arXiv subjects

Yujun Li

Publications and source records attributed to Yujun Li.

At least 19 recordsLinked to original sources

arXiv2026

HERMES: A Multi-Granularity Labeling Substrate for Pre-training Data Mixtures

Most data-mixing methods assume the corpus has already been partitioned into groups, and the choice of those groups determines what a mixer can express. Existing labels, including provenance, topic or format taxonomies, and flat embedding clusters, commit to one semantic axis at one granularity; changing the resolution rebuilds the labels. We argue the bottleneck is the label system, not the mixer, and provide a hierarchical one. HERMES is a data-derived labeling substrate: a Learned Semantic Transform followed by 3-stage residual vector quantization annotates each document once into a coarse-to-fine code whose prefix length controls granularity up to approximately 130k cells. At coarse granularity HERMES sits at a plateau with KMeans-family methods on standard clustering metrics, so the contribution is the substrate, not the clusterer. On 1B-parameter, 25B-token pre-training, the hierarchy exposes an interaction fixed-granularity pipelines cannot test: at one prefix length, a combined Stage-2 rule contrast, equal-subbucket coverage versus size-proportional within-bucket quality top-30%, lifts a 16-task capability macro-average by +0.0253; at the next finer level, the same rule loses its measurable edge as candidate pools contract approximately 5x. HERMES reframes data mixture design from choosing among fixed label sets to navigating a reusable, data-derived granularity hierarchy.

cs.LG↗

arXiv2026

GRASP: Geometry-aware Residual Alignment for Scalable Pretraining Data Attribution

Scalable data attribution methods typically assign isolated utility scores to individual training examples. This prevalent additive assumption fundamentally fails to capture critical subset dynamics, including data redundancy and complementary coverage. In this work, we reframe attribution as subset-level counterfactual utility prediction and introduce GRASP, an interaction-aware surrogate. Grounded in a theoretical smoothness lower bound, GRASP explicitly models subset interactions through a quadratic geometric penalty. To achieve pretraining-scale efficiency without relying on hidden oracle tuning, we couple low-dimensional feature sketches with a strictly finite lower-confidence bound selection protocol. Extensive subset-retraining evaluations demonstrate that GRASP decisively outperforms existing scalable baselines. It more than doubles the task-level rank correlation for counterfactual subset fidelity while reducing upfront artifact construction costs by nearly an order of magnitude. Downstream diagnostics further show that this scoring mechanism transfers to language model curation and cross-domain vision selection, establishing a robust foundation for optimizing massive pretraining corpora.

cs.LG↗

arXiv2026

Quality-Aware Collaborative Multi-Positive Contrastive Learning for Sequential Recommendation

The effectiveness of contrastive learning in sequential recommendation hinges on the construction of contrastive views, which ideally should be both semantically consistent and diverse. However, most existing CL-based methods rely on heuristic augmentations that are prone to removing crucial items or disrupting transition patterns, leading to semantic drift. While a few studies have explored learnable augmentations to improve view quality, they often suffer from limited diversity and still necessitate heuristic aids. Furthermore, the quality differences across views are rarely modeled explicitly and adaptively, aggravating the false-positive issue. To address these issues, we propose Quality-aware Collaborative Multi-Positive Contrastive Learning for sequential recommendation. First, we introduce a learnable collaborative sequence augmentation module that generates two augmented views under two complementary collaborative contexts, one based on same-target sequences and the other on similar sequences, thereby enhancing view diversity while preserving intent consistency.Second, we design a quality-aware mechanism, tightly integrated into the model representations, which estimates each view' s quality from the confidence of its augmentation operations and assigns adaptive weights to ensure that high-confidence views contribute more supervision while low-confidence ones contribute less.Extensive experiments on three real-world datasets demonstrate that QCMP-CL outperforms state-of-the-art CL-based sequential recommendation baselines.

cs.IR↗

arXiv2026

GRPO-TTA: Test-Time Visual Tuning for Vision-Language Models via GRPO-Driven Reinforcement Learning

Group Relative Policy Optimization (GRPO) has recently shown strong performance in post-training large language models and vision-language models. It raises a question of whether the GRPO also significantly promotes the test-time adaptation (TTA) of vision language models. In this paper, we propose Group Relative Policy Optimization for Test-Time Adaptation (GRPO-TTA), which adapts GRPO to the TTA setting by reformulating class-specific prompt prediction as a group-wise policy optimization problem. Specifically, we construct output groups by sampling top-K class candidates from CLIP similarity distributions, enabling probability-driven optimization without access to ground-truth labels. Moreover, we design reward functions tailored to test-time adaptation, including alignment rewards and dispersion rewards, to guide effective visual encoder tuning. Extensive experiments across diverse benchmarks demonstrate that GRPO-TTA consistently outperforms existing test-time adaptation methods, with notably larger performance gains under natural distribution shifts.

