arXiv subjects

Rui Yuan

Publications and source records attributed to Rui Yuan.

At least 19 recordsLinked to original sources

arXiv2026

YouZhi: Towards High-Concurrency Financial LLMs via Adaptive GQA-to-MLA Transition

Large language models (LLMs) drive significant financial innovations, yet their high-concurrency deployment is severely bottlenecked by KV cache memory overhead, which inflates infrastructure costs and throttles scalability. To address this, we propose YouZhi-LLM, a highly efficient financial LLM empowered by a comprehensive structural transition and training pipeline natively built on the Huawei Ascend ecosystem. At its algorithmic core, YouZhi-LLM features a layer-adaptive GQA-to-MLA transition framework that dynamically assigns per-layer FreqFold sizes, maximizing KV-cache compression while minimizing perplexity degradation. To recover representation capacity and inject domain expertise, the Ascend-based training pipeline seamlessly integrates generalized knowledge distillation with financial-specific supervised fine-tuning. Evaluations demonstrate the superiority of this systematic approach, with the adaptive transition reducing perplexity degradation by up to 35% over uniform baselines. Crucially, when evaluated on Ascend NPUs via vLLM-Ascend, the massive KV-cache reduction translates directly into deployment efficiency. Compared to their respective base models, YouZhi-7B yields a 12.3% improvement in average financial benchmark score alongside a 2.69$\times$ increase in maximum concurrency; similarly, YouZhi-14B achieves a 7.0% accuracy gain and a 2.43$\times$ concurrency boost, establishing a new paradigm for cost-effective, high-throughput financial inference.

cs.CL↗

arXiv2026

CAD 100K: A Comprehensive Multi-Task Dataset for Car Related Visual Anomaly Detection

Multi-task visual anomaly detection is critical for car-related manufacturing quality assessment. However, existing methods remain task-specific, hindered by the absence of a unified benchmark for multi-task evaluation. To fill in this gap, We present the CAD Dataset, a large-scale and comprehensive benchmark designed for car-related multi-task visual anomaly detection. The dataset contains over 100 images crossing 7 vehicle domains and 3 tasks, providing models a comprehensive view for car-related anomaly detection. It is the first car-related anomaly dataset specialized for multi-task learning(MTL), while combining synthesis data augmentation for few-shot anomaly images. We implement a multi-task baseline and conduct extensive empirical studies. Results show MTL promotes task interaction and knowledge transfer, while also exposing challenging conflicts between tasks. The CAD dataset serves as a standardized platform to drive future advances in car-related multi-task visual anomaly detection.

cs.CV↗

arXiv2026

Beyond Uniform Credit: Causal Credit Assignment for Policy Optimization

Policy gradient methods for language model reasoning, such as GRPO and DAPO, assign uniform credit to all generated tokens - the filler phrase "Let me think" receives the same gradient update as the critical calculation "23 + 45 = 68." We propose counterfactual importance weighting: mask reasoning spans, measure the drop in answer probability, and upweight tokens accordingly during policy gradient updates. Our method requires no auxiliary models or external annotation, instead importance is estimated directly from the policy model's own probability shifts. Experiments on GSM8K across three models spanning the Qwen and Llama families demonstrate consistent improvements over uniform baselines and faster convergence to equivalent accuracy. Inverting the importance signal hurts performance, confirming we capture genuine causal structure rather than noise. Analysis shows the method correctly prioritizes calculation steps over scaffolding text. We view these findings as establishing counterfactual importance weighting as a foundation for further research rather than a complete solution.

cs.CL↗

arXiv2026

AlignTune: Modular Toolkit for Post-Training Alignment of Large Language Models

Post-training alignment is central to deploying large language models (LLMs), yet practical workflows remain split across backend-specific tools and ad-hoc glue code, making experiments hard to reproduce. We identify backend interference, reward fragmentation, and irreproducible pipelines as key obstacles in alignment research. We introduce AlignTune, a modular toolkit exposing a unified interface for supervised fine-tuning (SFT) and RLHF-style optimization with interchangeable TRL and Unsloth backends. AlignTune standardizes configuration, provides an extensible reward layer (rule-based and learned), and integrates evaluation over standard benchmarks and custom tasks. By isolating backend-specific logic behind a single factory boundary, AlignTune enables controlled comparisons and reproducible alignment experiments.

