Ran Yan

Yanjie Song, Yutong Wu, Yangyang Guo et al.

Evolutionary algorithms (EA), a class of stochastic search methods based on the principles of natural evolution, have received widespread acclaim for their exceptional performance in various real-world optimization problems. While researchers worldwide have proposed a wide variety of EAs, certain limitations remain, such as slow convergence speed and poor generalization capabilities. Consequently, numerous scholars actively explore improvements to algorithmic structures, operators, search patterns, etc., to enhance their optimization performance. Reinforcement learning (RL) integrated as a component in the EA framework has demonstrated superior performance in recent years. This paper presents a comprehensive survey on integrating reinforcement learning into the evolutionary algorithm, referred to as reinforcement learning-assisted evolutionary algorithm (RL-EA). We begin with the conceptual outlines of reinforcement learning and the evolutionary algorithm. We then provide a taxonomy of RL-EA. Subsequently, we discuss the RL-EA integration method, the RL-assisted strategy adopted by RL-EA, and its applications according to the existing literature. The RL-assisted procedure is divided according to the implemented functions including solution generation, learnable objective function, algorithm/operator/sub-population selection, parameter adaptation, and other strategies. Additionally, different attribute settings of RL in RL-EA are discussed. In the applications of RL-EA section, we also demonstrate the excellent performance of RL-EA on several benchmarks and a range of public datasets to facilitate a quick comparative study. Finally, we analyze potential directions for future research.

Liyuan Zhang, Jiarui Zhang, Jinwei Yao et al.

Speculative decoding accelerates autoregressive large language model inference by drafting multiple tokens and verifying them in a single target-model forward pass. Recent diffusion-based drafters generate an entire block of tokens in parallel but usually commit to a single draft sequence per verification: once the first mismatch occurs, all subsequent draft tokens are discarded, resulting in a limited acceptance rate. Naively batching more draft candidate sequences only introduces a marginal improvement, as redundant or poorly placed branches increase the cost of drafting and verification without proportionally increasing the number of accepted tokens. We propose D^2SD, a dual diffusion draft speculative decoding framework that organizes candidates into a confidence-guided prefix tree, where the first diffusion drafter generates a block along with per-position confidence scores that are used to identify the most likely rejection boundary and select the top-K prefix ranges for recovery; the second variable-prefix diffusion drafter re-anchors at each selected prefix and proposes alternative continuations in one batched pass; the resulting shared-prefix candidates are jointly verified via cascade attention. Empirically, D^2SD shows clear improvements over both the underlying diffusion approach and strong autoregressive speculative decoding baselines.

Ran Yan, Youhe Jiang, Xiaonan Nie et al.

Training large language models (LLMs) is a computationally intensive task, which is typically conducted in data centers with homogeneous high-performance GPUs. In this paper, we explore an alternative approach by deploying training computations across heterogeneous GPUs to enable better flexibility and efficiency for heterogeneous resource utilization. Toward this end, we propose a novel system, HexiScale, that can flexibly support asymmetric partition of training computations in the scope of data-, pipeline-, and tensor model parallelism. We further formalize the allocation of asymmetric partitioned training computations over a set of heterogeneous GPUs as a constrained optimization problem and propose an efficient hierarchical graph partitioning algorithm. Our approach effectively allocates training computations across heterogeneous GPUs, fully leveraging the available computational power. We compare the performance of HexiScale with state-of-the-art homogeneous and heterogeneous training systems. When training LLMs at different scales (from 7B to 30B), empirical results demonstrate that: (i) compared to state-of-the-art homogeneous baselines running over homogeneous GPUs, HexiScale achieves similar performance when running over heterogeneous GPUs with the same theoretical FLOPS; (ii) compared to state-of-the-art heterogeneous baselines running on the same heterogeneous clusters, HexiScale delivers $1.5\times$ to $2.4\times$ higher throughput.

Wenzheng Song, Ran Yan, Boshu Lei et al.

Advanced visual localization techniques encompass image retrieval challenges and 6 Degree-of-Freedom (DoF) camera pose estimation, such as hierarchical localization. Thus, they must extract global and local features from input images. Previous methods have achieved this through resource-intensive or accuracy-reducing means, such as combinatorial pipelines or multi-task distillation. In this study, we present a novel method called SuperGF, which effectively unifies local and global features for visual localization, leading to a higher trade-off between localization accuracy and computational efficiency. Specifically, SuperGF is a transformer-based aggregation model that operates directly on image-matching-specific local features and generates global features for retrieval. We conduct experimental evaluations of our method in terms of both accuracy and efficiency, demonstrating its advantages over other methods. We also provide implementations of SuperGF using various types of local features, including dense and sparse learning-based or hand-crafted descriptors.

