A Regret Minimization Framework on Preference Learning in Large Language Models

Reinforcement learning with verifiable rewards (RLVR) has enabled progress on reasoning-intensive tasks by relying on task-specific verifiers that provide automated correctness signals. However, many realistic language tasks are difficult to equip with reliable verifiers, motivating a growing reliance on reinforcement learning from human feedback (RLHF). In this setting, we argue that a closer examination of how human feedback should be interpreted is essential. We introduce Regret-based Preference Optimization $(\textbf{RePO})$, which reframes RLHF through $\textit{regret minimization}$ rather than reward maximization. Human preferences are often shaped by $\textit{prospective}$ anticipation of outcomes and $\textit{counterfactual}$ comparisons to alternative behaviors, rather than by immediate, outcome-independent utility. $\textbf{RePO}$ captures this structure by modeling preferences as behavior-conditioned assessments of relative suboptimality. Experiments on mathematical reasoning benchmarks and human preference datasets demonstrate consistent performance gains, indicating that $\textbf{RePO}$ is an effective and human-aligned approach for training large language models.

Efficient Process Reward Modeling via Contrastive Mutual Information

Recent research has devoted considerable effort to verifying the intermediate reasoning steps of chain-of-thought (CoT) trajectories using process reward models (PRMs) and other verifier models. However, training a PRM typically requires human annotators to assign reward scores to each reasoning step, which is both costly and time-consuming. Existing automated approaches, such as Monte Carlo (MC) estimation, also demand substantial computational resources due to repeated LLM rollouts. To overcome these limitations, we propose contrastive pointwise mutual information (CPMI), a novel automatic reward labeling method that leverages the model's internal probability to infer step-level supervision while significantly reducing the computational burden of annotating dataset. CPMI quantifies how much a reasoning step increases the mutual information between the step and the correct target answer relative to hard-negative alternatives. This contrastive signal serves as a proxy for the step's contribution to the final solution and yields a reliable reward. The experimental results show that CPMI-based labeling reduces dataset construction time by 84% and token generation by 98% compared to MC estimation, while achieving higher accuracy on process-level evaluations and mathematical reasoning benchmarks.

QFlowNet: Fast, Diverse, and Efficient Unitary Synthesis with Generative Flow Networks

Unitary Synthesis, the decomposition of a unitary matrix into a sequence of quantum gates, is a fundamental challenge in quantum compilation. Prevailing reinforcement learning (RL) approaches are often hampered by sparse reward signals, which necessitate complex reward shaping or long training times, and typically converge to a single policy, lacking solution diversity. In this work, we propose QFlowNet, a novel framework that learns efficiently from sparse signals by pairing a Generative Flow Network (GFlowNet) with Transformers. Our approach addresses two key challenges. First, the GFlowNet framework is fundamentally designed to learn a diverse policy that samples solutions proportional to their reward, overcoming the single-solution limitation of RL while offering faster inference than other generative models like diffusion. Second, the Transformers act as a powerful encoder, capturing the non-local structure of unitary matrices and compressing a high-dimensional state into a dense latent representation for the policy network. Our agent achieves an overall success rate of 99.7% on a 3-qubit benchmark(lengths 1-12) and discovers a diverse set of compact circuits, establishing QFlowNet as an efficient and diverse paradigm for unitary synthesis.

MVP-LAM: Learning Action-Centric Latent Action via Cross-Viewpoint Reconstruction

Learning \emph{latent actions} from diverse human videos enables scaling robot learning beyond embodiment-specific robot datasets, and these latent actions have recently been used as pseudo-action labels for vision-language-action (VLA) model pretraining. To make VLA pretraining effective, latent actions should contain information about the underlying agent's actions despite the absence of ground-truth labels. We propose \textbf{M}ulti-\textbf{V}iew\textbf{P}oint \textbf{L}atent \textbf{A}ction \textbf{M}odel (\textbf{MVP-LAM}), which learns discrete latent actions that are highly informative about ground-truth actions from time-synchronized multi-view videos. MVP-LAM trains latent actions with a \emph{cross-viewpoint reconstruction} objective, so that a latent action inferred from one view must explain the future in another view, reducing reliance on viewpoint-specific cues. On Bridge V2, MVP-LAM produces more action-centric latent actions, achieving higher mutual information with ground-truth actions and improved action prediction, including under out-of-distribution evaluation. Finally, pretraining VLAs with MVP-LAM latent actions improves downstream manipulation performance on the SIMPLER and LIBERO-Long benchmarks.

