Ju Huang

Chuanming Tang, Xiao Wang, Ju Huang et al.

Combining the Color and Event cameras (also called Dynamic Vision Sensors, DVS) for robust object tracking is a newly emerging research topic in recent years. Existing color-event tracking framework usually contains multiple scattered modules which may lead to low efficiency and high computational complexity, including feature extraction, fusion, matching, interactive learning, etc. In this paper, we propose a single-stage backbone network for Color-Event Unified Tracking (CEUTrack), which achieves the above functions simultaneously. Given the event points and RGB frames, we first transform the points into voxels and crop the template and search regions for both modalities, respectively. Then, these regions are projected into tokens and parallelly fed into the unified Transformer backbone network. The output features will be fed into a tracking head for target object localization. Our proposed CEUTrack is simple, effective, and efficient, which achieves over 75 FPS and new SOTA performance. To better validate the effectiveness of our model and address the data deficiency of this task, we also propose a generic and large-scale benchmark dataset for color-event tracking, termed COESOT, which contains 90 categories and 1354 video sequences. Additionally, a new evaluation metric named BOC is proposed in our evaluation toolkit to evaluate the prominence with respect to the baseline methods. We hope the newly proposed method, dataset, and evaluation metric provide a better platform for color-event-based tracking. The dataset, toolkit, and source code will be released on: \url{https://github.com/Event-AHU/COESOT}.

Weixun Wang, XiaoXiao Xu, Wanhe An et al.

Agentic crafting requires LLMs to operate in real-world environments over multiple turns by taking actions, observing outcomes, and iteratively refining artifacts. Despite its importance, the open-source community lacks a principled, end-to-end ecosystem to streamline agent development. We introduce the Agentic Learning Ecosystem (ALE), a foundational infrastructure that optimizes the production pipeline for agentic model. ALE consists of three components: ROLL, a post-training framework for weight optimization; ROCK, a sandbox environment manager for trajectory generation; and iFlow CLI, an agent framework for efficient context engineering. We release ROME, an open-source agent grounded by ALE and trained on over one million trajectories. Our approach includes data composition protocols for synthesizing complex behaviors and a novel policy optimization algorithm, Interaction-Perceptive Agentic Policy Optimization (IPA), which assigns credit over semantic interaction chunks rather than individual tokens to improve long-horizon training stability. Empirically, we evaluate ROME within a structured setting and introduce Terminal Bench Pro, a benchmark with improved scale and contamination control. ROME demonstrates strong performance across benchmarks like SWE-bench Verified and Terminal Bench, proving the effectiveness of ALE.

Wei Gao, Yuheng Zhao, Tianyuan Wu et al.

Agentic Reinforcement Learning (RL) enables Large Language Models (LLMs) to perform autonomous decision-making and long-term planning. Unlike standard LLM post-training, agentic RL workloads are highly heterogeneous, combining compute-intensive prefill phases, bandwidth-bound decoding, and stateful, CPU-heavy environment simulations. We argue that efficient agentic RL training requires disaggregated infrastructure to leverage specialized, best-fit hardware. However, naive disaggregation introduces substantial synchronization overhead and resource underutilization due to the complex dependencies between stages. We present RollArc, a distributed system designed to maximize throughput for multi-task agentic RL on disaggregated infrastructure. RollArc is built on three core principles: (1) hardware-affinity workload mapping, which routes compute-bound and bandwidth-bound tasks to bestfit GPU devices, (2) fine-grained asynchrony, which manages execution at the trajectory level to mitigate resource bubbles, and (3) statefulness-aware computation, which offloads stateless components (e.g., reward models) to serverless infrastructure for elastic scaling. Our results demonstrate that RollArc effectively improves training throughput and achieves 1.35-2.05\(\times\) end-to-end training time reduction compared to monolithic and synchronous baselines. We also evaluate RollArc by training a hundreds-of-billions-parameter MoE model for Qoder product on an Alibaba cluster with more than 3,000 GPUs, further demonstrating RollArc scalability and robustness. The code is available at https://github.com/alibaba/ROLL.