Peihua Mai, Ran Yan, Zhe Huang et al.

Large Language Models (LLMs) excel in natural language understanding by capturing hidden semantics in vector space. This process enriches the value of text embeddings for various downstream tasks, thereby fostering the Embedding-as-a-Service (EaaS) business model. However, the risk of privacy leakage due to direct text transmission to servers remains a critical concern. To address this, we introduce Split-N-Denoise (SnD), an private inference framework that splits the model to execute the token embedding layer on the client side at minimal computational cost. This allows the client to introduce noise prior to transmitting the embeddings to the server, and subsequently receive and denoise the perturbed output embeddings for downstream tasks. Our approach is designed for the inference stage of LLMs and requires no modifications to the model parameters. Extensive experiments demonstrate SnD's effectiveness in optimizing the privacy-utility tradeoff across various LLM architectures and diverse downstream tasks. The results reveal an improvement in performance under the same privacy budget compared to the baselines by over 10\% on average, offering clients a privacy-preserving solution for local privacy protection.

Peihua Mai, Youjia Yang, Ran Yan et al.

State-of-the-art large language models (LLMs) are typically deployed as online services, requiring users to transmit detailed prompts to cloud servers. This raises significant privacy concerns. In response, we introduce ConfusionPrompt, a novel framework for private LLM inference that protects user privacy by: (i) decomposing the original prompt into smaller sub-prompts, and (ii) generating pseudo-prompts alongside the genuine sub-prompts, which are then sent to the LLM. The server responses are later recomposed by the user to reconstruct the final output. This approach offers key advantages over previous LLM privacy protection methods: (i) it integrates seamlessly with existing black-box LLMs, and (ii) it delivers a significantly improved privacy-utility trade-off compared to existing text perturbation methods. We also develop a $(λ, μ, ρ)$-privacy model to formulate the requirements for a privacy-preserving group of prompts and provide a complexity analysis to justify the role of prompt decomposition. Our empirical evaluation shows that ConfusionPrompt achieves significantly higher utility than local inference methods using open-source models and perturbation-based techniques, while also reducing memory consumption compared to open-source LLMs.

Ran Yan, Youhe Jiang, Tianyuan Wu et al.

Maximizing training throughput and cost-efficiency of RL for LLMs is essential to democratize this advanced technique. One promising but challenging approach is to deploy such a computational workflow over heterogeneous GPUs. Unlike conventional large-scale LLM pretraining, RL training generally decomposes into three coupled stages, i.e., rollout generation, reward computation, and policy/value updates, which exhibit markedly different compute intensities, memory footprints, and communication patterns. Recent research shows that fully asynchronous RL training can disaggregate these stages across disjoint hardware pools without sacrificing training stability, creating a great opportunity for real-world heterogeneous deployment. To this end, we present AReaL-Hex, a heterogeneity-aware asynchronous RL training system that effectively schedules how to execute rollout generation and policy model training over heterogeneous GPUs while enforcing data staleness bounds. Concretely, we use a two-phase scheduler: (i) a constrained search with MILP to select per-stage parallelization strategies and workload assignments given a resource budget, and (ii) a graph-partitioning step that allocates heterogeneous GPUs and interconnects to maximize end-to-end throughput. Built atop a fully asynchronous RL architecture, AReaL-Hex maps HBM-I/O-bound generation and compute-bound optimization to more cost-efficient resources and balances their producer-consumer interactions to avoid both idleness and stale rollout trajectories. On the mathematical reasoning task with various model scales (1.5B, 7B, and 14B), compared to homogeneous deployments of state-of-the-art asynchronous RL systems: (i) When maintaining the same total budgets, AReaL-Hex delivers up to 1.50x higher training throughput; (ii) When achieving the same training throughput, AReaL-Hex results in up to 1.46x reduction in training cost.

Ran Yan, Youhe Jiang, Zhuoming Chen et al.

