Yicheng Xiao

Lin Song, Wenbo Li, Guoqing Ma et al.

We present JoyAI-Image, a unified multimodal foundation model for visual understanding, text-to-image generation, and instruction-guided image editing. JoyAI-Image couples a spatially enhanced Multimodal Large Language Model (MLLM) with a Multimodal Diffusion Transformer (MMDiT), allowing perception and generation to interact through a shared multimodal interface. Around this architecture, we build a scalable training recipe that combines unified instruction tuning, long-text rendering supervision, spatially grounded data, and both general and spatial editing signals. This design gives the model broad multimodal capability while strengthening geometry-aware reasoning and controllable visual synthesis. Experiments across understanding, generation, long-text rendering, and editing benchmarks show that JoyAI-Image achieves state-of-the-art or highly competitive performance. More importantly, the bidirectional loop between enhanced understanding, controllable spatial editing, and novel-view-assisted reasoning enables the model to move beyond general visual competence toward stronger spatial intelligence. These results suggest a promising path for unified visual models in downstream applications such as vision-language-action systems and world models.

Yicheng Xiao, Yue Ma, Shuyan Li et al.

In this paper, we propose SemanticAC, a semantics-assisted framework for Audio Classification to better leverage the semantic information. Unlike conventional audio classification methods that treat class labels as discrete vectors, we employ a language model to extract abundant semantics from labels and optimize the semantic consistency between audio signals and their labels. We verify that simple textual information from labels and advanced pretraining models enable more abundant semantic supervision for better performance. Specifically, we design a text encoder to capture the semantic information from the text extension of labels. Then we map the audio signals to align with the semantics of corresponding class labels via an audio encoder and a similarity calculation module so as to enforce the semantic consistency. Extensive experiments on two audio datasets, ESC-50 and US8K demonstrate that our proposed method consistently outperforms the compared audio classification methods.

Yicheng Xiao, Yu Chen, Haoxuan Ma et al.

While the Contrastive Language-Image Pretraining(CLIP) model has achieved remarkable success in a variety of downstream vison language understanding tasks, enhancing its capability for fine-grained image-text alignment remains an active research focus. To this end, most existing works adopt the strategy of explicitly increasing the granularity of visual information processing, e.g., incorporating visual prompts to guide the model focus on specific local regions within the image. Meanwhile, researches on Multimodal Large Language Models(MLLMs) have demonstrated that training with long and detailed textual descriptions can effectively improve the model's fine-grained vision-language alignment. However, the inherent token length limitation of CLIP's text encoder fundamentally limits CLIP to process more granular textual information embedded in long text sequences. To synergistically leverage the advantages of enhancing both visual and textual content processing granularity, we propose PixCLIP, a novel framework designed to concurrently accommodate visual prompt inputs and process lengthy textual descriptions. Specifically, we first establish an automated annotation pipeline capable of generating pixel-level localized, long-form textual descriptions for images. Utilizing this pipeline, we construct LongGRIT, a high-quality dataset comprising nearly 1.5 million samples. Secondly, we replace CLIP's original text encoder with the LLM and propose a three-branch pixel-text alignment learning framework, facilitating fine-grained alignment between image regions and corresponding textual descriptions at arbitrary granularity. Experiments demonstrate that PixCLIP showcases breakthroughs in pixel-level interaction and handling long-form texts, achieving state-of-the-art performance.

Yicheng Xiao, Lin Song, Yukang Chen et al.

Recent text-to-image systems face limitations in handling multimodal inputs and complex reasoning tasks. We introduce MindOmni, a unified multimodal large language model that addresses these challenges by incorporating reasoning generation through reinforcement learning. MindOmni leverages a three-phase training strategy: i) design of a unified vision language model with a decoder-only diffusion module, ii) supervised fine-tuning with Chain-of-Thought (CoT) instruction data, and iii) our proposed Reasoning Generation Policy Optimization (RGPO) algorithm, utilizing multimodal feedback to effectively guide policy updates. Experimental results demonstrate that MindOmni outperforms existing models, achieving impressive performance on both understanding and generation benchmarks, meanwhile showcasing advanced fine-grained reasoning generation capabilities, especially with mathematical reasoning instruction. All codes will be made public at https://github.com/TencentARC/MindOmni