Yunze Tong, Mushui Liu, Canyu Zhao et al.

Deploying GRPO on Flow Matching models has proven effective for text-to-image generation. However, existing paradigms typically propagate an outcome-based reward to all preceding denoising steps without distinguishing the local effect of each step. Moreover, current group-wise ranking mainly compares trajectories at matched timesteps and ignores within-trajectory dependencies, where certain early denoising actions can affect later states via delayed, implicit interactions. We propose TurningPoint-GRPO (TP-GRPO), a GRPO framework that alleviates step-wise reward sparsity and explicitly models long-term effects within the denoising trajectory. TP-GRPO makes two key innovations: (i) it replaces outcome-based rewards with step-level incremental rewards, providing a dense, step-aware learning signal that better isolates each denoising action's "pure" effect, and (ii) it identifies turning points-steps that flip the local reward trend and make subsequent reward evolution consistent with the overall trajectory trend-and assigns these actions an aggregated long-term reward to capture their delayed impact. Turning points are detected solely via sign changes in incremental rewards, making TP-GRPO efficient and hyperparameter-free. Extensive experiments also demonstrate that TP-GRPO exploits reward signals more effectively and consistently improves generation. Demo code is available at https://github.com/YunzeTong/TurningPoint-GRPO.

Xiao Wang, Shiao Wang, Yuhe Ding et al.

In the post-deep learning era, the Transformer architecture has demonstrated its powerful performance across pre-trained big models and various downstream tasks. However, the enormous computational demands of this architecture have deterred many researchers. To further reduce the complexity of attention models, numerous efforts have been made to design more efficient methods. Among them, the State Space Model (SSM), as a possible replacement for the self-attention based Transformer model, has drawn more and more attention in recent years. In this paper, we give the first comprehensive review of these works and also provide experimental comparisons and analysis to better demonstrate the features and advantages of SSM. Specifically, we first give a detailed description of principles to help the readers quickly capture the key ideas of SSM. After that, we dive into the reviews of existing SSMs and their various applications, including natural language processing, computer vision, graph, multi-modal and multi-media, point cloud/event stream, time series data, and other domains. In addition, we give statistical comparisons and analysis of these models and hope it helps the readers to understand the effectiveness of different structures on various tasks. Then, we propose possible research points in this direction to better promote the development of the theoretical model and application of SSM. More related works will be continuously updated on the following GitHub: https://github.com/Event-AHU/Mamba_State_Space_Model_Paper_List.

Dilxat Muhtar, Jiashun Liu, Wei Gao et al.

Reinforcement Learning (RL) has emerged as a powerful paradigm for training LLM-based agents, yet remains limited by low sample efficiency, stemming not only from sparse outcome feedback but also from the agent's inability to leverage prior experience across episodes. While augmenting agents with historical experience offers a promising remedy, existing approaches suffer from a critical weakness: the experience distilled from history is either stored statically or fail to coevolve with the improving actor, causing a progressive misalignment between the experience and the actor's evolving capability that diminishes its utility over the course of training. Inspired by complementary learning systems in neuroscience, we present Complementary RL to achieve seamless co-evolution of an experience extractor and a policy actor within the RL optimization loop. Specifically, the actor is optimized via sparse outcome-based rewards, while the experience extractor is optimized according to whether its distilled experiences demonstrably contribute to the actor's success, thereby evolving its experience management strategy in lockstep with the actor's growing capabilities. Empirically, Complementary RL outperforms outcome-based agentic RL baselines that do not learn from experience, achieving 10% performance improvement in single-task scenarios and exhibits robust scalability in multi-task settings. These results establish Complementary RL as a paradigm for efficient experience-driven agent learning.

Ju Huang, Shiao Wang, Shuai Wang et al.