Recent advance in sparse attention mechanisms has demonstrated strong potential for reducing the computational cost of long-context training and inference in large language models (LLMs). Native Sparse Attention (NSA), one state-of-the-art approach, introduces natively trainable, hardware-aligned sparse attention that delivers substantial system-level performance boost while maintaining accuracy comparable to full attention. However, the kernel implementation of NSA forces a loop order that is only efficient with a relatively large number of query heads in each Grouped Query Attention (GQA) group, whereas existing LLMs widely adopt much smaller number of query heads in each GQA group -- such an inconsistency significantly limits the applicability of this sparse algorithmic advance. In this work, we propose Flash Sparse Attention (FSA), an alternative kernel implementation that enables efficient NSA computation across a wide range of popular LLMs with varied smaller number of heads in each GQA group on modern GPUs. Compared to vanilla NSA kernel implementation, our empirical evaluation demonstrates that FSA achieves (i) up to 3.5x and on average 1.6x kernel-level latency reduction, (ii) up to 1.25x and 1.09x on average end-to-end training speedup on state-of-the-art LLMs, and (iii) up to 1.36x and 1.11x on average for prefill-phase speedup in LLM generative inference. Github Repo at https://github.com/Relaxed-System-Lab/Flash-Sparse-Attention.

Jie Ren, Wenteng Liang, Ran Yan et al.

Large-scale Bundle Adjustment (BA) requires massive memory and computation resources which are difficult to be fulfilled by existing BA libraries. In this paper, we propose MegBA, a GPU-based distributed BA library. MegBA can provide massive aggregated memory by automatically partitioning large BA problems, and assigning the solvers of sub-problems to parallel nodes. The parallel solvers adopt distributed Precondition Conjugate Gradient and distributed Schur Elimination, so that an effective solution, which can match the precision of those computed by a single node, can be efficiently computed. To accelerate BA computation, we implement end-to-end BA computation using high-performance primitives available on commodity GPUs. MegBA exposes easy-to-use APIs that are compatible with existing popular BA libraries. Experiments show that MegBA can significantly outperform state-of-the-art BA libraries: Ceres (41.45$\times$), RootBA (64.576$\times$) and DeepLM (6.769$\times$) in several large-scale BA benchmarks. The code of MegBA is available at https://github.com/MegviiRobot/MegBA.

Yan Liang, Youhe Jiang, Ran Yan et al.

Long-context training of large language models (LLMs) is commonly distributed with Context Parallelism (CP) and Head Parallelism (HP), but existing training systems largely assume homogeneous GPU meshes. This paper extends CP and HP to heterogeneous GPU clusters with mixed GPU models and non-uniform network bandwidths, a common setting in production training. We introduce HexiSeq, a system that supports fully asymmetric CP--HP partitioning by assigning sequence shards and attention heads according to device compute, memory, and communication capabilities. We formalize heterogeneous CP--HP allocation as a constrained optimization problem and develop an efficient hierarchical scheduler for finding optimal schedules. We evaluate HexiSeq against state-of-the-art CP and HP baselines on both real and simulated heterogeneous clusters. Across models from 3B to 70B parameters and context lengths up to one million tokens, HexiSeq improves throughput by $1.11\times$ on average and up to $1.19\times$ on mixed H100--A100 testbeds, and by $1.36\times$ on average and up to $1.72\times$ in simulations with 32--128 GPUs spanning up to four GPU models. On FLOP-comparable pairs against homogeneous clusters, HexiSeq reaches throughput close to the strongest homogeneous baseline, showing that heterogeneous clusters can be used efficiently for long-context LLM training.

Zijun Wu, Bingyuan Liu, Ran Yan et al.

Recent advancements in Large Language Models (LLMs) have significantly enhanced their capacity to process long contexts. However, effectively utilizing this long context remains a challenge due to the issue of distraction, where irrelevant information dominates lengthy contexts, causing LLMs to lose focus on the most relevant segments. To address this, we propose a novel training method that enhances LLMs' ability to discern relevant information through a unique combination of retrieval-based data augmentation and contrastive learning. Specifically, during fine-tuning with long contexts, we employ a retriever to extract the most relevant segments, serving as augmented inputs. We then introduce an auxiliary contrastive learning objective to explicitly ensure that outputs from the original context and the retrieved sub-context are closely aligned. Extensive experiments on long single-document and multi-document QA benchmarks demonstrate the effectiveness of our proposed method.

Youhe Jiang, Ran Yan, You Peng et al.

