Xiangxiang Chu

Chuyi Li, Lulu Li, Hongliang Jiang et al.

For years, the YOLO series has been the de facto industry-level standard for efficient object detection. The YOLO community has prospered overwhelmingly to enrich its use in a multitude of hardware platforms and abundant scenarios. In this technical report, we strive to push its limits to the next level, stepping forward with an unwavering mindset for industry application. Considering the diverse requirements for speed and accuracy in the real environment, we extensively examine the up-to-date object detection advancements either from industry or academia. Specifically, we heavily assimilate ideas from recent network design, training strategies, testing techniques, quantization, and optimization methods. On top of this, we integrate our thoughts and practice to build a suite of deployment-ready networks at various scales to accommodate diversified use cases. With the generous permission of YOLO authors, we name it YOLOv6. We also express our warm welcome to users and contributors for further enhancement. For a glimpse of performance, our YOLOv6-N hits 35.9% AP on the COCO dataset at a throughput of 1234 FPS on an NVIDIA Tesla T4 GPU. YOLOv6-S strikes 43.5% AP at 495 FPS, outperforming other mainstream detectors at the same scale~(YOLOv5-S, YOLOX-S, and PPYOLOE-S). Our quantized version of YOLOv6-S even brings a new state-of-the-art 43.3% AP at 869 FPS. Furthermore, YOLOv6-M/L also achieves better accuracy performance (i.e., 49.5%/52.3%) than other detectors with a similar inference speed. We carefully conducted experiments to validate the effectiveness of each component. Our code is made available at https://github.com/meituan/YOLOv6.

Chuyi Li, Lulu Li, Yifei Geng et al.

The YOLO community has been in high spirits since our first two releases! By the advent of Chinese New Year 2023, which sees the Year of the Rabbit, we refurnish YOLOv6 with numerous novel enhancements on the network architecture and the training scheme. This release is identified as YOLOv6 v3.0. For a glimpse of performance, our YOLOv6-N hits 37.5% AP on the COCO dataset at a throughput of 1187 FPS tested with an NVIDIA Tesla T4 GPU. YOLOv6-S strikes 45.0% AP at 484 FPS, outperforming other mainstream detectors at the same scale (YOLOv5-S, YOLOv8-S, YOLOX-S and PPYOLOE-S). Whereas, YOLOv6-M/L also achieve better accuracy performance (50.0%/52.8% respectively) than other detectors at a similar inference speed. Additionally, with an extended backbone and neck design, our YOLOv6-L6 achieves the state-of-the-art accuracy in real-time. Extensive experiments are carefully conducted to validate the effectiveness of each improving component. Our code is made available at https://github.com/meituan/YOLOv6.

Liang Li, Qingyuan Li, Bo Zhang et al.

As the size of large language models (LLMs) continues to grow, model compression without sacrificing accuracy has become a crucial challenge for deployment. While some quantization methods, such as GPTQ, have made progress in achieving acceptable 4-bit weight-only quantization, attempts at lower-bit quantization often result in severe performance degradation. In this paper, we introduce a technique called norm tweaking, which can be used as a plugin in current PTQ methods to achieve high precision while being cost-efficient. Our approach is inspired by the observation that rectifying the quantized activation distribution to match its float counterpart can readily restore accuracy for LLMs. To achieve this, we carefully design a tweaking strategy that includes calibration data generation and channel-wise distance constraint to update the weights of normalization layers for better generalization. We conduct extensive experiments on various datasets using several open-sourced LLMs. Our method demonstrates significant improvements in both weight-only quantization and joint quantization of weights and activations, surpassing existing PTQ methods. On GLM-130B and OPT-66B, our method even achieves the same level of accuracy at 2-bit quantization as their float ones. Our simple and effective approach makes it more practical for real-world applications.

Sifan Zhou, Zhi Tian, Xiangxiang Chu et al.

The deployment of 3D detectors strikes one of the major challenges in real-world self-driving scenarios. Existing BEV-based (i.e., Bird Eye View) detectors favor sparse convolutions (known as SPConv) to speed up training and inference, which puts a hard barrier for deployment, especially for on-device applications. In this paper, to tackle the challenge of efficient 3D object detection from an industry perspective, we devise a deployment-friendly pillar-based 3D detector, termed FastPillars. First, we introduce a novel lightweight Max-and-Attention Pillar Encoding (MAPE) module specially for enhancing small 3D objects. Second, we propose a simple yet effective principle for designing a backbone in pillar-based 3D detection. We construct FastPillars based on these designs, achieving high performance and low latency without SPConv. Extensive experiments on two large-scale datasets demonstrate the effectiveness and efficiency of FastPillars for on-device 3D detection regarding both performance and speed. Specifically, FastPillars delivers state-of-the-art accuracy on Waymo Open Dataset with 1.8X speed up and 3.8 mAPH/L2 improvement over CenterPoint (SPConv-based). Our code is publicly available at: https://github.com/StiphyJay/FastPillars.