RGB-Event based tracking is an emerging research topic, focusing on how to effectively integrate heterogeneous multi-modal data (synchronized exposure video frames and asynchronous pulse Event stream). Existing works typically employ Transformer based networks to handle these modalities and achieve decent accuracy through input-level or feature-level fusion on multiple datasets. However, these trackers require significant memory consumption and computational complexity due to the use of self-attention mechanism. This paper proposes a novel RGB-Event tracking framework, Mamba-FETrack, based on the State Space Model (SSM) to achieve high-performance tracking while effectively reducing computational costs and realizing more efficient tracking. Specifically, we adopt two modality-specific Mamba backbone networks to extract the features of RGB frames and Event streams. Then, we also propose to boost the interactive learning between the RGB and Event features using the Mamba network. The fused features will be fed into the tracking head for target object localization. Extensive experiments on FELT and FE108 datasets fully validated the efficiency and effectiveness of our proposed tracker. Specifically, our Mamba-based tracker achieves 43.5/55.6 on the SR/PR metric, while the ViT-S based tracker (OSTrack) obtains 40.0/50.9. The GPU memory cost of ours and ViT-S based tracker is 13.98GB and 15.44GB, which decreased about $9.5\%$. The FLOPs and parameters of ours/ViT-S based OSTrack are 59GB/1076GB and 7MB/60MB, which decreased about $94.5\%$ and $88.3\%$, respectively. We hope this work can bring some new insights to the tracking field and greatly promote the application of the Mamba architecture in tracking. The source code of this work will be released on \url{https://github.com/Event-AHU/Mamba_FETrack}.

Zihe Liu, Jiashun Liu, Yancheng He et al.

Reinforcement learning for LLM reasoning has rapidly emerged as a prominent research area, marked by a significant surge in related studies on both algorithmic innovations and practical applications. Despite this progress, several critical challenges remain, including the absence of standardized guidelines for employing RL techniques and a fragmented understanding of their underlying mechanisms. Additionally, inconsistent experimental settings, variations in training data, and differences in model initialization have led to conflicting conclusions, obscuring the key characteristics of these techniques and creating confusion among practitioners when selecting appropriate techniques. This paper systematically reviews widely adopted RL techniques through rigorous reproductions and isolated evaluations within a unified open-source framework. We analyze the internal mechanisms, applicable scenarios, and core principles of each technique through fine-grained experiments, including datasets of varying difficulty, model sizes, and architectures. Based on these insights, we present clear guidelines for selecting RL techniques tailored to specific setups, and provide a reliable roadmap for practitioners navigating the RL for the LLM domain. Finally, we reveal that a minimalist combination of two techniques can unlock the learning capability of critic-free policies using vanilla PPO loss. The results demonstrate that our simple combination consistently improves performance, surpassing strategies like GRPO and DAPO.

Jia-Qi Yang, Chenglei Dai, Dan OU et al.

With the advancement of multimedia internet, the impact of visual characteristics on the decision of users to click or not within the online retail industry is increasingly significant. Thus, incorporating visual features is a promising direction for further performance improvements in click-through rate (CTR). However, experiments on our production system revealed that simply injecting the image embeddings trained with established pre-training methods only has marginal improvements. We believe that the main advantage of existing image feature pre-training methods lies in their effectiveness for cross-modal predictions. However, this differs significantly from the task of CTR prediction in recommendation systems. In recommendation systems, other modalities of information (such as text) can be directly used as features in downstream models. Even if the performance of cross-modal prediction tasks is excellent, it is challenging to provide significant information gain for the downstream models. We argue that a visual feature pre-training method tailored for recommendation is necessary for further improvements beyond existing modality features. To this end, we propose an effective user intention reconstruction module to mine visual features related to user interests from behavior histories, which constructs a many-to-one correspondence. We further propose a contrastive training method to learn the user intentions and prevent the collapse of embedding vectors. We conduct extensive experimental evaluations on public datasets and our production system to verify that our method can learn users' visual interests. Our method achieves $0.46\%$ improvement in offline AUC and $0.88\%$ improvement in Taobao GMV (Cross Merchandise Volume) with p-value$<$0.01.