Yicheng Xiao, Junpeng Zhang

Deep convolutional neural network with increased number of parameters has achieved improved precision in task of object detection on natural images, where objects of interests are annotated with horizontal boundary boxes. On aerial images captured from the bird-view perspective, these improvements on model architecture and deeper convolutional layers can also boost the performance on oriented object detection task. However, it is hard to directly apply those state-of-the-art object detectors on the devices with limited computation resources, which necessitates lightweight models through model compression. In order to address this issue, we present a model compression method for rotated object detection on aerial images by knowledge distillation, namely KD-RNet. With a well-trained teacher oriented object detector with a large number of parameters, the obtained object category and location information are both transferred to a compact student network in KD-RNet by collaborative training strategy. Transferring the category information is achieved by knowledge distillation on predicted probability distribution, and a soft regression loss is adopted for handling displacement in location information transfer. The experimental result on a large-scale aerial object detection dataset (DOTA) demonstrates that the proposed KD-RNet model can achieve improved mean-average precision (mAP) with reduced number of parameters, at the same time, KD-RNet boost the performance on providing high quality detections with higher overlap with groundtruth annotations.

Yukang Chen, Luozhou Wang, Wei Huang et al.

We present LongLive-2.0, an NVFP4-based parallel infrastructure throughout the full training and inference workflow of long video generation, addressing speed and memory bottlenecks. For training, we introduce sequence-parallel autoregressive (AR) training, instantiated as Balanced SP, which co-designs the efficient teacher-forcing layout with SP execution by pairing clean-history and noisy-target temporal chunks on each rank, enabling a natural teacher-forcing mask with SP-aware chunked VAE encoding. Combined with NVFP4 precision, it reduces GPU memory cost and accelerates GEMM computation during training, the proportion of which increases as video length grows. Moreover, we show that a high-quality infrastructure and dataset enable a remarkably clean training pipeline. Unlike existing Self-Forcing series methods that rely on ODE initialization and subsequent distribution matching distillation (DMD), LongLive-2.0 directly tunes a diffusion model into a long, multi-shot, interactive auto-regressive (AR) diffusion model. It can be further converted to real-time generation (4 to 2 denoising steps) with standalone LoRA weights. For inference on Blackwell GPUs, we enable W4A4 NVFP4 inference, quantize KV cache into NVFP4 for memory savings, and boost end-to-end throughput with asynchronous streaming VAE decoding. On non-Blackwell GPU architectures, we deploy SP inference to match the speed on Blackwell GPUs, while the quantized KV cache can lower inter-GPU communication of SP. Experiments show up to 2.15x speedup in training, and 1.84x in inference. LongLive-2.0-5B achieves 45.7 FPS inference while attaining strong performance on benchmarks. To our knowledge, LongLive-2.0 is the first NVFP4 training and inference system for long video generation.

Yicheng Xiao, Lin Song, Rui Yang et al. · tsinghua

With the advancement of language models, unified multimodal understanding and generation have made significant strides, with model architectures evolving from separated components to unified single-model frameworks. This paper explores an efficient training paradigm to build a single transformer for unified multimodal understanding and generation. Specifically, we propose a multimodal warmup strategy utilizing prior knowledge to extend capabilities. To address cross-modal compatibility challenges, we introduce feature pre-scaling and multimodal AdaLN techniques. Integrating the proposed technologies, we present the HaploOmni, a new single multimodal transformer. With limited training costs, HaploOmni achieves competitive performance across multiple image and video understanding and generation benchmarks over advanced unified models. All codes will be made public at https://github.com/Tencent/HaploVLM.

Yicheng Xiao, Wenhu Zhang, Lin Song et al.

Image spatial editing performs geometry-driven transformations, allowing precise control over object layout and camera viewpoints. Current models are insufficient for fine-grained spatial manipulations, motivating a dedicated assessment suite. Our contributions are listed: (i) We introduce SpatialEdit-Bench, a complete benchmark that evaluates spatial editing by jointly measuring perceptual plausibility and geometric fidelity via viewpoint reconstruction and framing analysis. (ii) To address the data bottleneck for scalable training, we construct SpatialEdit-500k, a synthetic dataset generated with a controllable Blender pipeline that renders objects across diverse backgrounds and systematic camera trajectories, providing precise ground-truth transformations for both object- and camera-centric operations. (iii) Building on this data, we develop SpatialEdit-16B, a baseline model for fine-grained spatial editing. Our method achieves competitive performance on general editing while substantially outperforming prior methods on spatial manipulation tasks. All resources will be made public at https://github.com/EasonXiao-888/SpatialEdit.