Qingyuan Li, Yifan Zhang, Liang Li et al.

In the era of large-scale language models, the substantial parameter size poses significant challenges for deployment. Being a prevalent compression technique, quantization has emerged as the mainstream practice to tackle this issue, which is mainly centered on two recipes W8A8 and W4A16 (i.e. weights and activations in such bit widths). In this study, we propose a novel W4A8 post-training quantization method for the available open-sourced LLMs, which combines the advantages of both two recipes. Therefore, we can leverage the benefit in the I/O utilization of 4-bit weight quantization and the acceleration due to 8-bit matrix computation. Nevertheless, the W4A8 faces notorious performance degradation. As a remedy, we involve layerwise activation quantization strategies which feature a novel logarithmic equalization for most intractable layers, and we combine them with fine-grained weight quantization. Without whistles and bells, we eliminate the necessity for further fine-tuning and obtain the state-of-the-art W4A8 quantized performance on BLOOM, LLaMA, and LLaMA-2 on standard benchmarks. We confirm that the W4A8 quantization is achievable for the deployment of large language models, fostering their wide-spreading real-world applications.

Bowen Zhang, Zhi Tian, Quan Tang et al.

We explore the capability of plain Vision Transformers (ViTs) for semantic segmentation and propose the SegVit. Previous ViT-based segmentation networks usually learn a pixel-level representation from the output of the ViT. Differently, we make use of the fundamental component -- attention mechanism, to generate masks for semantic segmentation. Specifically, we propose the Attention-to-Mask (ATM) module, in which the similarity maps between a set of learnable class tokens and the spatial feature maps are transferred to the segmentation masks. Experiments show that our proposed SegVit using the ATM module outperforms its counterparts using the plain ViT backbone on the ADE20K dataset and achieves new state-of-the-art performance on COCO-Stuff-10K and PASCAL-Context datasets. Furthermore, to reduce the computational cost of the ViT backbone, we propose query-based down-sampling (QD) and query-based up-sampling (QU) to build a Shrunk structure. With the proposed Shrunk structure, the model can save up to $40\%$ computations while maintaining competitive performance.

Zhi Tian, Xiangxiang Chu, Xiaoming Wang et al.

We present a simple yet effective fully convolutional one-stage 3D object detector for LiDAR point clouds of autonomous driving scenes, termed FCOS-LiDAR. Unlike the dominant methods that use the bird-eye view (BEV), our proposed detector detects objects from the range view (RV, a.k.a. range image) of the LiDAR points. Due to the range view's compactness and compatibility with the LiDAR sensors' sampling process on self-driving cars, the range view-based object detector can be realized by solely exploiting the vanilla 2D convolutions, departing from the BEV-based methods which often involve complicated voxelization operations and sparse convolutions. For the first time, we show that an RV-based 3D detector with standard 2D convolutions alone can achieve comparable performance to state-of-the-art BEV-based detectors while being significantly faster and simpler. More importantly, almost all previous range view-based detectors only focus on single-frame point clouds, since it is challenging to fuse multi-frame point clouds into a single range view. In this work, we tackle this challenging issue with a novel range view projection mechanism, and for the first time demonstrate the benefits of fusing multi-frame point clouds for a range-view based detector. Extensive experiments on nuScenes show the superiority of our proposed method and we believe that our work can be strong evidence that an RV-based 3D detector can compare favourably with the current mainstream BEV-based detectors.

Wangbo Zhao, Kai Wang, Xiangxiang Chu et al.

Text-based video segmentation aims to segment the target object in a video based on a describing sentence. Incorporating motion information from optical flow maps with appearance and linguistic modalities is crucial yet has been largely ignored by previous work. In this paper, we design a method to fuse and align appearance, motion, and linguistic features to achieve accurate segmentation. Specifically, we propose a multi-modal video transformer, which can fuse and aggregate multi-modal and temporal features between frames. Furthermore, we design a language-guided feature fusion module to progressively fuse appearance and motion features in each feature level with guidance from linguistic features. Finally, a multi-modal alignment loss is proposed to alleviate the semantic gap between features from different modalities. Extensive experiments on A2D Sentences and J-HMDB Sentences verify the performance and the generalization ability of our method compared to the state-of-the-art methods.