Modern Large Language Model (LLM) serving operates in highly volatile environments characterized by severe runtime dynamics, such as workload fluctuations and elastic cluster autoscaling. Traditional serving systems rely on static, human-engineered serving policies (e.g., scheduling algorithms and rescheduling strategies) to manage these dynamics. However, these policies must navigate deeply intertwined runtime trade-offs (e.g., scheduling overhead vs. execution efficiency, rescheduling frequency vs. reconfiguration overhead), whose optimal balance is workload-specific and shifts continuously as runtime conditions evolve, rendering any fixed policy fundamentally unable to adapt. We propose Autopoiesis, a novel online self-evolving system that shifts LLM serving from static policy deployment to continuous online policy evolution. First, Autopoiesis introduces an LLM-driven program synthesis workflow to evolve serving policies with respect to real-time observed dynamics, where the evolved policies reflect the optimal decision in navigating the complex, multi-dimensional trade-off space. Second, Autopoiesis enables this synthesis process to operate continuously during serving, observing real-world system behavior, and rewriting the policy code as runtime trade-offs shift, thereby transforming policy design from a one-time offline endeavor into an ongoing system component, enabling autonomous adaptation to evolving runtime conditions. Together, we establish a new paradigm: Serving policies are no longer static artifacts designed by humans before deployment, but living code that LLMs continuously evolve throughout deployment to navigate runtime trade-offs beyond human design. We evaluate Autopoiesis across diverse runtime dynamics and show up to 53% and on average 34% improvements over state-of-the-art LLM serving systems.

Guangxin He, Shen Nie, Fengqi Zhu et al.

Diffusion LLMs have attracted growing interest, with plenty of recent work emphasizing their great potential in various downstream tasks; yet the long-context behavior of diffusion LLMs remains largely uncharted. We present a case study of post-training techniques for extending the context window of diffusion LLMs (i.e., LLaDA) without retraining from scratch. We show that a simple modification to the standard Rotary Positional Embeddings (RoPE) extension effectively accommodates the probabilistic modeling inherent in the diffusion process, enabling stable scaling to longer context ranges. We further compare masking strategies used during post-training and analyze their impact on optimization stability and long-range recall. Instantiating these insights, we introduce UltraLLaDA, a diffusion LLM with a 128K-token context window that, in our empirical evaluation on long-context tasks, significantly outperforms training-free baselines. Our experimental results highlight the special positional extension as a key lever for scaling diffusion LLMs to extended contexts and offer practical guidance for practitioners seeking 128K-scale context via efficient post-training.

Huatian Gong, Jiuh-Biing Sheu, Zheng Wang et al.

Rapid post-disaster road damage assessment is critical for effective emergency response, yet traditional optimization methods suffer from excessive computational time and require domain knowledge for algorithm design, making them unsuitable for time-sensitive disaster scenarios. This study proposes an attention-based encoder-decoder model (AEDM) for real-time drone routing decision in post-disaster road damage assessment. The method employs deep reinforcement learning to determine high-quality drone assessment routes without requiring algorithmic design knowledge. A network transformation method is developed to convert link-based routing problems into equivalent node-based formulations, while a synthetic road network generation technique addresses the scarcity of large-scale training datasets. The model is trained using policy optimization with multiple optima (POMO) with multi-task learning capabilities to handle diverse parameter combinations. Experimental results demonstrate two key strengths of AEDM: it outperforms commercial solvers by 16--69\% in solution quality and achieves real-time inference (1--2 seconds) versus 100--2,000 seconds for traditional methods. The model exhibits strong generalization across varying problem scales, drone numbers, and time constraints, consistently outperforming baseline methods on unseen parameter distributions and real-world road networks. The proposed method effectively balances computational efficiency with solution quality, making it particularly suitable for time-critical disaster response applications where rapid decision-making is essential for saving lives.

Huatian Gong, Jiuh-Biing Sheu, Zheng Wang et al.

