arXiv subjects

Guoqing Ma

Publications and source records attributed to Guoqing Ma.

30 records · Page 2Linked to original sources

arXiv2025

Clustering-Based Weight Orthogonalization for Stabilizing Deep Reinforcement Learning

Reinforcement learning (RL) has made significant advancements, achieving superhuman performance in various tasks. However, RL agents often operate under the assumption of environmental stationarity, which poses a great challenge to learning efficiency since many environments are inherently non-stationary. This non-stationarity results in the requirement of millions of iterations, leading to low sample efficiency. To address this issue, we introduce the Clustering Orthogonal Weight Modified (COWM) layer, which can be integrated into the policy network of any RL algorithm and mitigate non-stationarity effectively. The COWM layer stabilizes the learning process by employing clustering techniques and a projection matrix. Our approach not only improves learning speed but also reduces gradient interference, thereby enhancing the overall learning efficiency. Empirically, the COWM outperforms state-of-the-art methods and achieves improvements of 9% and 12.6% in vision based and state-based DMControl benchmark. It also shows robustness and generality across various algorithms and tasks.

cs.LG↗

arXiv2025

Generative Pre-trained Autoregressive Diffusion Transformer

In this work, we present GPDiT, a Generative Pre-trained Autoregressive Diffusion Transformer that unifies the strengths of diffusion and autoregressive modeling for long-range video synthesis, within a continuous latent space. Instead of predicting discrete tokens, GPDiT autoregressively predicts future latent frames using a diffusion loss, enabling natural modeling of motion dynamics and semantic consistency across frames. This continuous autoregressive framework not only enhances generation quality but also endows the model with representation capabilities. Additionally, we introduce a lightweight causal attention variant and a parameter-free rotation-based time-conditioning mechanism, improving both the training and inference efficiency. Extensive experiments demonstrate that GPDiT achieves strong performance in video generation quality, video representation ability, and few-shot learning tasks, highlighting its potential as an effective framework for video modeling in continuous space.

cs.CV↗

arXiv2025

Automatic Failure Attribution and Critical Step Prediction Method for Multi-Agent Systems Based on Causal Inference

Multi-agent systems (MAS) are critical for automating complex tasks, yet their practical deployment is severely hampered by the challenge of failure attribution. Current diagnostic tools, which rely on statistical correlations, are fundamentally inadequate; on challenging benchmarks like Who\&When, state-of-the-art methods achieve less than 15\% accuracy in locating the root-cause step of a failure. To address this critical gap, we introduce the first failure attribution framework for MAS grounded in multi-granularity causal inference. Our approach makes two key technical contributions: (1) a performance causal inversion principle, which correctly models performance dependencies by reversing the data flow in execution logs, combined with Shapley values to accurately assign agent-level blame; (2) a novel causal discovery algorithm, CDC-MAS, that robustly identifies critical failure steps by tackling the non-stationary nature of MAS interaction data. The framework's attribution results directly fuel an automated optimization loop, generating targeted suggestions whose efficacy is validated via counterfactual simulations. Evaluations on the Who\&When and TRAIL benchmarks demonstrate a significant leap in performance. Our method achieves up to 36.2\% step-level accuracy. Crucially, the generated optimizations boost overall task success rates by an average of 22.4\%. This work provides a principled and effective solution for debugging complex agent interactions, paving the way for more reliable and interpretable multi-agent systems.

cs.AI↗

arXiv2025

Frame-Level Captions for Long Video Generation with Complex Multi Scenes

Generating long videos that can show complex stories, like movie scenes from scripts, has great promise and offers much more than short clips. However, current methods that use autoregression with diffusion models often struggle because their step-by-step process naturally leads to a serious error accumulation (drift). Also, many existing ways to make long videos focus on single, continuous scenes, making them less useful for stories with many events and changes. This paper introduces a new approach to solve these problems. First, we propose a novel way to annotate datasets at the frame-level, providing detailed text guidance needed for making complex, multi-scene long videos. This detailed guidance works with a Frame-Level Attention Mechanism to make sure text and video match precisely. A key feature is that each part (frame) within these windows can be guided by its own distinct text prompt. Our training uses Diffusion Forcing to provide the model with the ability to handle time flexibly. We tested our approach on difficult VBench 2.0 benchmarks ("Complex Plots" and "Complex Landscapes") based on the WanX2.1-T2V-1.3B model. The results show our method is better at following instructions in complex, changing scenes and creates high-quality long videos. We plan to share our dataset annotation methods and trained models with the research community. Project page: https://zgctroy.github.io/frame-level-captions .