Qingnan Ren, Shun Zou, Shiting Huang et al.

As autonomous coding agents become capable of handling increasingly long-horizon tasks, they have gradually demonstrated the potential to complete end-to-end software development. Although existing benchmarks have recently evolved from localized code editing to from-scratch project generation, they remain confined to structurally simplified, single-stack applications. Consequently, they fail to capture the heterogeneous environments, full-stack orchestration, and system-level complexity of real enterprise Software as a Service (SaaS) systems, leaving a critical gap in assessing agents under realistic engineering constraints. To fill this gap, we introduce SaaSBench, the first benchmark designed to explore the boundaries of AI agents in enterprise SaaS engineering. Spanning 30 complex tasks across 6 SaaS domains with 5,370 validation nodes, it incorporates 8 programming languages, 6 databases, and 13 frameworks to meticulously mirror real-world software heterogeneity. Furthermore, we design a dependency-aware hybrid evaluation paradigm tailored for complex systems with long horizons and multi-component coupling, enabling fine-grained, reproducible assessment. Crucially, our extensive experiments reveal a striking insight: the primary bottleneck for state-of-the-art agents is not generating isolated code logic, but successfully configuring and integrating a multi-component system. Over 95\% of task failures occur before agents even reach deep business logic, with models often falling victim to overconfidence and prematurely halting during foundational system setup, or getting trapped in ineffective debugging loops. We hope SaaSBench serves as a practical and challenging testbed to drive the evolution of reliable, system-level coding agents. The code is available at \url{https://github.com/ShadeCloak/SaaSbench}.

Xiangxiang Chu, Liang Li, Bo Zhang

The tradeoff between performance and inference speed is critical for practical applications. Architecture reparameterization obtains better tradeoffs and it is becoming an increasingly popular ingredient in modern convolutional neural networks. Nonetheless, its quantization performance is usually too poor to deploy (more than 20% top-1 accuracy drop on ImageNet) when INT8 inference is desired. In this paper, we dive into the underlying mechanism of this failure, where the original design inevitably enlarges quantization error. We propose a simple, robust, and effective remedy to have a quantization-friendly structure that also enjoys reparameterization benefits. Our method greatly bridges the gap between INT8 and FP32 accuracy for RepVGG. Without bells and whistles, the top-1 accuracy drop on ImageNet is reduced within 2% by standard post-training quantization. Moreover, our method also achieves similar FP32 performance as RepVGG. Extensive experiments on detection and semantic segmentation tasks verify its generalization.

Chengjian Feng, Zequn Jie, Yujie Zhong et al.

Recent LSS-based multi-view 3D object detection has made tremendous progress, by processing the features in Brid-Eye-View (BEV) via the convolutional detector. However, the typical convolution ignores the radial symmetry of the BEV features and increases the difficulty of the detector optimization. To preserve the inherent property of the BEV features and ease the optimization, we propose an azimuth-equivariant convolution (AeConv) and an azimuth-equivariant anchor. The sampling grid of AeConv is always in the radial direction, thus it can learn azimuth-invariant BEV features. The proposed anchor enables the detection head to learn predicting azimuth-irrelevant targets. In addition, we introduce a camera-decoupled virtual depth to unify the depth prediction for the images with different camera intrinsic parameters. The resultant detector is dubbed Azimuth-equivariant Detector (AeDet). Extensive experiments are conducted on nuScenes, and AeDet achieves a 62.0% NDS, surpassing the recent multi-view 3D object detectors such as PETRv2 and BEVDepth by a large margin. Project page: https://fcjian.github.io/aedet.

Kaiheng Weng, Xiangxiang Chu, Xiaoming Xu et al.

We present a hardware-efficient architecture of convolutional neural network, which has a repvgg-like architecture. Flops or parameters are traditional metrics to evaluate the efficiency of networks which are not sensitive to hardware including computing ability and memory bandwidth. Thus, how to design a neural network to efficiently use the computing ability and memory bandwidth of hardware is a critical problem. This paper proposes a method how to design hardware-aware neural network. Based on this method, we designed EfficientRep series convolutional networks, which are high-computation hardware(e.g. GPU) friendly and applied in YOLOv6 object detection framework. YOLOv6 has published YOLOv6N/YOLOv6S/YOLOv6M/YOLOv6L models in v1 and v2 versions.