Zhuoyan Luo, Yinghao Wu, Tianheng Cheng et al.

The newly proposed Generalized Referring Expression Segmentation (GRES) amplifies the formulation of classic RES by involving complex multiple/non-target scenarios. Recent approaches address GRES by directly extending the well-adopted RES frameworks with object-existence identification. However, these approaches tend to encode multi-granularity object information into a single representation, which makes it difficult to precisely represent comprehensive objects of different granularity. Moreover, the simple binary object-existence identification across all referent scenarios fails to specify their inherent differences, incurring ambiguity in object understanding. To tackle the above issues, we propose a \textbf{Co}unting-Aware \textbf{H}ierarchical \textbf{D}ecoding framework (CoHD) for GRES. By decoupling the intricate referring semantics into different granularity with a visual-linguistic hierarchy, and dynamic aggregating it with intra- and inter-selection, CoHD boosts multi-granularity comprehension with the reciprocal benefit of the hierarchical nature. Furthermore, we incorporate the counting ability by embodying multiple/single/non-target scenarios into count- and category-level supervision, facilitating comprehensive object perception. Experimental results on gRefCOCO, Ref-ZOM, R-RefCOCO, and RefCOCO benchmarks demonstrate the effectiveness and rationality of CoHD which outperforms state-of-the-art GRES methods by a remarkable margin. Code is available at \href{https://github.com/RobertLuo1/CoHD}{here}.

Zhuoyan Luo, Yicheng Xiao, Yong Liu et al.

The recent transformer-based models have dominated the Referring Video Object Segmentation (RVOS) task due to the superior performance. Most prior works adopt unified DETR framework to generate segmentation masks in query-to-instance manner. In this work, we integrate strengths of that leading RVOS models to build up an effective paradigm. We first obtain binary mask sequences from the RVOS models. To improve the consistency and quality of masks, we propose Two-Stage Multi-Model Fusion strategy. Each stage rationally ensembles RVOS models based on framework design as well as training strategy, and leverages different video object segmentation (VOS) models to enhance mask coherence by object propagation mechanism. Our method achieves 75.7% J&F on Ref-Youtube-VOS validation set and 70% J&F on test set, which ranks 1st place on 5th Large-scale Video Object Segmentation Challenge (ICCV 2023) track 3. Code is available at https://github.com/RobertLuo1/iccv2023_RVOS_Challenge.

Jiajun Luo, Yicheng Xiao, Jianru Xu et al.

Diffusion models produce realistic images and videos but require substantial computational resources, necessitating multi-accelerator parallelism for real-time deployment. However, parallel inference introduces significant communication overhead from exchanging large activations between devices, limiting efficiency and scalability. We present CompactFusion, a compression framework that significantly reduces communication while preserving generation quality. Our key observation is that diffusion activations exhibit strong temporal redundancy-adjacent steps produce highly similar activations, saturating bandwidth with near-duplicate data carrying little new information. To address this inefficiency, we seek a more compact representation that encodes only the essential information. CompactFusion achieves this via Residual Compression that transmits only compressed residuals (step-wise activation differences). Based on empirical analysis and theoretical justification, we show that it effectively removes redundant data, enabling substantial data reduction while maintaining high fidelity. We also integrate lightweight error feedback to prevent error accumulation. CompactFusion establishes a new paradigm for parallel diffusion inference, delivering lower latency and significantly higher generation quality than prior methods. On 4xL20, it achieves 3.0x speedup while greatly improving fidelity. It also uniquely supports communication-heavy strategies like sequence parallelism on slow networks, achieving 6.7x speedup over prior overlap-based method. CompactFusion applies broadly across diffusion models and parallel settings, and integrates easily without requiring pipeline rework. Portable implementation demonstrated on xDiT is publicly available at https://github.com/Cobalt-27/CompactFusion

Jiangshan Wang, Yue Ma, Jiayi Guo et al.