Post-disaster road assessment (PDRA) is essential for emergency response, enabling rapid evaluation of infrastructure conditions and efficient allocation of resources. Although drones provide a flexible and effective tool for PDRA, routing them in large-scale networks remains challenging. Traditional optimization methods scale poorly and demand domain expertise, while existing deep reinforcement learning (DRL) approaches adopt a single-task paradigm, requiring separate models for each problem variant and lacking adaptability to evolving operational needs. This study proposes a unified model (UM) for drone routing that simultaneously addresses eight PDRA variants. By training a single neural network across multiple problem configurations, UM captures shared structural knowledge while adapting to variant-specific constraints through a modern transformer encoder-decoder architecture. A lightweight adapter mechanism further enables efficient finetuning to unseen attributes without retraining, enhancing deployment flexibility in dynamic disaster scenarios. Extensive experiments demonstrate that the UM reduces training time and parameters by a factor of eight compared with training separate models, while consistently outperforming single-task DRL methods by 6--14\% and traditional optimization approaches by 24--82\% in terms of solution quality (total collected information value). The model achieves real-time solutions (1--10 seconds) across networks of up to 1,000 nodes, with robustness confirmed through sensitivity analyses. Moreover, finetuning experiments show that unseen attributes can be effectively incorporated with minimal cost while retaining high solution quality. The proposed UM advances neural combinatorial optimization for time-critical applications, offering a computationally efficient, high-quality, and adaptable solution for drone-based PDRA.

Fuyao Zhang, Xinyu Yan, Tiantong Wu et al.

Large Language Models (LLMs) increasingly leverage Federated Learning (FL) to utilize private, task-specific datasets for fine-tuning while preserving data privacy. However, while federated LLM frameworks effectively enable collaborative training without raw data sharing, they critically lack built-in mechanisms for regulatory compliance like GDPR's right to be forgotten. Integrating private data heightens concerns over data quality and long-term governance, yet existing distributed training frameworks offer no principled way to selectively remove specific client contributions post-training. Due to distributed data silos, stringent privacy constraints, and the intricacies of interdependent model aggregation, federated LLM unlearning is significantly more complex than centralized LLM unlearning. To address this gap, we introduce Oblivionis, a lightweight learning and unlearning framework that enables clients to selectively remove specific private data during federated LLM training, enhancing trustworthiness and regulatory compliance. By unifying FL and unlearning as a dual optimization objective, we incorporate 6 FL and 5 unlearning algorithms for comprehensive evaluation and comparative analysis, establishing a robust pipeline for federated LLM unlearning. Extensive experiments demonstrate that Oblivionis outperforms local training, achieving a robust balance between forgetting efficacy and model utility, with cross-algorithm comparisons providing clear directions for future LLM development.

Jiongchao Jin, Xiuju Fu, Xiaowei Gao et al.

Maritime transportation is the backbone of global trade, making ship inspection essential for ensuring maritime safety and environmental protection. Port State Control (PSC), conducted by national ports, enforces compliance with safety regulations, with ship detention being the most severe consequence, impacting both ship schedules and company reputations. Traditional machine learning methods for ship detention prediction are limited by the capacity of representation learning and thus suffer from low accuracy. Meanwhile, autoencoder-based deep learning approaches face challenges due to the severe data imbalance in learning historical PSC detention records. To address these limitations, we propose Maritime Ship Detention with Large Language Models (MSD-LLM), integrating a dual robust subspace recovery (DSR) layer-based autoencoder with a progressive learning pipeline to handle imbalanced data and extract meaningful PSC representations. Then, a large language model groups and ranks features to identify likely detention cases, enabling dynamic thresholding for flexible detention predictions. Extensive evaluations on 31,707 PSC inspection records from the Asia-Pacific region show that MSD-LLM outperforms state-of-the-art methods more than 12\% on Area Under the Curve (AUC) for Singapore ports. Additionally, it demonstrates robustness to real-world challenges, making it adaptable to diverse maritime risk assessment scenarios.

Shuxue Peng, Zihang He, Haotian Zhang et al.

In this paper, we present a visual localization pipeline, namely MegLoc, for robust and accurate 6-DoF pose estimation under varying scenarios, including indoor and outdoor scenes, different time across a day, different seasons across a year, and even across years. MegLoc achieves state-of-the-art results on a range of challenging datasets, including winning the Outdoor and Indoor Visual Localization Challenge of ICCV 2021 Workshop on Long-term Visual Localization under Changing Conditions, as well as the Re-localization Challenge for Autonomous Driving of ICCV 2021 Workshop on Map-based Localization for Autonomous Driving.

Xiaopeng Bi, Ran Yan, Zheng Chai et al.

This report describes Megvii-3D team's approach towards SimLocMatch Challenge @ CVPR 2021 Image Matching Workshop.

Xiaopeng Bi, Yu Chen, Xinyang Liu et al.

This report describes Megvii-3D team's approach towards CVPR 2021 Image Matching Workshop.