Qingyuan Li, Ran Meng, Yiduo Li et al.

The large language model era urges faster and less costly inference. Prior model compression works on LLMs tend to undertake a software-centric approach primarily focused on the simulated quantization performance. By neglecting the feasibility of deployment, these approaches are typically disabled in real practice. They used to drastically push down the quantization bit range for a reduced computation which might not be supported by the mainstream hardware, or involve sophisticated algorithms that introduce extra computation or memory access overhead. We argue that pursuing a hardware-centric approach in the construction of quantization algorithms is crucial. In this regard, we are driven to build our compression method on top of hardware awareness, eliminating impractical algorithm choices while maximizing the benefit of hardware acceleration. Our method, OdysseyLLM, comes with a novel W4A8 kernel implementation called FastGEMM and a combined recipe of quantization strategies. Extensive experiments manifest the superiority of our W4A8 method which brings the actual speed boosting up to \textbf{4$\times$} compared to Hugging Face FP16 inference and \textbf{2.23$\times$} vs. the state-of-the-art inference engine TensorRT-LLM in FP16, and \textbf{1.45$\times$} vs. TensorRT-LLM in INT8, yet without substantially harming the performance.

Qingyuan Li, Bo Zhang, Xiangxiang Chu

Structured pruning greatly eases the deployment of large neural networks in resource-constrained environments. However, current methods either involve strong domain expertise, require extra hyperparameter tuning, or are restricted only to a specific type of network, which prevents pervasive industrial applications. In this paper, we undertake a simple and effective approach that can be easily applied to both vision transformers and convolutional neural networks. Specifically, we consider pruning as an evolution process of sub-network structures that inherit weights through reconstruction techniques. We achieve a 50% FLOPS reduction for ResNet50 and MobileNetV1, leading to 1.37x and 1.34x speedup respectively. For DeiT-Base, we reach nearly 40% FLOPs reduction and 1.4x speedup. Our code will be made available.

Xiangxiang Chu, Limeng Qiao, Xinyu Zhang et al.

We introduce MobileVLM V2, a family of significantly improved vision language models upon MobileVLM, which proves that a delicate orchestration of novel architectural design, an improved training scheme tailored for mobile VLMs, and rich high-quality dataset curation can substantially benefit VLMs' performance. Specifically, MobileVLM V2 1.7B achieves better or on-par performance on standard VLM benchmarks compared with much larger VLMs at the 3B scale. Notably, our 3B model outperforms a large variety of VLMs at the 7B+ scale. Our models will be released at https://github.com/Meituan-AutoML/MobileVLM .

Zhengze Xu, Dongyue Wu, Changqian Yu et al.

Recent real-time semantic segmentation methods usually adopt an additional semantic branch to pursue rich long-range context. However, the additional branch incurs undesirable computational overhead and slows inference speed. To eliminate this dilemma, we propose SCTNet, a single branch CNN with transformer semantic information for real-time segmentation. SCTNet enjoys the rich semantic representations of an inference-free semantic branch while retaining the high efficiency of lightweight single branch CNN. SCTNet utilizes a transformer as the training-only semantic branch considering its superb ability to extract long-range context. With the help of the proposed transformer-like CNN block CFBlock and the semantic information alignment module, SCTNet could capture the rich semantic information from the transformer branch in training. During the inference, only the single branch CNN needs to be deployed. We conduct extensive experiments on Cityscapes, ADE20K, and COCO-Stuff-10K, and the results show that our method achieves the new state-of-the-art performance. The code and model is available at https://github.com/xzz777/SCTNet

Yiming Hu, Xiangxiang Chu, Yong Wang

Neural Architecture Search (NAS) relies heavily on labeled data, which is labor-intensive and time-consuming to obtain. In this paper, we propose a novel NAS method based on an unsupervised paradigm, specifically Masked Autoencoders (MAE), thereby eliminating the need for labeled data. By replacing the supervised learning objective with an image reconstruction task, our approach enables the efficient discovery of network architectures without compromising performance and generalization ability. Additionally, we address the problem of performance collapse encountered in the widely-used Differentiable Architecture Search (DARTS) in the unsupervised setting by designing a hierarchical decoder. Extensive experiments across various datasets demonstrate the effectiveness and robustness of our method, offering empirical evidence of its superiority over the counterparts.

Ziyu Ma, Chenhui Gou, Yiming Hu et al.