Video editing is an emerging task, in which most current methods adopt the pre-trained text-to-image (T2I) diffusion model to edit the source video in a zero-shot manner. Despite extensive efforts, maintaining the temporal consistency of edited videos remains challenging due to the lack of temporal constraints in the regular T2I diffusion model. To address this issue, we propose COrrespondence-guided Video Editing (COVE), leveraging the inherent diffusion feature correspondence to achieve high-quality and consistent video editing. Specifically, we propose an efficient sliding-window-based strategy to calculate the similarity among tokens in the diffusion features of source videos, identifying the tokens with high correspondence across frames. During the inversion and denoising process, we sample the tokens in noisy latent based on the correspondence and then perform self-attention within them. To save GPU memory usage and accelerate the editing process, we further introduce the temporal-dimensional token merging strategy, which can effectively reduce redundancy. COVE can be seamlessly integrated into the pre-trained T2I diffusion model without the need for extra training or optimization. Extensive experiment results demonstrate that COVE achieves the start-of-the-art performance in various video editing scenarios, outperforming existing methods both quantitatively and qualitatively. The code will be release at https://github.com/wangjiangshan0725/COVE.

Rui Yang, Lin Song, Yicheng Xiao et al.

Recent advancements in large language models (LLMs) have significantly propelled the development of large multi-modal models (LMMs), highlighting the potential for general and intelligent assistants. However, most LMMs model visual and textual modalities separately, leading to recent efforts to develop native LMMs using a single transformer. Despite the promise, these native models are resource-intensive and often exhibit performance gaps compared to their compositional counterparts. To alleviate this issue, we propose a simple yet efficient method to construct a baseline for the native and end-to-end large multi-modal model in a single transformer. First, we propose a new early-fusion LMM that can fuse multi-modal inputs in the early stage and respond to visual instructions in an auto-regressive manner. Second, we devise an efficient training recipe for the proposed model, which harnesses the prior knowledge of the pre-trained models, addressing both the performance limitations and the challenge of resource consumption. The proposed model demonstrates superior performance compared to other LMMs using one transformer and significantly narrows the performance gap with compositional LMMs.

Shuai Yang, Wei Huang, Ruihang Chu et al.

We present LongLive, a frame-level autoregressive (AR) framework for real-time and interactive long video generation. Long video generation presents challenges in both efficiency and quality. Diffusion and Diffusion-Forcing models can produce high-quality videos but suffer from low efficiency due to bidirectional attention. Causal attention AR models support KV caching for faster inference, but often degrade in quality on long videos due to memory challenges during long-video training. In addition, beyond static prompt-based generation, interactive capabilities, such as streaming prompt inputs, are critical for dynamic content creation, enabling users to guide narratives in real time. This interactive requirement significantly increases complexity, especially in ensuring visual consistency and semantic coherence during prompt transitions. To address these challenges, LongLive adopts a causal, frame-level AR design that integrates a KV-recache mechanism that refreshes cached states with new prompts for smooth, adherent switches; streaming long tuning to enable long video training and to align training and inference (train-long-test-long); and short window attention paired with a frame-level attention sink, shorten as frame sink, preserving long-range consistency while enabling faster generation. With these key designs, LongLive fine-tunes a 1.3B-parameter short-clip model to minute-long generation in just 32 GPU-days. At inference, LongLive sustains 20.7 FPS on a single NVIDIA H100, achieves strong performance on VBench in both short and long videos. LongLive supports up to 240-second videos on a single H100 GPU. LongLive further supports INT8-quantized inference with only marginal quality loss.

Jiaqi Huang, Zunnan Xu, Jun Zhou et al. · tsinghua

Leveraging multimodal large models for image segmentation has become a prominent research direction. However, existing approaches typically rely heavily on manually annotated datasets that include explicit reasoning processes, which are costly and time-consuming to produce. Recent advances suggest that reinforcement learning (RL) can endow large models with reasoning capabilities without requiring such reasoning-annotated data. In this paper, we propose SAM-R1, a novel framework that enables multimodal large models to perform fine-grained reasoning in image understanding tasks. Our approach is the first to incorporate fine-grained segmentation settings during the training of multimodal reasoning models. By integrating task-specific, fine-grained rewards with a tailored optimization objective, we further enhance the model's reasoning and segmentation alignment. We also leverage the Segment Anything Model (SAM) as a strong and flexible reward provider to guide the learning process. With only 3k training samples, SAM-R1 achieves strong performance across multiple benchmarks, demonstrating the effectiveness of reinforcement learning in equipping multimodal models with segmentation-oriented reasoning capabilities.

