Reinforcement learning (RL) post-training for Large Language Models (LLMs) is now scaling to large clusters and running for extended durations to enhance model reasoning performance. However, the scalability of existing RL frameworks is limited, as extreme long-tail skewness in RL trajectory generation causes severe GPU underutilization. Current asynchronous RL systems attempt to mitigate this, but they rely on global weight synchronization between the actor and all rollouts, which creates a rigid model update schedule. This global synchronization is ill-suited for the highly skewed and evolving distribution of trajectory generation latency in RL training, crippling training efficiency. Our key insight is that efficient scaling requires breaking this lockstep through trajectory-level asynchrony, which generates and consumes each trajectory independently. We propose Laminar, a scalable and robust RL post-training system built on a fully decoupled architecture. First, we replace global updates with a tier of relay workers acting as a distributed parameter service. This enables asynchronous and fine-grained weight synchronization, allowing rollouts to pull the latest weight anytime without stalling the actor's training loop. Second, a dynamic repack mechanism consolidates long-tail trajectories onto a few dedicated rollouts, maximizing generation throughput. The fully decoupled design also isolates failures, ensuring robustness for long-running jobs. Our evaluation on a 1024-GPU cluster shows that Laminar achieves up to 5.48$\times$ training throughput speedup over state-of-the-art systems, while reducing model convergence time.

Knowledge distillation is often considered a compression mechanism when judged on the resulting student's accuracy and loss, yet its functional impact is poorly understood. In this work, we quantify the compression capacity of knowledge distillation and the resulting knowledge transfer from a functional perspective, decoupling compression from architectural reduction, which provides an improved understanding of knowledge distillation. We employ hypothesis testing, controls, and random control distillation to understand knowledge transfer mechanisms across data modalities. To rigorously test the breadth and limits of our analyses, we explore multiple distillation variants and analyse distillation scaling laws across model sizes. Our findings demonstrate that, while there is statistically significant knowledge transfer in some modalities and architectures, the extent of this transfer is less pronounced than anticipated, even under conditions designed to maximise knowledge sharing. Notably, in cases of significant knowledge transfer, we identify a consistent and severe asymmetric transfer of negative knowledge to the student, raising safety concerns in knowledge distillation applications. Across 12 experimental setups, 9 architectures, and 7 datasets, our findings show that knowledge distillation functions less as a compression mechanism and more as a data-dependent regulariser with a negative asymmetric payoff.

In the evolving landscape of high-speed agile quadrotor flight, achieving precise trajectory tracking at the platform's operational limits is paramount. Controllers must handle actuator constraints, exhibit robustness to disturbances, and remain computationally efficient for safety-critical applications. In this work, we present a novel neural-augmented feedback controller for agile flight control. The controller addresses individual limitations of existing state-of-the-art control paradigms and unifies their strengths. We demonstrate the controller's capabilities, including the accurate tracking of highly aggressive trajectories that surpass the feasibility of the actuators. Notably, the controller provides universal stability guarantees, enhancing its robustness and tracking performance even in exceedingly disturbance-prone settings. Its nonlinear feedback structure is highly efficient enabling fast computation at high update rates. Moreover, the learning process in simulation is both fast and stable, and the controller's inherent robustness allows direct deployment to real-world platforms without the need for training augmentations or fine-tuning.

With the increasing integration of large language models (LLMs) into open-domain writing, detecting machine-generated text has become a critical task for ensuring content authenticity and trust. Existing approaches rely on statistical discrepancies or model-specific heuristics to distinguish between LLM-generated and human-written text. However, these methods struggle in real-world scenarios due to limited generalization, vulnerability to paraphrasing, and lack of explainability, particularly when facing stylistic diversity or hybrid human-AI authorship. In this work, we propose StyleDecipher, a robust and explainable detection framework that revisits LLM-generated text detection using combined feature extractors to quantify stylistic differences. By jointly modeling discrete stylistic indicators and continuous stylistic representations derived from semantic embeddings, StyleDecipher captures distinctive style-level divergences between human and LLM outputs within a unified representation space. This framework enables accurate, explainable, and domain-agnostic detection without requiring access to model internals or labeled segments. Extensive experiments across five diverse domains, including news, code, essays, reviews, and academic abstracts, demonstrate that StyleDecipher consistently achieves state-of-the-art in-domain accuracy. Moreover, in cross-domain evaluations, it surpasses existing baselines by up to 36.30%, while maintaining robustness against adversarial perturbations and mixed human-AI content. Further qualitative and quantitative analysis confirms that stylistic signals provide explainable evidence for distinguishing machine-generated text. Our source code can be accessed at https://github.com/SiyuanLi00/StyleDecipher.

Large language models (LLMs) often miscommunicate their uncertainty: repeated queries can produce divergent answers, yet generated responses are typically unhedged or hedged in ways that do not reflect this variability. This conveys unfaithful information about the uncertain state of the LLMs' knowledge, creating a faithfulness gap that affects even strong LLMs. We introduce Faithful Uncertainty Tuning (FUT): a fine-tuning approach that teaches instruction-tuned LLMs to express uncertainty faithfully without altering their underlying answer distribution. We construct training data by augmenting model samples with uncertainty hedges (i.e. verbal cues such as 'possibly' or 'likely') aligned with sample consistency, requiring no supervision beyond the model and a set of prompts. We evaluate FUT on open-domain question answering (QA) across multiple models and datasets. Our results show that FUT substantially reduces the faithfulness gap, while preserving QA accuracy and introducing minimal semantic distribution shift. Further analyses demonstrate robustness across decoding strategies, choice of hedgers, and other forms of uncertainty expression (i.e. numerical). These findings establish FUT as a simple and effective way to teach LLMs to communicate uncertainty faithfully.