Large Multimodal Models (LMMs) have shown promising in-context learning (ICL) capabilities, but scaling to many-shot settings remains difficult due to limited context length and high inference cost. To address these challenges, task-vector-based methods have been explored by inserting compact representations of many-shot in-context demonstrations into model activations. However, existing task-vector-based methods either overlook the importance of where to insert task vectors or struggle to determine suitable values for each location. To this end, we propose a novel Sensitivity-aware Task Vector insertion framework (STV) to figure out where and what to insert. Our key insight is that activation deltas across query-context pairs exhibit consistent structural patterns, providing a reliable cue for insertion. Based on the identified sensitive-aware locations, we construct a pre-clustered activation bank for each location by clustering the activation values, and then apply reinforcement learning to choose the most suitable one to insert. We evaluate STV across a range of multimodal models (e.g., Qwen-VL, Idefics-2) and tasks (e.g., VizWiz, OK-VQA), demonstrating its effectiveness and showing consistent improvements over previous task-vector-based methods with strong generalization.

Xiangxiang Chu, Hailang Huang, Xiao Zhang et al.

Reinforcement Learning (RL) can directly enhance the reasoning capabilities of large language models without extensive reliance on Supervised Fine-Tuning (SFT). In this work, we revisit the traditional Policy Gradient (PG) mechanism and propose a minimalist RL approach termed Group Policy Gradient (GPG). Unlike conventional methods, GPG directly optimize the original RL objective, thus obviating the need for surrogate loss functions. By eliminating the critic and reference models, avoiding KL divergence constraints, and addressing the advantage and gradient estimation bias, our approach significantly simplifies the training process compared to Group Relative Policy Optimization (GRPO). Our approach achieves superior performance without relying on auxiliary techniques or adjustments. As illustrated in Figure 1, extensive experiments demonstrate that our method not only reduces computational costs but also consistently outperforms GRPO across various unimodal and multimodal tasks. Our code is available at https://github.com/AMAP-ML/GPG.

Xiangxiang Chu, Limeng Qiao, Xinyang Lin et al.

We present MobileVLM, a competent multimodal vision language model (MMVLM) targeted to run on mobile devices. It is an amalgamation of a myriad of architectural designs and techniques that are mobile-oriented, which comprises a set of language models at the scale of 1.4B and 2.7B parameters, trained from scratch, a multimodal vision model that is pre-trained in the CLIP fashion, cross-modality interaction via an efficient projector. We evaluate MobileVLM on several typical VLM benchmarks. Our models demonstrate on par performance compared with a few much larger models. More importantly, we measure the inference speed on both a Qualcomm Snapdragon 888 CPU and an NVIDIA Jeston Orin GPU, and we obtain state-of-the-art performance of 21.5 tokens and 65.3 tokens per second, respectively. Our code will be made available at: https://github.com/Meituan-AutoML/MobileVLM.

Honghao Chen, Yurong Zhang, Xiaokun Feng et al.

Robustness is a vital aspect to consider when deploying deep learning models into the wild. Numerous studies have been dedicated to the study of the robustness of vision transformers (ViTs), which have dominated as the mainstream backbone choice for vision tasks since the dawn of 2020s. Recently, some large kernel convnets make a comeback with impressive performance and efficiency. However, it still remains unclear whether large kernel networks are robust and the attribution of their robustness. In this paper, we first conduct a comprehensive evaluation of large kernel convnets' robustness and their differences from typical small kernel counterparts and ViTs on six diverse robustness benchmark datasets. Then to analyze the underlying factors behind their strong robustness, we design experiments from both quantitative and qualitative perspectives to reveal large kernel convnets' intriguing properties that are completely different from typical convnets. Our experiments demonstrate for the first time that pure CNNs can achieve exceptional robustness comparable or even superior to that of ViTs. Our analysis on occlusion invariance, kernel attention patterns and frequency characteristics provide novel insights into the source of robustness.

Sifan Zhou, Liang Li, Xinyu Zhang et al.