Haoxiao Wang, Kaichen Zhou, Binrui Gu et al.

Manipulating transparent objects presents significant challenges due to the complexities introduced by their reflection and refraction properties, which considerably hinder the accurate estimation of their 3D shapes. To address these challenges, we propose a single-view RGB-D-based depth completion framework, TransDiff, that leverages the Denoising Diffusion Probabilistic Models(DDPM) to achieve material-agnostic object grasping in desktop. Specifically, we leverage features extracted from RGB images, including semantic segmentation, edge maps, and normal maps, to condition the depth map generation process. Our method learns an iterative denoising process that transforms a random depth distribution into a depth map, guided by initially refined depth information, ensuring more accurate depth estimation in scenarios involving transparent objects. Additionally, we propose a novel training method to better align the noisy depth and RGB image features, which are used as conditions to refine depth estimation step by step. Finally, we utilized an improved inference process to accelerate the denoising procedure. Through comprehensive experimental validation, we demonstrate that our method significantly outperforms the baselines in both synthetic and real-world benchmarks with acceptable inference time. The demo of our method can be found on https://wang-haoxiao.github.io/TransDiff/

Cheng Cheng, Lin Song, Yicheng Xiao et al.

Autoregressive (AR) image generators offer a language-model-friendly approach to image generation by predicting discrete image tokens in a causal sequence. However, unlike diffusion models, AR models lack a mechanism to refine previous predictions, limiting their generation quality. In this paper, we introduce TensorAR, a new AR paradigm that reformulates image generation from next-token prediction to next-tensor prediction. By generating overlapping windows of image patches (tensors) in a sliding fashion, TensorAR enables iterative refinement of previously generated content. To prevent information leakage during training, we propose a discrete tensor noising scheme, which perturbs input tokens via codebook-indexed noise. TensorAR is implemented as a plug-and-play module compatible with existing AR models. Extensive experiments on LlamaGEN, Open-MAGVIT2, and RAR demonstrate that TensorAR significantly improves the generation performance of autoregressive models.

Wei Huang, Yi Ge, Shuai Yang et al.

We propose QeRL, a Quantization-enhanced Reinforcement Learning framework for large language models (LLMs). While RL is essential for LLMs' reasoning capabilities, it is resource-intensive, requiring substantial GPU memory and long rollout durations. QeRL addresses these issues by combining NVFP4 quantization with Low-Rank Adaptation (LoRA), accelerating rollout phase of RL while reducing memory overhead. Beyond efficiency, our findings show that quantization noise increases policy entropy, enhancing exploration, and enabling the discovery of better strategies during RL. To further optimize exploration, QeRL introduces an Adaptive Quantization Noise (AQN) mechanism, which dynamically adjusts noise during training. Experiments demonstrate that QeRL delivers over 1.5 times speedup in the rollout phase. Moreover, this is the first framework to enable RL training of a 32B LLM on a single H100 80GB GPU, while delivering overall speedups for RL training. It also achieves faster reward growth and higher final accuracy than 16-bit LoRA and QLoRA, while matching the performance of full-parameter fine-tuning on mathematical benchmarks such as GSM8K (90.8%) and MATH 500 (77.4%) in the 7B model. These results establish QeRL as an efficient and effective framework for RL training in LLMs.

Yicheng Xiao, Lin Song, Rui Yang et al.