We propose LayerSync, a domain-agnostic approach for improving the generation quality and the training efficiency of diffusion models. Prior studies have highlighted the connection between the quality of generation and the representations learned by diffusion models, showing that external guidance on model intermediate representations accelerates training. Building on the observation that representation quality varies across diffusion model layers, we show that the most semantically rich representations can act as an intrinsic guidance for weaker ones, reducing the need for external supervision. Our approach, LayerSync, is a self-sufficient, plug-and-play regularizer term with no overhead on diffusion model training and generalizes beyond the visual domain to other modalities. LayerSync requires no pretrained models nor additional data. We extensively evaluate the method on image generation and demonstrate its applicability to other domains such as audio, video, and motion generation. We show that it consistently improves the generation quality and the training efficiency. For example, we speed up the training of flow-based transformer by over 8.75 on ImageNet dataset and improved the generation quality by 23.6%. Figure 1: LayerSync improves training efficiency and generation quality via internal representation alignment. Denoising generative models, such as diffusion (Ho et al., 2020; Song et al., 2020; Song & Ermon, 2019) and flow matching models (Lipman et al., 2023), have demonstrated remarkable success in However, this success comes at a significant computation cost. Thus, a new promising line of research has emerged to improve the training efficiency of these models by improving the models' intermediate representations (Y u et al., 2024; Wang et al., 2025; Wang & He, 2025). It has been shown that the quality of a diffusion model's intermediate representations is intrinsically linked to its generative performance. As a result, explicitly guiding these representations accelerates training and improves generation quality (Y u et al., 2024). Building on this insight, the most dominant approach (Y u et al., 2024; Wang et al., 2025) has been to leverage powerful external guidance from large pre-trained models, by aligning the internal features of a diffusion model with those of high-capacity vision models like DINOv2 (Oquab et al., 2023) or vision-language models (VLMs) like Qwen2-VL (Wang et al., 2024).

End-to-end autonomous driving models trained solely with imitation learning (IL) often suffer from poor generalization. In contrast, reinforcement learning (RL) promotes exploration through reward maximization but faces challenges such as sample inefficiency and unstable convergence. A natural solution is to combine IL and RL. Moving beyond the conventional two-stage paradigm (IL pretraining followed by RL fine-tuning), we propose CoIRL-AD, a competitive dual-policy framework that enables IL and RL agents to interact during training. CoIRL-AD introduces a competition-based mechanism that facilitates knowledge exchange while preventing gradient conflicts. Experiments on the nuScenes dataset show an 18% reduction in collision rate compared to baselines, along with stronger generalization and improved performance on long-tail scenarios. Code is available at: https://github.com/SEU-zxj/CoIRL-AD.

The increasing popularity of portable ECG systems and the growing demand for privacy-compliant, energy-efficient real-time analysis require new approaches to signal processing at the point of data acquisition. In this context, the edge domain is acquiring increasing importance, as it not only reduces latency times, but also enables an increased level of data security. The FACE project aims to develop an innovative machine learning solution for analysing long-term electrocardiograms that synergistically combines the strengths of edge and cloud computing. In this thesis, various pre-processing steps of ECG signals are analysed with regard to their applicability in the project. The selection of suitable methods in the edge area is based in particular on criteria such as energy efficiency, processing capability and real-time capability.

Recent work has explored a range of model families for human motion generation, including Variational Autoencoders (VAEs), Generative Adversarial Networks (GANs), and diffusion-based models. Despite their differences, many methods rely on over-parameterized input features and auxiliary losses to improve empirical results. These strategies should not be strictly necessary for diffusion models to match the human motion distribution. We show that on par with state-of-the-art results in unconditional human motion generation are achievable with a score-based diffusion model using only careful feature-space normalization and analytically derived weightings for the standard L2 score-matching loss, while generating both motion and shape directly, thereby avoiding slow post hoc shape recovery from joints. We build the method step by step, with a clear theoretical motivation for each component, and provide targeted ablations demonstrating the effectiveness of each proposed addition in isolation.

Tactile perception is crucial for embodied intelligent robots to recognize objects. Vision-based tactile sensors extract object physical attributes multidimensionally using high spatial resolution; however, this process generates abundant redundant information. Furthermore, single-dimensional extraction, lacking effective fusion, fails to fully characterize object attributes. These challenges hinder the improvement of recognition accuracy. To address this issue, this study introduces a two-stream network feature extraction and fusion perception strategy for vision-based tactile systems. This strategy employs a distributed approach to extract internal and external object features. It obtains depth map information through three-dimensional reconstruction while simultaneously acquiring hardness information by measuring contact force data. After extracting features with a convolutional neural network (CNN), weighted fusion is applied to create a more informative and effective feature representation. In standard tests on objects of varying shapes and hardness, the force prediction error is 0.06 N (within a 12 N range). Hardness recognition accuracy reaches 98.0%, and shape recognition accuracy reaches 93.75%. With fusion algorithms, object recognition accuracy in actual grasping scenarios exceeds 98.5%. Focused on object physical attributes perception, this method enhances the artificial tactile system ability to transition from perception to cognition, enabling its use in embodied perception applications.