cs.CV↗

arXiv2025

STORYANCHORS: Generating Consistent Multi-Scene Story Frames for Long-Form Narratives

This paper introduces StoryAnchors, a unified framework for generating high-quality, multi-scene story frames with strong temporal consistency. The framework employs a bidirectional story generator that integrates both past and future contexts to ensure temporal consistency, character continuity, and smooth scene transitions throughout the narrative. Specific conditions are introduced to distinguish story frame generation from standard video synthesis, facilitating greater scene diversity and enhancing narrative richness. To further improve generation quality, StoryAnchors integrates Multi-Event Story Frame Labeling and Progressive Story Frame Training, enabling the model to capture both overarching narrative flow and event-level dynamics. This approach supports the creation of editable and expandable story frames, allowing for manual modifications and the generation of longer, more complex sequences. Extensive experiments show that StoryAnchors outperforms existing open-source models in key areas such as consistency, narrative coherence, and scene diversity. Its performance in narrative consistency and story richness is also on par with GPT-4o. Ultimately, StoryAnchors pushes the boundaries of story-driven frame generation, offering a scalable, flexible, and highly editable foundation for future research.

cs.CV↗

arXiv2025

Step-Video-TI2V Technical Report: A State-of-the-Art Text-Driven Image-to-Video Generation Model

We present Step-Video-TI2V, a state-of-the-art text-driven image-to-video generation model with 30B parameters, capable of generating videos up to 102 frames based on both text and image inputs. We build Step-Video-TI2V-Eval as a new benchmark for the text-driven image-to-video task and compare Step-Video-TI2V with open-source and commercial TI2V engines using this dataset. Experimental results demonstrate the state-of-the-art performance of Step-Video-TI2V in the image-to-video generation task. Both Step-Video-TI2V and Step-Video-TI2V-Eval are available at https://github.com/stepfun-ai/Step-Video-TI2V.

cs.CV↗

arXiv2025

Step-Video-T2V Technical Report: The Practice, Challenges, and Future of Video Foundation Model

We present Step-Video-T2V, a state-of-the-art text-to-video pre-trained model with 30B parameters and the ability to generate videos up to 204 frames in length. A deep compression Variational Autoencoder, Video-VAE, is designed for video generation tasks, achieving 16x16 spatial and 8x temporal compression ratios, while maintaining exceptional video reconstruction quality. User prompts are encoded using two bilingual text encoders to handle both English and Chinese. A DiT with 3D full attention is trained using Flow Matching and is employed to denoise input noise into latent frames. A video-based DPO approach, Video-DPO, is applied to reduce artifacts and improve the visual quality of the generated videos. We also detail our training strategies and share key observations and insights. Step-Video-T2V's performance is evaluated on a novel video generation benchmark, Step-Video-T2V-Eval, demonstrating its state-of-the-art text-to-video quality when compared with both open-source and commercial engines. Additionally, we discuss the limitations of current diffusion-based model paradigm and outline future directions for video foundation models. We make both Step-Video-T2V and Step-Video-T2V-Eval available at https://github.com/stepfun-ai/Step-Video-T2V. The online version can be accessed from https://yuewen.cn/videos as well. Our goal is to accelerate the innovation of video foundation models and empower video content creators.

cs.CV↗

arXiv2025

Efficient Reinforcement Learning Through Adaptively Pretrained Visual Encoder

While Reinforcement Learning (RL) agents can successfully learn to handle complex tasks, effectively generalizing acquired skills to unfamiliar settings remains a challenge. One of the reasons behind this is the visual encoders used are task-dependent, preventing effective feature extraction in different settings. To address this issue, recent studies have tried to pretrain encoders with diverse visual inputs in order to improve their performance. However, they rely on existing pretrained encoders without further exploring the impact of pretraining period. In this work, we propose APE: efficient reinforcement learning through Adaptively Pretrained visual Encoder -- a framework that utilizes adaptive augmentation strategy during the pretraining phase and extracts generalizable features with only a few interactions within the task environments in the policy learning period. Experiments are conducted across various domains, including DeepMind Control Suite, Atari Games and Memory Maze benchmarks, to verify the effectiveness of our method. Results show that mainstream RL methods, such as DreamerV3 and DrQ-v2, achieve state-of-the-art performance when equipped with APE. In addition, APE significantly improves the sampling efficiency using only visual inputs during learning, approaching the efficiency of state-based method in several control tasks. These findings demonstrate the potential of adaptive pretraining of encoder in enhancing the generalization ability and efficiency of visual RL algorithms.