cs.CV↗

arXiv2026

GEM: Geometric Entropy Mixing for Optimal LLM Data Curation

LLM pre-training efficacy increasingly depends on data composition rather than sheer volume. Yet, optimal mixing is hindered by categorization flaws: human taxonomies suffer from ontological misalignment, and Euclidean clustering fails to address embedding anisotropy. We introduce GEM (Geometric Entropy Mixing), a framework reformulating data curation as a variational problem on the hypersphere augmented with a mixing-balance regularizer. By decoupling the generative prior and optimizing the objective via a provable MM (Minorize-Maximize) algorithm, GEM effectively counteracts the cluster collapse to discover balanced semantic structures invisible to Euclidean heuristics. We employ teacher-student distillation to scale this geometric fidelity to web-scale corpora and introduce the Geometric Influence Score (GIS) for interpretable taxonomy generation. Experiments with 1.1B-parameter models demonstrate that GEM establishes a new state-of-the-art when integrated into mixing strategies like DoReMi and RegMix, improving average downstream accuracy by up to 1.2% and offering a robust coordinate system for predictable data mixing.

cs.LG↗

arXiv2026

PLL Based Sub-/Super-synchronous Resonance Damping Controller for D-PMSG Wind Farm Integrated Power Systems

Existing sub-/super-synchronous (SSO) suppression methods for the direct-drive permanent magnet synchronous generators (D-PMSG) integrated power systems are mainly achieved by external devices or sub-synchronous resonance damping controller (SSRDC) at the converters, facing challenges of considerable control costs, complex parameters tuning, or inadaptability to various operating conditions. To address these problems, this paper proposes an adaptive SSRDC based on the phase-locked loop (PLL) for D-PMSG integrated power systems. Firstly, the PLL parameter is found critical to SSO suppression by a comprehensive sensitivity analysis on the dominant poles of the impedance closed-loop transfer function. Motivated by this finding, this paper then designs a PLL-based SSRDC, which features a simple structure, easy parameter tuning, and flexible adaptability to various operating modes. The simplicity in structure is guaranteed by the avoidance of phase compensation. Benefiting from the simple structure, only one key parameter needs to be tuned. Moreover, two principles of parameter tuning are proposed to enhance the efficiency, robustness, and adaptability of the proposed SSRDC. The controller-hardware-in-the-loop (CHIL) tests verify the validity of the proposed SSRDC under various operating conditions. Finally, some concerns about this method such as frequency estimation, computational efficiency and potential impacts on PLL are thoroughly analyzed and clarified.

eess.SY↗

arXiv2025

Robust Design for Multi-Antenna LEO Satellite Communications with Fractional Delay and Doppler Shifts: An RSMA-OTFS Approach

Low-Earth-orbit (LEO) satellite communication systems face challenges due to high satellite mobility, which hinders the reliable acquisition of instantaneous channel state information at the transmitter (CSIT) and subsequently degrades multi-user transmission performance. This paper investigates a downlink multi-user multi-antenna system, and tackles the above challenges by introducing orthogonal time frequency space (OTFS) modulation and rate-splitting multiple access (RSMA) transmission. Specifically, OTFS enables stable characterization of time-varying channels by representing them in the delay-Doppler domain. However, realistic propagation introduces various inter-symbol and inter-user interference due to non-orthogonal yet practical rectangular pulse shaping, fractional delays, Doppler shifts, and imperfect (statistical) CSIT. In this context, RSMA offers promising robustness for interference mitigation and CSIT imperfections, and hence is integrated with OTFS to provide a comprehensive solution. A compact cross-domain input-output relationship for RSMA-OTFS is established, and an ergodic sum-rate maximization problem is formulated and solved using a weighted minimum mean-square-error based alternating optimization algorithm that does not depend on channel sparsity. Simulation results reveal that the considered practical propagation effects significantly degrade performance if unaddressed. Furthermore, the RSMA-OTFS scheme demonstrates improved ergodic sum-rate and robustness against CSIT uncertainty across various user deployments and CSIT qualities.

eess.SP↗

arXiv2025

Can Language Models Discover Scaling Laws?