cs.CL↗

arXiv2026

Beyond KL Divergence: Policy Optimization with Flexible Bregman Divergences for LLM Reasoning

Policy optimization methods like Group Relative Policy Optimization (GRPO) and its variants have achieved strong results on mathematical reasoning and code generation tasks. Despite extensive exploration of reward processing strategies and training dynamics, all existing group-based methods exclusively use KL divergence for policy regularization, leaving the choice of divergence function unexplored. We introduce Group-Based Mirror Policy Optimization (GBMPO), a framework that extends group-based policy optimization to flexible Bregman divergences, including hand-designed alternatives (L2 in probability space) and learned neural mirror maps. On GSM8K mathematical reasoning, hand-designed ProbL2-GRPO achieves 86.7% accuracy, improving +5.5 points over the Dr. GRPO baseline. On MBPP code generation, neural mirror maps reach 60.1-60.8% pass@1, with random initialization already capturing most of the benefit. While evolutionary strategies meta-learning provides marginal accuracy improvements, its primary value lies in variance reduction ($\pm$0.2 versus $\pm$0.6) and efficiency gains (15% shorter responses on MBPP), suggesting that random initialization of neural mirror maps is sufficient for most practical applications. These results establish divergence choice as a critical, previously unexplored design dimension in group-based policy optimization for LLM reasoning.

cs.LG↗

arXiv2025

The Initialization Determines Whether In-Context Learning Is Gradient Descent

In-context learning (ICL) in large language models (LLMs) is a striking phenomenon, yet its underlying mechanisms remain only partially understood. Previous work connects linear self-attention (LSA) to gradient descent (GD), this connection has primarily been established under simplified conditions with zero-mean Gaussian priors and zero initialization for GD. However, subsequent studies have challenged this simplified view by highlighting its overly restrictive assumptions, demonstrating instead that under conditions such as multi-layer or nonlinear attention, self-attention performs optimization-like inference, akin to but distinct from GD. We investigate how multi-head LSA approximates GD under more realistic conditions specifically when incorporating non-zero Gaussian prior means in linear regression formulations of ICL. We first extend multi-head LSA embedding matrix by introducing an initial estimation of the query, referred to as the initial guess. We prove an upper bound on the number of heads needed for ICL linear regression setup. Our experiments confirm this result and further observe that a performance gap between one-step GD and multi-head LSA persists. To address this gap, we introduce yq-LSA, a simple generalization of single-head LSA with a trainable initial guess yq. We theoretically establish the capabilities of yq-LSA and provide experimental validation on linear regression tasks, thereby extending the theory that bridges ICL and GD. Finally, inspired by our findings in the case of linear regression, we consider widespread LLMs augmented with initial guess capabilities, and show that their performance is improved on a semantic similarity task.

cs.LG↗

arXiv2025

HunyuanVideo 1.5 Technical Report

We present HunyuanVideo 1.5, a lightweight yet powerful open-source video generation model that achieves state-of-the-art visual quality and motion coherence with only 8.3 billion parameters, enabling efficient inference on consumer-grade GPUs. This achievement is built upon several key components, including meticulous data curation, an advanced DiT architecture featuring selective and sliding tile attention (SSTA), enhanced bilingual understanding through glyph-aware text encoding, progressive pre-training and post-training, and an efficient video super-resolution network. Leveraging these designs, we developed a unified framework capable of high-quality text-to-video and image-to-video generation across multiple durations and resolutions. Extensive experiments demonstrate that this compact and proficient model establishes a new state-of-the-art among open-source video generation models. By releasing the code and model weights, we provide the community with a high-performance foundation that lowers the barrier to video creation and research, making advanced video generation accessible to a broader audience. All open-source assets are publicly available at https://github.com/Tencent-Hunyuan/HunyuanVideo-1.5.

cs.CV↗

arXiv2025

Wireless Power Transfer and Intent-Driven Network Optimization in AAVs-assisted IoT for 6G Sustainable Connectivity

