arXiv subjects

Xinrui Zu

Publications and source records attributed to Xinrui Zu.

6 recordsLinked to original sources

arXiv2026

Modularized Reinforcement Learning on LLMs: From MDP Creation to Exploration and Learning

Reinforcement learning (RL) has become central to LLM post-training, yet the methods that dominate current pipelines, PPO and GRPO, represent only a narrow slice of what RL offers. Understanding why these methods prevail, and what alternatives exist, requires a principled examination of the design decisions that underlie any RL algorithm. This survey organizes that examination around three stages of algorithm construction. We begin with MDP creation: how the reward function, state space, action space, termination condition, and discount factor are, or could be, defined for LLM training. We then turn to exploration, covering temperature sampling, entropy regularization, intrinsic motivation, tree search, and curriculum learning. Finally, we address learning along four classical RL dimensions: model-free versus model-based, value-based versus policy-based versus actor-critic, on-policy versus off-policy, and credit assignment, including both Monte Carlo methods, which rely on full return estimates, and bootstrapping methods, which update estimates using other learned predictions. Mapping the LLM literature onto this taxonomy reveals a strikingly non-uniform distribution of research effort. Critic-free policy gradients and Monte Carlo credit assignment are densely populated, while value-based methods, off-policy actor-critic training, and bootstrapping-based credit assignment remain largely unexplored despite well-established counterparts in classical RL. These gaps represent concrete opportunities for transferring proven RL techniques to LLM training. By making these gaps explicit alongside the methods that have proven effective, this survey offers researchers in both RL and LLMs a shared framework for understanding current practice and identifying promising directions for future work.

cs.LG↗

arXiv2026

Reinforcement Learning from Cross-domain Videos with Video Prediction Model

Reinforcement learning from expert videos across visually distinct domains is challenging due to the absence of reward signals and the presence of domain gaps. We introduce XIPER (Cross-domain Video Prediction Reward), a reward model for learning from expert videos collected in a visually different domain, where the agent's appearance differs due to factors such as color, morphology, or the sim-to-real gap. More specifically, XIPER trains a cross-domain video prediction model that maps agent observations into the expert domain and uses the prediction likelihood as a reward signal. Experiments on the DMC Color Suite (8 tasks) and DMC Body Suite (3 tasks) show that XIPER consistently outperforms baselines despite domain gaps such as differences in agent color and morphology. We further analyze XIPER on a sim-to-real transfer dataset, demonstrating that it produces meaningful reward signals for real-robot observations given only simulated expert videos. Code, pretrained models, datasets and video demonstrations can be found on our project webpage: https://sites.google.com/view/xiper

cs.CV↗

arXiv2026

OVA-IB: One vs All Information Bottleneck for Multi-Modal Alignment

Contrastive learning is effective for aligning paired views or modalities, but alignment beyond two modalities remains non-trivial and comparatively underexplored. Pairwise CLIP-style losses decompose multi-modal alignment into independent two-way comparisons and therefore do not explicitly model higher-order dependencies among multiple modalities. Recent beyond-pairwise objectives approach this problem from statistical or geometric perspectives, but arbitrary-modality alignment still lacks a principled criterion for defining what each modality should preserve and compress relative to the others. We revisit arbitrary-modality alignment through the Information Bottleneck principle. In multi-modal learning, sufficiency should preserve information predictable from the remaining modalities, while minimality should compress modality-specific information not supported by them. This naturally leads to a One-vs-All view, where each modality is characterized with respect to the remaining modalities. We propose OVA-IB, an Information Bottleneck framework for arbitrary-modality alignment. OVA-IB optimizes a tractable One-vs-All contrastive lower bound for sufficiency connected to a Dual Total Correlation-style objective, uses a parameter-free geometry-aware projection score, and derives a tractable upper-bound regularizer for minimality by bounding each representation's dependence on its own input with representation distributions induced by the remaining modalities. Experiments on classification, regression, modality-agnostic evaluation, and cross-modal retrieval benchmarks demonstrate strong and robust performance.

cs.LG↗

arXiv2026

StretchBot: A Neuro-Symbolic Framework for Adaptive Guidance with Assistive Robots

Assistive robots have growing potential to support physical wellbeing in home and healthcare settings, for example, by guiding users through stretching or rehabilitation routines. However, existing systems remain largely scripted, which limits their ability to adapt to user state, environmental context, and interaction dynamics. In this work, we present StretchBot, a hybrid neuro-symbolic robotic coach for adaptive assistive guidance. The system combines multimodal perception with knowledge-graph-grounded large language model reasoning to support context-aware adjustments during short stretching sessions while maintaining a structured routine. To complement the system description, we report an exploratory pilot comparison between scripted and adaptive guidance with three participants. The pilot findings suggest that the adaptive condition improved perceived adaptability and contextual relevance, while scripted guidance remained competitive in smoothness and predictability. These results provide preliminary evidence that structured actionable knowledge can help ground language-model-based adaptation in embodied assistive interaction, while also highlighting the need for larger, longitudinal studies to evaluate robustness, generalizability, and long-term user experience.

cs.RO↗

arXiv2024

COT Flow: Learning Optimal-Transport Image Sampling and Editing by Contrastive Pairs

Diffusion models have demonstrated strong performance in sampling and editing multi-modal data with high generation quality, yet they suffer from the iterative generation process which is computationally expensive and slow. In addition, most methods are constrained to generate data from Gaussian noise, which limits their sampling and editing flexibility. To overcome both disadvantages, we present Contrastive Optimal Transport Flow (COT Flow), a new method that achieves fast and high-quality generation with improved zero-shot editing flexibility compared to previous diffusion models. Benefiting from optimal transport (OT), our method has no limitation on the prior distribution, enabling unpaired image-to-image (I2I) translation and doubling the editable space (at both the start and end of the trajectory) compared to other zero-shot editing methods. In terms of quality, COT Flow can generate competitive results in merely one step compared to previous state-of-the-art unpaired image-to-image (I2I) translation methods. To highlight the advantages of COT Flow through the introduction of OT, we introduce the COT Editor to perform user-guided editing with excellent flexibility and quality. The code will be released at https://github.com/zuxinrui/cot_flow.

cs.CV↗

arXiv2022

Deep Recursive Embedding for High-Dimensional Data

Embedding high-dimensional data onto a low-dimensional manifold is of both theoretical and practical value. In this paper, we propose to combine deep neural networks (DNN) with mathematics-guided embedding rules for high-dimensional data embedding. We introduce a generic deep embedding network (DEN) framework, which is able to learn a parametric mapping from high-dimensional space to low-dimensional space, guided by well-established objectives such as Kullback-Leibler (KL) divergence minimization. We further propose a recursive strategy, called deep recursive embedding (DRE), to make use of the latent data representations for boosted embedding performance. We exemplify the flexibility of DRE by different architectures and loss functions, and benchmarked our method against the two most popular embedding methods, namely, t-distributed stochastic neighbor embedding (t-SNE) and uniform manifold approximation and projection (UMAP). The proposed DRE method can map out-of-sample data and scale to extremely large datasets. Experiments on a range of public datasets demonstrated improved embedding performance in terms of local and global structure preservation, compared with other state-of-the-art embedding methods.

cs.LG↗