Due to highly constrained computing power and memory, deploying 3D lidar-based detectors on edge devices equipped in autonomous vehicles and robots poses a crucial challenge. Being a convenient and straightforward model compression approach, Post-Training Quantization (PTQ) has been widely adopted in 2D vision tasks. However, applying it directly to 3D lidar-based tasks inevitably leads to performance degradation. As a remedy, we propose an effective PTQ method called LiDAR-PTQ, which is particularly curated for 3D lidar detection (both SPConv-based and SPConv-free). Our LiDAR-PTQ features three main components, \textbf{(1)} a sparsity-based calibration method to determine the initialization of quantization parameters, \textbf{(2)} a Task-guided Global Positive Loss (TGPL) to reduce the disparity between the final predictions before and after quantization, \textbf{(3)} an adaptive rounding-to-nearest operation to minimize the layerwise reconstruction error. Extensive experiments demonstrate that our LiDAR-PTQ can achieve state-of-the-art quantization performance when applied to CenterPoint (both Pillar-based and Voxel-based). To our knowledge, for the very first time in lidar-based 3D detection tasks, the PTQ INT8 model's accuracy is almost the same as the FP32 model while enjoying $3\times$ inference speedup. Moreover, our LiDAR-PTQ is cost-effective being $30\times$ faster than the quantization-aware training method. Code will be released at \url{https://github.com/StiphyJay/LiDAR-PTQ}.

Fei Wei, Xinyu Zhang, Ailing Zhang et al.

With the fast-paced development of multimodal large language models (MLLMs), we can now converse with AI systems in natural languages to understand images. However, the reasoning power and world knowledge embedded in the large language models have been much less investigated and exploited for image perception tasks. In this paper, we propose Lenna, a language-enhanced reasoning detection assistant, which utilizes the robust multimodal feature representation of MLLMs, while preserving location information for detection. This is achieved by incorporating an additional <DET> token in the MLLM vocabulary that is free of explicit semantic context but serves as a prompt for the detector to identify the corresponding position. To evaluate the reasoning capability of Lenna, we construct a ReasonDet dataset to measure its performance on reasoning-based detection. Remarkably, Lenna demonstrates outstanding performance on ReasonDet and comes with significantly low training costs. It also incurs minimal transferring overhead when extended to other tasks. Our code and model will be available at https://git.io/Lenna.

Xiangxiang Chu, Jianlin Su, Bo Zhang et al.

Large language models are built on top of a transformer-based architecture to process textual inputs. For example, the LLaMA stands out among many open-source implementations. Can the same transformer be used to process 2D images? In this paper, we answer this question by unveiling a LLaMA-like vision transformer in plain and pyramid forms, termed VisionLLaMA, which is tailored for this purpose. VisionLLaMA is a unified and generic modelling framework for solving most vision tasks. We extensively evaluate its effectiveness using typical pre-training paradigms in a good portion of downstream tasks of image perception and especially image generation. In many cases, VisionLLaMA have exhibited substantial gains over the previous state-of-the-art vision transformers. We believe that VisionLLaMA can serve as a strong new baseline model for vision generation and understanding. Our code is released at https://github.com/Meituan-AutoML/VisionLLaMA.

Xiangxiang Chu, Renda Li, Yong Wang

Recent studies have highlighted the interplay between diffusion models and representation learning. Intermediate representations from diffusion models can be leveraged for downstream visual tasks, while self-supervised vision models can enhance the convergence and generation quality of diffusion models. However, transferring pretrained weights from vision models to diffusion models is challenging due to input mismatches and the use of latent spaces. To address these challenges, we propose Unified Self-supervised Pretraining (USP), a framework that initializes diffusion models via masked latent modeling in a Variational Autoencoder (VAE) latent space. USP achieves comparable performance in understanding tasks while significantly improving the convergence speed and generation quality of diffusion models. Our code will be publicly available at https://github.com/AMAP-ML/USP.

Rui Lan, Yancheng Bai, Xu Duan et al.

Scene text editing aims to modify or add texts on images while ensuring text fidelity and overall visual quality consistent with the background. Recent methods are primarily built on UNet-based diffusion models, which have improved scene text editing results, but still struggle with complex glyph structures, especially for non-Latin ones (\eg, Chinese, Korean, Japanese). To address these issues, we present \textbf{FLUX-Text}, a simple and advanced multilingual scene text editing DiT method. Specifically, our FLUX-Text enhances glyph understanding and generation through lightweight Visual and Text Embedding Modules, while preserving the original generative capability of FLUX. We further propose a Regional Text Perceptual Loss tailored for text regions, along with a matching two-stage training strategy to better balance text editing and overall image quality. Benefiting from the DiT-based architecture and lightweight feature injection modules, FLUX-Text can be trained with only $0.1$M training examples, a \textbf{97\%} reduction compared to $2.9$M required by popular methods. Extensive experiments on multiple public datasets, including English and Chinese benchmarks, demonstrate that our method surpasses other methods in visual quality and text fidelity. All the code is available at https://github.com/AMAP-ML/FluxText.