Recent advances have highlighted the benefits of scaling language models to enhance performance across a wide range of NLP tasks. However, these approaches still face limitations in effectiveness and efficiency when applied to domain-specific tasks, particularly for small edge-side models. We propose the LoRA-Gen framework, which utilizes a large cloud-side model to generate LoRA parameters for edge-side models based on task descriptions. By employing the reparameterization technique, we merge the LoRA parameters into the edge-side model to achieve flexible specialization. Our method facilitates knowledge transfer between models while significantly improving the inference efficiency of the specialized model by reducing the input context length. Without specialized training, LoRA-Gen outperforms conventional LoRA fine-tuning, which achieves competitive accuracy and a 2.1x speedup with TinyLLaMA-1.1B in reasoning tasks. Besides, our method delivers a compression ratio of 10.1x with Gemma-2B on intelligent agent tasks.

Yicheng Xiao, Yangyang Sun, Yicheng Lin

Current scientific research witnesses various attempts at applying Large Language Models for scenario generation but is inclined only to comprehensive or dangerous scenarios. In this paper, we seek to build a three-stage framework that not only lets users regain controllability over the generated scenarios but also generates comprehensive scenarios containing danger factors in uncontrolled intersection settings. In the first stage, LLM agents will contribute to translating the key components of the description of the expected scenarios into Functional Scenarios. For the second stage, we use Answer Set Programming (ASP) solver Clingo to help us generate comprehensive logical traffic within intersections. During the last stage, we use LLM to update relevant parameters to increase the critical level of the concrete scenario.

Yicheng Xiao, Lin Song, Shaoli Huang et al.

The state space models, employing recursively propagated features, demonstrate strong representation capabilities comparable to Transformer models and superior efficiency. However, constrained by the inherent geometric constraints of sequences, it still falls short in modeling long-range dependencies. To address this issue, we propose the GrootVL network, which first dynamically generates a tree topology based on spatial relationships and input features. Then, feature propagation is performed based on this graph, thereby breaking the original sequence constraints to achieve stronger representation capabilities. Additionally, we introduce a linear complexity dynamic programming algorithm to enhance long-range interactions without increasing computational cost. GrootVL is a versatile multimodal framework that can be applied to both visual and textual tasks. Extensive experiments demonstrate that our method significantly outperforms existing structured state space models on image classification, object detection and segmentation. Besides, by fine-tuning large language models, our approach achieves consistent improvements in multiple textual tasks at minor training cost.

Yicheng Xiao, Zhuoyan Luo, Yong Liu et al.

Video Moment Retrieval (MR) and Highlight Detection (HD) have attracted significant attention due to the growing demand for video analysis. Recent approaches treat MR and HD as similar video grounding problems and address them together with transformer-based architecture. However, we observe that the emphasis of MR and HD differs, with one necessitating the perception of local relationships and the other prioritizing the understanding of global contexts. Consequently, the lack of task-specific design will inevitably lead to limitations in associating the intrinsic specialty of two tasks. To tackle the issue, we propose a Unified Video COMprehension framework (UVCOM) to bridge the gap and jointly solve MR and HD effectively. By performing progressive integration on intra and inter-modality across multi-granularity, UVCOM achieves the comprehensive understanding in processing a video. Moreover, we present multi-aspect contrastive learning to consolidate the local relation modeling and global knowledge accumulation via well aligned multi-modal space. Extensive experiments on QVHighlights, Charades-STA, TACoS , YouTube Highlights and TVSum datasets demonstrate the effectiveness and rationality of UVCOM which outperforms the state-of-the-art methods by a remarkable margin.

Zhuoyan Luo, Yicheng Xiao, Yong Liu et al.

This paper studies referring video object segmentation (RVOS) by boosting video-level visual-linguistic alignment. Recent approaches model the RVOS task as a sequence prediction problem and perform multi-modal interaction as well as segmentation for each frame separately. However, the lack of a global view of video content leads to difficulties in effectively utilizing inter-frame relationships and understanding textual descriptions of object temporal variations. To address this issue, we propose Semantic-assisted Object Cluster (SOC), which aggregates video content and textual guidance for unified temporal modeling and cross-modal alignment. By associating a group of frame-level object embeddings with language tokens, SOC facilitates joint space learning across modalities and time steps. Moreover, we present multi-modal contrastive supervision to help construct well-aligned joint space at the video level. We conduct extensive experiments on popular RVOS benchmarks, and our method outperforms state-of-the-art competitors on all benchmarks by a remarkable margin. Besides, the emphasis on temporal coherence enhances the segmentation stability and adaptability of our method in processing text expressions with temporal variations. Code will be available.