Xiao Wang, Xufeng Lou, Shiao Wang et al.

Existing event stream based trackers undergo evaluation on short-term tracking datasets, however, the tracking of real-world scenarios involves long-term tracking, and the performance of existing tracking algorithms in these scenarios remains unclear. In this paper, we first propose a new long-term, large-scale frame-event visual object tracking dataset, termed FELT. It contains 1,044 long-term videos that involve 1.9 million RGB frames and event stream pairs, 60 different target objects, and 14 challenging attributes. To build a solid benchmark, we retrain and evaluate 21 baseline trackers on our dataset for future work to compare. In addition, we propose a novel Associative Memory Transformer based RGB-Event long-term visual tracker, termed AMTTrack. It follows a one-stream tracking framework and aggregates the multi-scale RGB/event template and search tokens effectively via the Hopfield retrieval layer. The framework also embodies another aspect of associative memory by maintaining dynamic template representations through an associative memory update scheme, which addresses the appearance variation in long-term tracking. Extensive experiments on FELT, FE108, VisEvent, and COESOT datasets fully validated the effectiveness of our proposed tracker. Both the dataset and source code will be released on https://github.com/Event-AHU/FELT_SOT_Benchmark

AMAP AI Agent Team, Yulan Hu, Xiangwen Zhang et al.

We present STAgent, an agentic large language model tailored for spatio-temporal understanding, designed to solve complex tasks such as constrained point-of-interest discovery and itinerary planning. STAgent is a specialized model capable of interacting with ten distinct tools within spatio-temporal scenarios, enabling it to explore, verify, and refine intermediate steps during complex reasoning. Notably, STAgent effectively preserves its general capabilities. We empower STAgent with these capabilities through three key contributions: (1) a stable tool environment that supports over ten domain-specific tools, enabling asynchronous rollout and training; (2) a hierarchical data curation framework that identifies high-quality data like a needle in a haystack, curating high-quality queries by retaining less than 1\% of the raw data, emphasizing both diversity and difficulty; and (3) a cascaded training recipe that starts with a seed SFT stage acting as a guardian to measure query difficulty, followed by a second SFT stage fine-tuned on queries with high certainty, and an ultimate RL stage that leverages data of low certainty. Initialized with Qwen3-30B-A3B to establish a strong SFT foundation and leverage insights into sample difficulty, STAgent yields promising performance on TravelBench while maintaining its general capabilities across a wide range of general benchmarks, thereby demonstrating the effectiveness of our proposed agentic model.

Hendrik Kraß, Ju Huang, Seyed Mohamad Moosavi

Universal machine learning interatomic potentials (uMLIPs) have emerged as powerful tools for accelerating atomistic simulations, offering scalable and efficient modeling with accuracy close to quantum calculations. However, their reliability and effectiveness in practical, real-world applications remain an open question. Metal-organic frameworks (MOFs) and related nanoporous materials are highly porous crystals with critical relevance in carbon capture, energy storage, and catalysis applications. Modeling nanoporous materials presents distinct challenges for uMLIPs due to their diverse chemistry, structural complexity, including porosity and coordination bonds, and the absence from existing training datasets. Here, we introduce MOFSimBench, a benchmark to evaluate uMLIPs on key materials modeling tasks for nanoporous materials, including structural optimization, molecular dynamics (MD) stability, the prediction of bulk properties, such as bulk modulus and heat capacity, and guest-host interactions. Evaluating over 20 models from various architectures on a chemically and structurally diverse materials set, we find that top-performing uMLIPs consistently outperform classical force fields and fine-tuned machine learning potentials across all tasks, demonstrating their readiness for deployment in nanoporous materials modeling. Our analysis highlights that data quality, particularly the diversity of training sets and inclusion of out-of-equilibrium conformations, plays a more critical role than model architecture in determining performance across all evaluated uMLIPs. We release our modular and extendable benchmarking framework at https://github.com/AI4ChemS/mofsim-bench, providing an open resource to guide the adoption for nanoporous materials modeling and further development of uMLIPs.