Rui Chen, Lei Sun, Jing Tang et al.

Recent advances in video generation have posed great challenges in the assessment of AI-generated content, particularly with the emergence of increasingly sophisticated models. The various inconsistencies and defects observed in such videos are inherently complex, making overall scoring notoriously difficult. In this paper, we emphasize the critical importance of integrating fine-grained reasoning into video evaluation, and we propose $\textbf{F}$ing$\textbf{ER}$, a novel entity-level reasoning evaluation framework that first automatically generates $\textbf{F}$ine-grained $\textbf{E}$ntity-level questions, and then answers those questions by a $\textbf{R}$easoning model with scores, which can be subsequently weighted summed to an overall score for different applications. Specifically, we leverage LLMs to derive entity-level questions across five distinct perspectives, which (i) often focus on some specific entities of the content, thereby making answering or scoring much easier by MLLMs, and (ii) are more interpretable. Then we construct a FingER dataset, consisting of approximately 3.3k videos and corresponding 60k fine-grained QA annotations, each with detailed reasons. Based on that, we further investigate various training protocols to best incentivize the reasoning capability of MLLMs for correct answer prediction. Extensive experiments demonstrate that a reasoning model trained using Group Relative Policy Optimization (GRPO) with a cold-start strategy achieves the best performance. Notably, our model surpasses existing methods by a relative margin of $11.8\%$ on GenAI-Bench and $5.5\%$ on MonetBench with only 3.3k training videos, which is at most one-tenth of the training samples utilized by other methods. Our code and dataset will be released soon.

Yurong Zhang, Honghao Chen, Xinyu Zhang et al.

Parameter-efficient transfer learning (PETL) is a promising task, aiming to adapt the large-scale pre-trained model to downstream tasks with a relatively modest cost. However, current PETL methods struggle in compressing computational complexity and bear a heavy inference burden due to the complete forward process. This paper presents an efficient visual recognition paradigm, called Dynamic Adapter (Dyn-Adapter), that boosts PETL efficiency by subtly disentangling features in multiple levels. Our approach is simple: first, we devise a dynamic architecture with balanced early heads for multi-level feature extraction, along with adaptive training strategy. Second, we introduce a bidirectional sparsity strategy driven by the pursuit of powerful generalization ability. These qualities enable us to fine-tune efficiently and effectively: we reduce FLOPs during inference by 50%, while maintaining or even yielding higher recognition accuracy. Extensive experiments on diverse datasets and pretrained backbones demonstrate the potential of Dyn-Adapter serving as a general efficiency booster for PETL in vision recognition tasks.

Chubin Chen, Jiashu Zhu, Xiaokun Feng et al.

Classifier-free Guidance (CFG) is a widely used technique in modern diffusion models for enhancing sample quality and prompt adherence. However, through an empirical analysis on Gaussian mixture modeling with a closed-form solution, we observe a discrepancy between the suboptimal results produced by CFG and the ground truth. The model's excessive reliance on these suboptimal predictions often leads to semantic incoherence and low-quality outputs. To address this issue, we first empirically demonstrate that the model's suboptimal predictions can be effectively refined using sub-networks of the model itself. Building on this insight, we propose S^2-Guidance, a novel method that leverages stochastic block-dropping during the forward process to construct stochastic sub-networks, effectively guiding the model away from potential low-quality predictions and toward high-quality outputs. Extensive qualitative and quantitative experiments on text-to-image and text-to-video generation tasks demonstrate that S^2-Guidance delivers superior performance, consistently surpassing CFG and other advanced guidance strategies. Our code will be released.

Ryan Xu, Dongyang Jin, Yancheng Bai et al.

Controllable image synthesis, which enables fine-grained control over generated outputs, has emerged as a key focus in visual generative modeling. However, controllable generation remains challenging for Visual Autoregressive (VAR) models due to their hierarchical, next-scale prediction style. Existing VAR-based methods often suffer from inefficient control encoding and disruptive injection mechanisms that compromise both fidelity and efficiency. In this work, we present SCALAR, a controllable generation method based on VAR, incorporating a novel Scale-wise Conditional Decoding mechanism. SCALAR leverages a pretrained image encoder to extract semantic control signal encodings, which are projected into scale-specific representations and injected into the corresponding layers of the VAR backbone. This design provides persistent and structurally aligned guidance throughout the generation process. Building on SCALAR, we develop SCALAR-Uni, a unified extension that aligns multiple control modalities into a shared latent space, supporting flexible multi-conditional guidance in a single model. Extensive experiments show that SCALAR achieves superior generation quality and control precision across various tasks. The code is released at https://github.com/AMAP-ML/SCALAR.