Autonomous Aerial Vehicle (AAV)-assisted Internet of Things (IoT) represents a collaborative architecture in which AAV allocate resources over 6G links to jointly enhance user-intent interpretation and overall network performance. Owing to this mutual dependence, improvements in intent inference and policy decisions on one component reinforce the efficiency of others, making highly reliable intent prediction and low-latency action execution essential. Although numerous approaches can model intent relationships, they encounter severe obstacles when scaling to high-dimensional action sequences and managing intensive on-board computation. We propose an Intent-Driven Framework for Autonomous Network Optimization comprising prediction and decision modules. First, implicit intent modeling is adopted to mitigate inaccuracies arising from ambiguous user expressions. For prediction, we introduce Hyperdimensional Transformer (HDT), which embeds data into a Hyperdimensional space via Hyperdimensional vector encoding and replaces standard matrix and attention operations with symbolic Hyperdimensional computations. For decision-making, where AAV must respond to user intent while planning trajectories, we design Double Actions based Multi-Agent Proximal Policy Optimization (DA-MAPPO). Building upon MAPPO, it samples actions through two independently parameterized networks and cascades the user-intent network into the trajectory network to maintain action dependencies. We evaluate our framework on a real IoT action dataset with authentic wireless data. Experimental results demonstrate that HDT and DA-MAPPO achieve superior performance across diverse scenarios.

cs.AI↗

arXiv2025

A Digital SRAM-Based Compute-In-Memory Macro for Weight-Stationary Dynamic Matrix Multiplication in Transformer Attention Score Computation

Compute-in-memory (CIM) techniques are widely employed in energy-efficient artificial intelligent (AI) processors. They alleviate power and latency bottlenecks caused by extensive data movements between compute and storage units. To extend these benefits to Transformer, this brief proposes a digital CIM macro to compute attention score. To eliminate dynamic matrix multiplication (MM), we reconstruct the computation as static MM using a combined QK-weight matrix, so that inputs can be directly fed to a single CIM macro to obtain the score results. However, this introduces a new challenge of 2-input static MM. The computation is further decomposed into four groups of bit-serial logical and addition operations. This allows 2-input to directly activate the word line via AND gate, thus realizing 2-input static MM with minimal overhead. A hierarchical zero-value bit skipping mechanism is introduced to prioritize skipping zero-value bits in the 2-input case. This mechanism effectively utilizes data sparsity of 2-input, significantly reducing redundant operations. Implemented in a 65-nm process, the 0.35 mm2 macro delivers 42.27 GOPS at 1.24 mW, yielding 34.1 TOPS/W energy and 120.77 GOPS/mm2 area efficiency. Compared to CPUs and GPUs, it achieves ~25x and ~13x higher efficiency, respectively. Against other Transformer-CIMs, it demonstrates at least 7x energy and 2x area efficiency gains, highlighting its strong potential for edge intelligence.

cs.AR↗

arXiv2025

From predictions to confidence intervals: an empirical study of conformal prediction methods for in-context learning

Transformers have become a standard architecture in machine learning, demonstrating strong in-context learning (ICL) abilities that allow them to learn from the prompt at inference time. However, uncertainty quantification for ICL remains an open challenge, particularly in noisy regression tasks. This paper investigates whether ICL can be leveraged for distribution-free uncertainty estimation, proposing a method based on conformal prediction to construct prediction intervals with guaranteed coverage. While traditional conformal methods are computationally expensive due to repeated model fitting, we exploit ICL to efficiently generate confidence intervals in a single forward pass. Our empirical analysis compares this approach against ridge regression-based conformal methods, showing that conformal prediction with in-context learning (CP with ICL) achieves robust and scalable uncertainty estimates. Additionally, we evaluate its performance under distribution shifts and establish scaling laws to guide model training. These findings bridge ICL and conformal prediction, providing a theoretically grounded and new framework for uncertainty quantification in transformer-based models.

stat.ML↗

arXiv2025

Rethinking Pseudo-Label Guided Learning for Weakly Supervised Temporal Action Localization from the Perspective of Noise Correction