Discovering scaling laws for predicting model performance at scale is a fundamental and open-ended challenge, mostly reliant on slow, case specific human experimentation. To investigate the potential for LLMs to automate this process, we collect over 5,000 experiments from existing literature and curate eight diverse scaling law discovery tasks. While existing agents struggle to produce accurate law formulas, this paper introduces SLDAgent, an evolution-based agent that co-optimize the scaling law model and the parameters, enabling it to autonomously explore complex relationships between variables. For the first time, we demonstrates that SLDAgent can automatically discover laws that exhibit consistently more accurate extrapolation than their established, human-derived counterparts across all tasks. Through comprehensive analysis, we elucidate why these discovered laws are superior and verify their practical utility in both pretraining and finetuning applications. This work establishes a new paradigm for agentic scientific discovery, showing that AI systems can understand their own scaling behavior, and can contribute novel and practical knowledge back to the research community.

cs.LG↗

arXiv2025

3D-UIR: 3D Gaussian for Underwater 3D Scene Reconstruction via Physics Based Appearance-Medium Decoupling

Novel view synthesis for underwater scene reconstruction presents unique challenges due to complex light-media interactions. Optical scattering and absorption in water body bring inhomogeneous medium attenuation interference that disrupts conventional volume rendering assumptions of uniform propagation medium. While 3D Gaussian Splatting (3DGS) offers real-time rendering capabilities, it struggles with underwater inhomogeneous environments where scattering media introduces artifacts and inconsistent appearance. In this study, we propose a physics-based framework that disentangles object appearance from water medium effects through tailored Gaussian modeling. Our approach introduces appearance embeddings, which are explicit medium representations for backscatter and attenuation, enhancing scene consistency. In addition, we propose a depth-guided optimization strategy that leverages pseudo-depth maps as supervision with depth regularization and scale penalty terms to improve geometric fidelity. By integrating the proposed appearance and medium modeling components via an underwater imaging model, our approach achieves both high-quality novel view synthesis and physically accurate scene restoration. Experiments demonstrate our significant improvements in rendering quality and restoration accuracy over existing methods. The project page is available at https://bilityniu.github.io/3D-UIR.

cs.CV↗

arXiv2025

Pangu Ultra MoE: How to Train Your Big MoE on Ascend NPUs

Sparse large language models (LLMs) with Mixture of Experts (MoE) and close to a trillion parameters are dominating the realm of most capable language models. However, the massive model scale poses significant challenges for the underlying software and hardware systems. In this paper, we aim to uncover a recipe to harness such scale on Ascend NPUs. The key goals are better usage of the computing resources under the dynamic sparse model structures and materializing the expected performance gain on the actual hardware. To select model configurations suitable for Ascend NPUs without repeatedly running the expensive experiments, we leverage simulation to compare the trade-off of various model hyperparameters. This study led to Pangu Ultra MoE, a sparse LLM with 718 billion parameters, and we conducted experiments on the model to verify the simulation results. On the system side, we dig into Expert Parallelism to optimize the communication between NPU devices to reduce the synchronization overhead. We also optimize the memory efficiency within the devices to further reduce the parameter and activation management overhead. In the end, we achieve an MFU of 30.0% when training Pangu Ultra MoE, with performance comparable to that of DeepSeek R1, on 6K Ascend NPUs, and demonstrate that the Ascend system is capable of harnessing all the training stages of the state-of-the-art language models. Extensive experiments indicate that our recipe can lead to efficient training of large-scale sparse language models with MoE. We also study the behaviors of such models for future reference.

cs.CL↗

arXiv2025

Pangu Ultra: Pushing the Limits of Dense Large Language Models on Ascend NPUs

We present Pangu Ultra, a Large Language Model (LLM) with 135 billion parameters and dense Transformer modules trained on Ascend Neural Processing Units (NPUs). Although the field of LLM has been witnessing unprecedented advances in pushing the scale and capability of LLM in recent years, training such a large-scale model still involves significant optimization and system challenges. To stabilize the training process, we propose depth-scaled sandwich normalization, which effectively eliminates loss spikes during the training process of deep models. We pre-train our model on 13.2 trillion diverse and high-quality tokens and further enhance its reasoning capabilities during post-training. To perform such large-scale training efficiently, we utilize 8,192 Ascend NPUs with a series of system optimizations. Evaluations on multiple diverse benchmarks indicate that Pangu Ultra significantly advances the state-of-the-art capabilities of dense LLMs such as Llama 405B and Mistral Large 2, and even achieves competitive results with DeepSeek-R1, whose sparse model structure contains much more parameters. Our exploration demonstrates that Ascend NPUs are capable of efficiently and effectively training dense models with more than 100 billion parameters. Our model and system will be available for our commercial customers.