Learning Generalizable Visuomotor Policy through Dynamics-Alignment

Behavior cloning methods for robot learning suffer from poor generalization due to limited data support beyond expert demonstrations. Recent approaches leveraging video prediction models have shown promising results by learning rich spatiotemporal representations from large-scale datasets. However, these models learn action-agnostic dynamics that cannot distinguish between different control inputs, limiting their utility for precise manipulation tasks and requiring large pretraining datasets. We propose a Dynamics-Aligned Flow Matching Policy (DAP) that integrates dynamics prediction into policy learning. Our method introduces a novel architecture where policy and dynamics models provide mutual corrective feedback during action generation, enabling self-correction and improved generalization. Empirical validation demonstrates generalization performance superior to baseline methods on real-world robotic manipulation tasks, showing particular robustness in OOD scenarios including visual distractions and lighting variations.

Mitigating Attention Localization in Small Scale: Self-Attention Refinement via One-step Belief Propagation

Transformer-based self-attention mechanism serves as the core of modern language models, yet it often suffers from localization, where attentions collapse onto a limited subset of tokens and fail to capture long-range dependencies. To address this issue, we propose Self-Attention One-step Belief Propagation (SAOBP), a refinement framework that injects multi-hop relationships through a belief propagation process. To interpret and quantify these interactions, we introduce Global Token Dependency (GTD) that captures the relative contribution of multihop connections within the attention graph. Empirical results indicate that SAOBP helps prevent entropy collapse in deeper layers and adaptively maintains GTD at task-appropriate levels, thereby supporting improvements in model performance. Importantly, we observe competitive gains in small-scale models, highlighting its potential for improving inference quality in resource-constrained scenarios.

RAISE: Enhancing Scientific Reasoning in LLMs via Step-by-Step Retrieval

Scientific reasoning requires not only long-chain reasoning processes, but also knowledge of domain-specific terminologies and adaptation to updated findings. To deal with these challenges for scientific reasoning, we introduce RAISE, a step-by-step retrieval-augmented framework which retrieves logically relevant documents from in-the-wild corpus. RAISE is divided into three steps: problem decomposition, logical query generation, and logical retrieval. We observe that RAISE consistently outperforms other baselines on scientific reasoning benchmarks. We analyze that unlike other baselines, RAISE retrieves documents that are not only similar in terms of the domain knowledge, but also documents logically more relevant.

Policy-labeled Preference Learning: Is Preference Enough for RLHF?

To design rewards that align with human goals, Reinforcement Learning from Human Feedback (RLHF) has emerged as a prominent technique for learning reward functions from human preferences and optimizing policies via reinforcement learning algorithms. However, existing RLHF methods often misinterpret trajectories as being generated by an optimal policy, causing inaccurate likelihood estimation and suboptimal learning. Inspired by Direct Preference Optimization framework which directly learns optimal policy without explicit reward, we propose policy-labeled preference learning (PPL), to resolve likelihood mismatch issues by modeling human preferences with regret, which reflects behavior policy information. We also provide a contrastive KL regularization, derived from regret-based principles, to enhance RLHF in sequential decision making. Experiments in high-dimensional continuous control tasks demonstrate PPL's significant improvements in offline RLHF performance and its effectiveness in online settings.

Constructing Fair Latent Space for Intersection of Fairness and Explainability

As the use of machine learning models has increased, numerous studies have aimed to enhance fairness. However, research on the intersection of fairness and explainability remains insufficient, leading to potential issues in gaining the trust of actual users. Here, we propose a novel module that constructs a fair latent space, enabling faithful explanation while ensuring fairness. The fair latent space is constructed by disentangling and redistributing labels and sensitive attributes, allowing the generation of counterfactual explanations for each type of information. Our module is attached to a pretrained generative model, transforming its biased latent space into a fair latent space. Additionally, since only the module needs to be trained, there are advantages in terms of time and cost savings, without the need to train the entire generative model. We validate the fair latent space with various fairness metrics and demonstrate that our approach can effectively provide explanations for biased decisions and assurances of fairness.

Mix from Failure: Confusion-Pairing Mixup for Long-Tailed Recognition

Long-tailed image recognition is a computer vision problem considering a real-world class distribution rather than an artificial uniform. Existing methods typically detour the problem by i) adjusting a loss function, ii) decoupling classifier learning, or iii) proposing a new multi-head architecture called experts. In this paper, we tackle the problem from a different perspective to augment a training dataset to enhance the sample diversity of minority classes. Specifically, our method, namely Confusion-Pairing Mixup (CP-Mix), estimates the confusion distribution of the model and handles the data deficiency problem by augmenting samples from confusion pairs in real-time. In this way, CP-Mix trains the model to mitigate its weakness and distinguish a pair of classes it frequently misclassifies. In addition, CP-Mix utilizes a novel mixup formulation to handle the bias in decision boundaries that originated from the imbalanced dataset. Extensive experiments demonstrate that CP-Mix outperforms existing methods for long-tailed image recognition and successfully relieves the confusion of the classifier.