Pseudo-label learning methods have been widely applied in weakly-supervised temporal action localization. Existing works directly utilize weakly-supervised base model to generate instance-level pseudo-labels for training the fully-supervised detection head. We argue that the noise in pseudo-labels would interfere with the learning of fully-supervised detection head, leading to significant performance leakage. Issues with noisy labels include:(1) inaccurate boundary localization; (2) undetected short action clips; (3) multiple adjacent segments incorrectly detected as one segment. To target these issues, we introduce a two-stage noisy label learning strategy to harness every potential useful signal in noisy labels. First, we propose a frame-level pseudo-label generation model with a context-aware denoising algorithm to refine the boundaries. Second, we introduce an online-revised teacher-student framework with a missing instance compensation module and an ambiguous instance correction module to solve the short-action-missing and many-to-one problems. Besides, we apply a high-quality pseudo-label mining loss in our online-revised teacher-student framework to add different weights to the noisy labels to train more effectively. Our model outperforms the previous state-of-the-art method in detection accuracy and inference speed greatly upon the THUMOS14 and ActivityNet v1.2 benchmarks.

cs.CV↗

arXiv2024

DarkSHINE Baseline Design Report: Physics Prospects and Detector Technologies

DarkSHINE is a newly proposed fixed-target experiment initiative to search for the invisible decay of Dark Photon via missing energy/momentum signatures, based on the high repetition rate electron beam to be deployed/delivered by the Shanghai High repetition rate XFEL and Extreme light facility (SHINE). This report elaborates the baseline design of DarkSHINE experiment by introducing the physics goals, experimental setups, details of each sub-detector system technical designs, signal and backgground modelings, expected search sensitivities and future prospects, which mark an important step towards the further prototyping and technical demonstrations.

physics.ins-det↗

arXiv2024

Weakly Supervised Temporal Action Localization via Dual-Prior Collaborative Learning Guided by Multimodal Large Language Models

Recent breakthroughs in Multimodal Large Language Models (MLLMs) have gained significant recognition within the deep learning community, where the fusion of the Video Foundation Models (VFMs) and Large Language Models(LLMs) has proven instrumental in constructing robust video understanding systems, effectively surmounting constraints associated with predefined visual tasks. These sophisticated MLLMs exhibit remarkable proficiency in comprehending videos, swiftly attaining unprecedented performance levels across diverse benchmarks. However, their operation demands substantial memory and computational resources, underscoring the continued importance of traditional models in video comprehension tasks. In this paper, we introduce a novel learning paradigm termed MLLM4WTAL. This paradigm harnesses the potential of MLLM to offer temporal action key semantics and complete semantic priors for conventional Weakly-supervised Temporal Action Localization (WTAL) methods. MLLM4WTAL facilitates the enhancement of WTAL by leveraging MLLM guidance. It achieves this by integrating two distinct modules: Key Semantic Matching (KSM) and Complete Semantic Reconstruction (CSR). These modules work in tandem to effectively address prevalent issues like incomplete and over-complete outcomes common in WTAL methods. Rigorous experiments are conducted to validate the efficacy of our proposed approach in augmenting the performance of various heterogeneous WTAL models.

cs.CV↗

arXiv2024

Hunyuan-Large: An Open-Source MoE Model with 52 Billion Activated Parameters by Tencent

In this paper, we introduce Hunyuan-Large, which is currently the largest open-source Transformer-based mixture of experts model, with a total of 389 billion parameters and 52 billion activation parameters, capable of handling up to 256K tokens. We conduct a thorough evaluation of Hunyuan-Large's superior performance across various benchmarks including language understanding and generation, logical reasoning, mathematical problem-solving, coding, long-context, and aggregated tasks, where it outperforms LLama3.1-70B and exhibits comparable performance when compared to the significantly larger LLama3.1-405B model. Key practice of Hunyuan-Large include large-scale synthetic data that is orders larger than in previous literature, a mixed expert routing strategy, a key-value cache compression technique, and an expert-specific learning rate strategy. Additionally, we also investigate the scaling laws and learning rate schedule of mixture of experts models, providing valuable insights and guidances for future model development and optimization. The code and checkpoints of Hunyuan-Large are released to facilitate future innovations and applications. Codes: https://github.com/Tencent/Hunyuan-Large Models: https://huggingface.co/tencent/Tencent-Hunyuan-Large