cs.CL↗

arXiv2025

Learnable Sequence Augmenter for Triplet Contrastive Learning in Sequential Recommendation

Most existing contrastive learning-based sequential recommendation (SR) methods rely on random operations (e.g., crop, reorder, and substitute) to generate augmented sequences. These methods often struggle to create positive sample pairs that closely resemble the representations of the raw sequences, potentially disrupting item correlations by deleting key items or introducing noisy iterac, which misguides the contrastive learning process. To address this limitation, we propose Learnable sequence Augmentor for triplet Contrastive Learning in sequential Recommendation (LACLRec). Specifically, the self-supervised learning-based augmenter can automatically delete noisy items from sequences and insert new items that better capture item transition patterns, generating a higher-quality augmented sequence. Subsequently, we randomly generate another augmented sequence and design a ranking-based triplet contrastive loss to differentiate the similarities between the raw sequence, the augmented sequence from augmenter, and the randomly augmented sequence, providing more fine-grained contrastive signals. Extensive experiments on three real-world datasets demonstrate that both the sequence augmenter and the triplet contrast contribute to improving recommendation accuracy. LACLRec significantly outperforms the baseline model CL4SRec, and demonstrates superior performance compared to several state-of-the-art sequential recommendation algorithms.

cs.IR↗

arXiv2024

Edge Contrastive Learning: An Augmentation-Free Graph Contrastive Learning Model

Graph contrastive learning (GCL) aims to learn representations from unlabeled graph data in a self-supervised manner and has developed rapidly in recent years. However, edgelevel contrasts are not well explored by most existing GCL methods. Most studies in GCL only regard edges as auxiliary information while updating node features. One of the primary obstacles of edge-based GCL is the heavy computation burden. To tackle this issue, we propose a model that can efficiently learn edge features for GCL, namely AugmentationFree Edge Contrastive Learning (AFECL) to achieve edgeedge contrast. AFECL depends on no augmentation consisting of two parts. Firstly, we design a novel edge feature generation method, where edge features are computed by embedding concatenation of their connected nodes. Secondly, an edge contrastive learning scheme is developed, where edges connecting the same nodes are defined as positive pairs, and other edges are defined as negative pairs. Experimental results show that compared with recent state-of-the-art GCL methods or even some supervised GNNs, AFECL achieves SOTA performance on link prediction and semi-supervised node classification of extremely scarce labels. The source code is available at https://github.com/YujunLi361/AFECL.

cs.LG↗

arXiv2024

Dynamics of Adaptive Continuous Attractor Neural Networks

Attractor neural networks consider that neural information is stored as stationary states of a dynamical system formed by a large number of interconnected neurons. The attractor property empowers a neural system to encode information robustly, but it also incurs the difficulty of rapid update of network states, which can impair information update and search in the brain. To overcome this difficulty, a solution is to include adaptation in the attractor network dynamics, whereby the adaptation serves as a slow negative feedback mechanism to destabilize which are otherwise permanently stable states. In such a way, the neural system can, on one hand, represent information reliably using attractor states, and on the other hand, perform computations wherever rapid state updating is involved. Previous studies have shown that continuous attractor neural networks with adaptation (A-CANNs) exhibits rich dynamical behaviors accounting for various brain functions. In this paper, we present a comprehensive view of the rich diverse dynamics of A-CANNs. Moreover, we provide a unified mathematical framework to understand these different dynamical behaviors, and briefly discuss about their biological implications.

q-bio.NC↗

arXiv2024

Flat-LoRA: Low-Rank Adaptation over a Flat Loss Landscape

Fine-tuning large-scale pre-trained models is prohibitively expensive in terms of computation and memory costs. Low-Rank Adaptation (LoRA), a popular Parameter-Efficient Fine-Tuning (PEFT) method, offers an efficient solution by optimizing only low-rank matrices. Despite recent progress in improving LoRA's performance, the relationship between the LoRA optimization space and the full parameter space is often overlooked. A solution that appears flat in the loss landscape of the LoRA space may still exhibit sharp directions in the full parameter space, potentially compromising generalization. We introduce Flat-LoRA, which aims to identify a low-rank adaptation situated in a flat region of the full parameter space. Instead of adopting the well-established sharpness-aware minimization approach, which incurs significant computation and memory overheads, we employ a Bayesian expectation loss objective to preserve training efficiency. Further, we design a refined random perturbation generation strategy for improved performance and carefully manage memory overhead using random seeds. Experiments across diverse tasks-including mathematical reasoning, coding abilities, dialogue generation, instruction following, and text-to-image generation-demonstrate that Flat-LoRA improves both in-domain and out-of-domain generalization. Code is available at https://github.com/nblt/Flat-LoRA.