cs.CV↗

arXiv2022

Paving the Way for Distributed Artificial Intelligence over the Air

Distributed Artificial Intelligence (DAI) is regarded as one of the most promising techniques to provide intelligent services under strict privacy protection regulations for multiple clients. By applying DAI, training on raw data is carried out locally, while the trained outputs, e.g., model parameters, from multiple local clients, are sent back to a central server for aggregation. Recently, for achieving better practicality, DAI is studied in conjunction with wireless communication networks, incorporating various random effects brought by wireless channels. However, because of the complex and case-dependent nature of wireless channels, a generic simulator for applying DAI in wireless communication networks is still lacking. To accelerate the development of DAI applied in wireless communication networks, we propose a generic system design in this paper as well as an associated simulator that can be set according to wireless channels and system-level configurations. Details of the system design and analysis of the impacts of wireless environments are provided to facilitate further implementations and updates. We employ a series of experiments to verify the effectiveness and efficiency of the proposed system design and reveal its superior scalability.

cs.DC↗

arXiv2021

Smart Buildings Enabled by 6G Communications

Smart building (SB), a promising solution to fast-paced and continuous urbanization around the world, includes the integration of a wide range of systems and services and involves the construction of multiple layers. SB is capable of sensing, acquiring, and processing a very large amount of data as well as performing appropriate actions and adaptation. Rapid increases in the number of connected nodes and thereby the data transmission demand of SB have led to conventional transmission and processing techniques becoming insufficient to provide satisfactory services. In order to enhance the intelligence of SBs and achieve efficient monitoring and control, sixth generation (6G) communication technologies, particularly indoor visible light communications (VLC) and machine learning (ML), are required to be incorporated in SBs. Herein, we envision a novel SB framework featuring a reliable data transmission network, powerful data processing, and reasoning abilities, all of which are enabled by 6G communications. Primary simulation results support the promising visions of the proposed SB framework.

cs.NI↗

arXiv2019

Enhanced Orthogonal Frequency-Division Multiplexing with Subcarrier Number Modulation

A novel modulation scheme termed orthogonal frequency-division multiplexing with subcarrier number modulation (OFDM-SNM) has been proposed and regarded as one of the promising candidate modulation schemes for next generation networks. Although OFDM-SNM is capable of having a higher spectral efficiency (SE) than OFDM with index modulation (OFDM-IM) and plain OFDM under certain conditions, its reliability is relatively inferior to these existing schemes, because the number of active subcarriers varies. In this regard, we propose an enhanced OFDM-SNM scheme in this paper, which utilizes the flexibility of placing subcarriers to harvest a coding gain in the high signal-to-noise ratio (SNR) region. In particular, we stipulate a methodology that optimizes the subcarrier activation pattern (SAP) by subcarrier assignment using instantaneous channel state information (CSI) and therefore the subcarriers with higher channel power gains will be granted the priority to be activated, given the number of subcarriers is fixed. We also analyze the proposed enhanced OFDM-SNM system in terms of outage and error performance. The average outage probability and block error rate (BLER) are derived and approximated in closed-form expressions, which are further verified by numerical results generated by Monte Carlo simulations. The high-reliability nature of the enhanced OFDM-SNM makes it a promising candidate for implementing in the Internet of Things (IoT) with stationary machine-type devices (MTDs), which are subject to slow fading and supported by proper power supply.

eess.SP↗

arXiv2019

A Novel Error Performance Analysis Methodology for OFDM-IM

Orthogonal frequency-division multiplexing with index modulation (OFDM-IM) has become a high-profile candidate for the modulation scheme in next generation networks, and many works have been published to analyze it in recent years. Error performance is one of the most important and interesting aspects of OFDM-IM, which has been analyzed in most works. However, most of them employ two key approximations to derive the closed-form expressions of block error rate (BLER) and/or average bit error rate (BER). The first one is to utilize the union bound assisted by the pairwise error probability (PEP) analysis, and the second one is to employ an exponential approximation of Q-function. In this letter, we apply Craig's formula to analyze the error performance for OFDM-IM in place of the exponential approximation. Because Craig's formula is the exact expression of Q-function, the accuracy of analytical results regarding error performance for OFDM-IM can be improved. We examine the proposed error performance analysis methodology based on Craig's formula by investigating both average BLER and BER.

cs.IT↗