cs.CL↗

arXiv2024

Design of a LYSO Crystal Electromagnetic Calorimeter for DarkSHINE Experiment

This paper presents the design and optimization of a LYSO crystal electromagnetic calorimeter (ECAL) for the DarkSHINE experiment, which aims to search for dark photons as potential mediators of dark forces. The ECAL design was evaluated through comprehensive simulations, focusing on optimizing dimensions, material selection, energy distribution, and energy resolution. The ECAL configuration consists of 21$\times$21$\times$11 LYSO crystals, each measuring 2.5$\times$2.5$\times$4 cm$^3$, arranged in a staggered layout to improve signal detection efficiency. A 4 GeV energy dynamic range was established to ensure accurate energy measurements without saturation, which is essential for background rejection and signal identification. A detailed digitization model was developed to simulate the scintillation, SiPM, and ADC behaviors, providing a more realistic representation of detector performance. Additionally, the study assessed radiation damage in the ECAL region, highlighting the necessity of radiation-resistant scintillators and silicon sensors.

physics.ins-det↗

arXiv2024

Probing Neutral Triple Gauge Couplings via $\boldsymbol{Z\gamma\,(\ell^+\ell^-\gamma)}$ Production at $\boldsymbol{e^+e^-}$ Colliders

Neutral triple gauge couplings (nTGCs) are absent in the Standard Model (SM) and at the dimension-6 level in the Standard Model Effective Field Theory (SMEFT), arising first from dimension-8 operators. As such, they provide a unique window for probing new physics beyond the SM. These dimension-8 operators can be mapped to nTGC form factors whose structure is consistent with the spontaneously-broken electroweak gauge symmetry of the SM. In this work, we study the probes of nTGCs in the reaction $e^+e^-\to Z\gamma$ with $Z\to\ell^+\ell^-\,(\ell =e,\mu)$ at an $e^+e^-$ collider. We perform a detector-level simulation and analysis of this reaction at the Circular Electron Positron Collider (CEPC) with collision energy $\sqrt{s} = 240$ GeV and an integrated luminosity of 20 ab$^{-1}$. We present the sensitivity limits on probing the new physics scales of dimension-8 nTGC operators via measurements of the corresponding nTGC form factors.

hep-ph↗

arXiv2024

Enhancing Policy Gradient with the Polyak Step-Size Adaption

Policy gradient is a widely utilized and foundational algorithm in the field of reinforcement learning (RL). Renowned for its convergence guarantees and stability compared to other RL algorithms, its practical application is often hindered by sensitivity to hyper-parameters, particularly the step-size. In this paper, we introduce the integration of the Polyak step-size in RL, which automatically adjusts the step-size without prior knowledge. To adapt this method to RL settings, we address several issues, including unknown f* in the Polyak step-size. Additionally, we showcase the performance of the Polyak step-size in RL through experiments, demonstrating faster convergence and the attainment of more stable policies.

cs.LG↗

arXiv2024

Dynamic and Memory-efficient Shape Based Methodologies for User Type Identification in Smart Grid Applications

Detecting behind-the-meter (BTM) equipment and major appliances at the residential level and tracking their changes in real time is important for aggregators and traditional electricity utilities. In our previous work, we developed a systematic solution called IRMAC to identify residential users' BTM equipment and applications from their imported energy data. As a part of IRMAC, a Similarity Profile (SP) was proposed for dimensionality reduction and extracting self-join similarity from the end users' daily electricity usage data. The proposed SP calculation, however, was computationally expensive and required a significant amount of memory at the user's end. To realise the benefits of edge computing, in this paper, we propose and assess three computationally-efficient updating solutions, namely additive, fixed memory, and codebook-based updating methods. Extensive simulation studies are carried out using real PV users' data to evaluate the performance of the proposed methods in identifying PV users, tracking changes in real time, and examining memory usage. We found that the Codebook-based solution reduces more than 30\% of the required memory without compromising the performance of extracting users' features. When the end users' data storage and computation speed are concerned, the fixed-memory method outperforms the others. In terms of tracking the changes, different variations of the fixed-memory method show various inertia levels, making them suitable for different applications.

eess.SY↗