Shiao Wang, Ju Huang, Qingchuan Ma et al.

Combining traditional RGB cameras with bio-inspired event cameras for robust object tracking has garnered increasing attention in recent years. However, most existing multimodal tracking algorithms depend heavily on high-complexity Vision Transformer architectures for feature extraction and fusion across modalities. This not only leads to substantial computational overhead but also limits the effectiveness of cross-modal interactions. In this paper, we propose an efficient RGB-Event object tracking framework based on the linear-complexity Vision Mamba network, termed Mamba-FETrack V2. Specifically, we first design a lightweight Prompt Generator that utilizes embedded features from each modality, together with a shared prompt pool, to dynamically generate modality-specific learnable prompt vectors. These prompts, along with the modality-specific embedded features, are then fed into a Vision Mamba-based FEMamba backbone, which facilitates prompt-guided feature extraction, cross-modal interaction, and fusion in a unified manner. Finally, the fused representations are passed to the tracking head for accurate target localization. Extensive experimental evaluations on multiple RGB-Event tracking benchmarks, including short-term COESOT dataset and long-term datasets, i.e., FE108 and FELT V2, demonstrate the superior performance and efficiency of the proposed tracking framework. The source code and pre-trained models will be released on https://github.com/Event-AHU/Mamba_FETrack

Wei Gao, Yuheng Zhao, Dilxat Muhtar et al.

Agentic reinforcement learning (RL) has emerged as a key driver for improving the multi-step reasoning and tool-use capabilities of LLMs. However, its efficiency is bottlenecked by long-tail rollouts with multi-turn environment interactions, making static GPU provisioning a poor fit: overprovisioning wastes GPUs on stragglers, while underprovisioning increases contention and slows training. We observe that production serving clusters routinely leave substantial GPU compute and memory headroom. Based on this observation, we argue for cooperative elasticity: opportunistically repurposing underutilized serving GPUs to execute rollouts. Realizing cooperative elasticity is non-trivial because it must preserve serving Service Level Objectives (SLOs) under bursty traffic and minimize communication overhead. To address these challenges, we present ROSE, a cooperative, resource-elastic post-training system that safely harvests idle compute and memory on serving GPUs to accelerate agentic RL rollouts. ROSE consists of three components: (1) an SLO-safe co-serving executor that improves rollout throughput while preserving serving SLOs through efficient GPU memory and compute sharing; (2) a cross-cluster weight transfer engine that leverages weight shards and sparsity for fast weight synchronization across clusters; and (3) an elastic rollout scheduler that dynamically provisions cooperative capacity and routes trajectory rollouts across dedicated rollout GPUs and opportunistic serving GPUs. Experiments across multiple model sizes and cluster scales show that ROSE improves average end-to-end throughput by 1.20-3.31 x compared with state-of-the-art resource-fixed and elastic baselines.

Wei Gao, Yuheng Zhao, Dakai An et al.

Reinforcement Learning (RL) is a pivotal post-training technique for enhancing the reasoning capabilities of Large Language Models (LLMs). However, synchronous RL post-training often suffers from significant GPU underutilization, referred to as bubbles, caused by imbalanced response lengths within rollout steps. Many RL systems attempt to alleviate this problem by relaxing synchronization, but this can compromise training accuracy. In this paper, we introduce tail batching, a novel rollout scheduling strategy for synchronous RL that systematically consolidates prompts leading to long-tail responses into a small subset of rollout steps (long rounds), while ensuring that the majority of steps (short rounds) involve only balanced, short rollouts. By excluding long responses from short rounds and rescheduling them into a few designated long rounds, tail batching effectively reduces GPU idle time during rollouts and significantly accelerates RL training without sacrificing accuracy. We present RollPacker, a system that fully harnesses the benefits of tail batching through holistic optimizations across all three RL stages: elastic parallelism adaptation for rollout, dynamic resource allocation and scheduling for reward, and stream-based training. Empirical results show that RollPacker achieves a 2.03x-2.56x end-to-end training time reduction compared to veRL and up to 2.24x speedup compared to RLHFuse for the Qwen2.5 family of LLMs on up to 128 H800 GPUs.

