set. Extensive experiments on six diverse standard benchmarks demonstrate that our approach significantly improves overall performance, while maintaining efficient inference without introducing substantial latency. The implementation is available at \url{https://github.com/zhaoxlpku/DynaAct}.

General Machine Learning · Causality

Muralikrishnna Guruswamy Sethuraman, Faramarz Fekri

Understanding causal relationships between variables is fundamental across scientific disciplines. Most causal discovery algorithms rely on two key assumptions: (i) all variables are observed, and (ii) the underlying causal graph is acyclic. While these assumptions simplify theoretical analysis, they are often violated in real-world systems, such as biological networks. Existing methods that account for confounders either assume linearity or struggle with scalability. To address these limitations, we propose DCCD-CONF, a novel framework for differentiable learning of nonlinear cyclic causal graphs in the presence of unmeasured confounders using interventional data. Our approach alternates between optimizing the graph structure and estimating the confounder distribution by maximizing the log-likelihood of the data. Through experiments on synthetic data and real-world gene perturbation datasets, we show that DCCD-CONF outperforms state-of-the-art methods in both causal graph recovery and confounder identification. Additionally, we provide consistency guarantees for our framework, reinforcing its theoretical soundness.

Data-centric AI · Data-centric AI methods and tools

Ishika Agarwal, Dilek Hakkani-Tur

Influence functions provide crucial insights into model training, but existing methods suffer from large computational costs and limited generalization. Particularly, recent works have proposed various metrics and algorithms to calculate the influence of data using language models, which do not scale well with large models and datasets. This is because of the expensive forward and backward passes required for computation, substantial memory requirements to store large models, and poor generalization of influence estimates to new data. In this paper, we explore the use of small neural networks -- which we refer to as the InfluenceNetwork -- to estimate influence values, achieving up to 99% cost reduction. Our evaluation demonstrates that influence values can be estimated with models just 0.0007% the size of full language models (we average across 1.5B-22B versions). We apply our algorithm of estimating influence values (called NN-CIFT: Neural Networks for effiCient Instruction Fine-Tuning) to the downstream task of subset selection for general instruction fine-tuning. In our study, we include four state-of-the-art influence functions and show no compromise in performance, despite large speedups, between NN-CIFT and the original influence functions. We provide an in-depth hyperparameter analyses of NN-CIFT. The code for our method can be found here: https://github.com/agarwalishika/NN-CIFT/tree/main.

Reinforcement Learning · Deep RL

Jingcheng Hu, Yinmin Zhang, Qi Han, Daxin Jiang, Xiangyu Zhang, Heung-Yeung Shum

We introduce Open-Reasoner-Zero, the first open source implementation of large-scale reasoning-oriented RL training on the base model focusing on scalability, simplicity and accessibility. Through extensive experiments, we demonstrate that a minimalist approach, vanilla PPO with GAE ($\lambda=1$, $\gamma=1$) and straightforward rule-based rewards, without any KL regularization, is sufficient to scale up both benchmark performance and response length, replicating the scaling phenomenon observed in DeepSeek-R1-Zero. Using the same base model as DeepSeek-R1-Zero-Qwen-32B, our implementation achieves superior performance across AIME2024, MATH500, and GPQA Diamond, while demonstrating remarkable efficiency—requiring only 1/10 of the training steps compared to the DeepSeek-R1-Zero pipeline. We validate that this recipe generalizes well across diverse training domains and different model families without algorithmic modifications. Moreover, our analysis not only covers training dynamics and ablation for critical design choices, but also quantitatively show how the learned critic in Reasoner-Zero training effectively identifies and devalues repetitive response patterns, yielding more robust advantage estimations and enhancing training stability. Embracing the principles of open-source, we release our source code, parameter settings, training data, and model weights across various sizes, fostering reproducibility and encouraging further exploration of the properties of related models.

Viet-Hoang Tran, Van Hoan Trinh, Khanh-Vinh Bui, Tan Nguyen

Linear Mode Connectivity (LMC) is a notable phenomenon in the loss landscapes of neural networks, wherein independently trained models have been observed to be connected—up to permutation symmetries—by linear paths in parameter space along which the loss remains consistently low. This observation challenges classical views of non-convex optimization and has implications for model ensembling, generalization, and our understanding of neural loss geometry. Inspired by recent studies on LMC in standard neural networks, we systematically investigate this phenomenon within Mixture-of-Experts (MoE) architectures—a class of models known for their scalability and computational efficiency, which combine traditional neural networks—referred to as experts—through a learnable gating mechanism. We begin by conducting a comprehensive analysis of both dense and sparse gating regimes, demonstrating that the symmetries inherent to MoE architectures are fully characterized by permutations acting on both the expert components and the gating function. Building on these foundational findings, we propose a matching algorithm that enables alignment between independently trained MoEs, thereby facilitating the discovery of LMC. Finally, we empirically validate the presence of LMC using our proposed algorithm across diverse MoE configurations—including dense, sparse, and shared-expert variants—under a wide range of model settings and datasets of varying scales and modalities. Our results confirm the existence of LMC in MoE architectures and offer fundamental insights into the functional landscape and optimization dynamics of deep learning models.