cs.LG↗

arXiv2024

Copyleft for Alleviating AIGC Copyright Dilemma: What-if Analysis, Public Perception and Implications

As AIGC has impacted our society profoundly in the past years, ethical issues have received tremendous attention. The most urgent one is the AIGC copyright dilemma, which can immensely stifle the development of AIGC and greatly cost the entire society. Given the complexity of AIGC copyright governance and the fact that no perfect solution currently exists, previous work advocated copyleft on AI governance but without substantive analysis. In this paper, we take a step further to explore the feasibility of copyleft to alleviate the AIGC copyright dilemma. We conduct a mixed-methods study from two aspects: qualitatively, we use a formal what-if analysis to clarify the dilemma and provide case studies to show the feasibility of copyleft; quantitatively, we perform a carefully designed survey to find out how the public feels about copylefting AIGC. The key findings include: a) people generally perceive the dilemma, b) they prefer to use authorized AIGC under loose restriction, and c) they are positive to copyleft in AIGC and willing to use it in the future.

cs.CY↗

arXiv2024

Privacy-Preserving Sequential Recommendation with Collaborative Confusion

Sequential recommendation has attracted a lot of attention from both academia and industry, however the privacy risks associated to gathering and transferring users' personal interaction data are often underestimated or ignored. Existing privacy-preserving studies are mainly applied to traditional collaborative filtering or matrix factorization rather than sequential recommendation. Moreover, these studies are mostly based on differential privacy or federated learning, which often leads to significant performance degradation, or has high requirements for communication. In this work, we address privacy-preserving from a different perspective. Unlike existing research, we capture collaborative signals of neighbor interaction sequences and directly inject indistinguishable items into the target sequence before the recommendation process begins, thereby increasing the perplexity of the target sequence. Even if the target interaction sequence is obtained by attackers, it is difficult to discern which ones are the actual user interaction records. To achieve this goal, we propose a CoLlaborative-cOnfusion seqUential recommenDer, namely CLOUD, which incorporates a collaborative confusion mechanism to edit the raw interaction sequences before conducting recommendation. Specifically, CLOUD first calculates the similarity between the target interaction sequence and other neighbor sequences to find similar sequences. Then, CLOUD considers the shared representation of the target sequence and similar sequences to determine the operation to be performed: keep, delete, or insert. We design a copy mechanism to make items from similar sequences have a higher probability to be inserted into the target sequence. Finally, the modified sequence is used to train the recommender and predict the next item.

cs.IR↗

arXiv2022

Finding Second-Order Stationary Points in Nonconvex-Strongly-Concave Minimax Optimization

We study the smooth minimax optimization problem $\min_{\bf x}\max_{\bf y} f({\bf x},{\bf y})$, where $f$ is $\ell$-smooth, strongly-concave in ${\bf y}$ but possibly nonconvex in ${\bf x}$. Most of existing works focus on finding the first-order stationary points of the function $f({\bf x},{\bf y})$ or its primal function $P({\bf x})\triangleq \max_{\bf y} f({\bf x},{\bf y})$, but few of them focus on achieving second-order stationary points. In this paper, we propose a novel approach for minimax optimization, called Minimax Cubic Newton (MCN), which could find an $\big(\varepsilon,κ^{1.5}\sqrt{ρ\varepsilon}\,\big)$-second-order stationary point of $P({\bf x})$ with calling ${\mathcal O}\big(κ^{1.5}\sqrtρ\varepsilon^{-1.5}\big)$ times of second-order oracles and $\tilde{\mathcal O}\big(κ^{2}\sqrtρ\varepsilon^{-1.5}\big)$ times of first-order oracles, where $κ$ is the condition number and $ρ$ is the Lipschitz continuous constant for the Hessian of $f({\bf x},{\bf y})$. In addition, we propose an inexact variant of MCN for high-dimensional problems to avoid calling expensive second-order oracles. Instead, our method solves the cubic sub-problem inexactly via gradient descent and matrix Chebyshev expansion. This strategy still obtains the desired approximate second-order stationary point with high probability but only requires $\tilde{\mathcal O}\big(κ^{1.5}\ell\varepsilon^{-2}\big)$ Hessian-vector oracle calls and $\tilde{\mathcal O}\big(κ^{2}\sqrtρ\varepsilon^{-1.5}\big)$ first-order oracle calls. To the best of our knowledge, this is the first work that considers the non-asymptotic convergence behavior of finding second-order stationary points for minimax problems without the convex-concave assumptions.

math.OC↗