Han Lu, Zichen Liu, Shaopan Xiong et al.

Synchronous Reinforcement Learning (RL) post-training has emerged as a crucial step for enhancing Large Language Models (LLMs) with diverse capabilities. However, many systems designed to accelerate RL post-training still suffer from low resource utilization and limited scalability. We present ROLL Flash, a system that extends ROLL with native support for asynchronous RL post-training. ROLL Flash is built upon two core design principles: fine-grained parallelism and rollout-train decoupling. Guided by these principles, ROLL Flash provides flexible programming interfaces that enable a fully asynchronous training architecture and support efficient rollout mechanisms, including queue scheduling and environment-level asynchronous execution. Through comprehensive theoretical analysis and extensive experiments, we demonstrate that ROLL Flash significantly improves resource utilization and scalability over synchronous RL post-training. ROLL Flash achieves up to 2.24x speedup on RLVR tasks and 2.72x on agentic tasks, using the same GPU budget as synchronous baselines. Furthermore, we implement several popular off-policy algorithms and verify that asynchronous training can achieve performance on par with synchronous training.

Hua Zong, Qingtao Zeng, Zhengxiong Zhou et al.

In this paper, we propose RecIS, a unified Sparse-Dense training framework designed to achieve two primary goals: 1. Unified Framework To create a Unified sparse-dense training framework based on the PyTorch ecosystem that meets the training needs of industrial-grade recommendation models that integrated with large models. 2.System Optimization To optimize the sparse component, offering superior efficiency over the TensorFlow-based recommendation models. The dense component, meanwhile, leverages existing optimization technologies within the PyTorch ecosystem. Currently, RecIS is being used in Alibaba for numerous large-model enhanced recommendation training tasks, and some traditional sparse models have also begun training in it.

Chao Yi, Dian Chen, Gaoyang Guo et al.

Recommender systems are among the most impactful applications of artificial intelligence, serving as critical infrastructure connecting users, merchants, and platforms. However, most current industrial systems remain heavily reliant on historical co-occurrence patterns and log-fitting objectives, i.e., optimizing for past user interactions without explicitly modeling user intent. This log-fitting approach often leads to overfitting to narrow historical preferences, failing to capture users' evolving and latent interests. As a result, it reinforces filter bubbles and long-tail phenomena, ultimately harming user experience and threatening the sustainability of the whole recommendation ecosystem. To address these challenges, we rethink the overall design paradigm of recommender systems and propose RecGPT, a next-generation framework that places user intent at the center of the recommendation pipeline. By integrating large language models (LLMs) into key stages of user interest mining, item retrieval, and explanation generation, RecGPT transforms log-fitting recommendation into an intent-centric process. To effectively align general-purpose LLMs to the above domain-specific recommendation tasks at scale, RecGPT incorporates a multi-stage training paradigm, which integrates reasoning-enhanced pre-alignment and self-training evolution, guided by a Human-LLM cooperative judge system. Currently, RecGPT has been fully deployed on the Taobao App. Online experiments demonstrate that RecGPT achieves consistent performance gains across stakeholders: users benefit from increased content diversity and satisfaction, merchants and the platform gain greater exposure and conversions. These comprehensive improvement results across all stakeholders validates that LLM-driven, intent-centric design can foster a more sustainable and mutually beneficial recommendation ecosystem.