Hanyu Guo, Chao Chen, Longfei Xu et al.

We argue that the decision-making essence of route recommendation is comparative judgment: users choose a route because it is better than alternatives in specific aspects. The decision-critical information resides in segment-level spatial differences of non-overlapping parts between routes, which is irreversibly lost through item-level feature aggregation. Existing methods, whether attention-based or pairwise ranking approaches, follow an item-first paradigm that can only infer pairwise relations indirectly from individual route representations. To address this, we propose the Comprehensive Comparison Network (CCN), which inverts the information flow by constructing explicit comparison features from non-overlapping segments between route pairs and reasoning directly in the pairwise space. CCN introduces a Comprehensive Comparison Block that enables context-aware pairwise reasoning, where the comparison between two routes is informed by how both compare against all other candidates. We further develop an interpretable Pair Scoring Network that decomposes pairwise preferences into independent physical fields, providing field-level explanations for route selection. CCN has served as the production ranking model in Amap for over two years, achieving 85.70% offline route-trajectory coverage rate and +1.2% online improvement over the previous production model. We also release a large-scale route recommendation dataset comprising 175 million users, 512 million samples, and 6 billion routes across 370 cities.

Huaqiu Li, Yong Wang, Tongwen Huang et al.

Unified image restoration is a significantly challenging task in low-level vision. Existing methods either make tailored designs for specific tasks, limiting their generalizability across various types of degradation, or rely on training with paired datasets, thereby suffering from closed-set constraints. To address these issues, we propose a novel, dataset-free, and unified approach through recurrent posterior sampling utilizing a pretrained latent diffusion model. Our method incorporates the multimodal understanding model to provide sematic priors for the generative model under a task-blind condition. Furthermore, it utilizes a lightweight module to align the degraded input with the generated preference of the diffusion model, and employs recurrent refinement for posterior sampling. Extensive experiments demonstrate that our method outperforms state-of-the-art methods, validating its effectiveness and robustness. Our code and data are available at https://github.com/AMAP-ML/LD-RPS.

Renda Li, Hailang Huang, Fei Wei et al.

Reinforcement learning (RL) has demonstrated considerable potential for enhancing reasoning in large language models (LLMs). However, existing methods suffer from Gradient Starvation and Policy Degradation when training directly on samples with mixed difficulty. To mitigate this, prior approaches leverage Chain-of-Thought (CoT) data, but the construction of high-quality CoT annotations remains labor-intensive. Alternatively, curriculum learning strategies have been explored but frequently encounter challenges, such as difficulty mismatch, reliance on manual curriculum design, and catastrophic forgetting. To address these issues, we propose AdaCuRL, a Adaptive Curriculum Reinforcement Learning framework that integrates coarse-to-fine difficulty estimation with adaptive curriculum scheduling. This approach dynamically aligns data difficulty with model capability and incorporates a data revisitation mechanism to mitigate catastrophic forgetting. Furthermore, AdaCuRL employs adaptive reference and sparse KL strategies to prevent Policy Degradation. Extensive experiments across diverse reasoning benchmarks demonstrate that AdaCuRL consistently achieves significant performance improvements on both LLMs and MLLMs.

Xiao Zhang, Fei Wei, Yong Wang et al.

Zero-shot domain adaptation (ZSDA) presents substantial challenges due to the lack of images in the target domain. Previous approaches leverage Vision-Language Models (VLMs) to tackle this challenge, exploiting their zero-shot learning capabilities. However, these methods primarily address domain distribution shifts and overlook the misalignment between the detection task and VLMs, which rely on manually crafted prompts. To overcome these limitations, we propose the unified prompt and representation enhancement (UPRE) framework, which jointly optimizes both textual prompts and visual representations. Specifically, our approach introduces a multi-view domain prompt that combines linguistic domain priors with detection-specific knowledge, and a visual representation enhancement module that produces domain style variations. Furthermore, we introduce multi-level enhancement strategies, including relative domain distance and positive-negative separation, which align multi-modal representations at the image level and capture diverse visual representations at the instance level, respectively. Extensive experiments conducted on nine benchmark datasets demonstrate the superior performance of our framework in ZSDA detection scenarios. Code is available at https://github.com/AMAP-ML/UPRE.

Xiangxiang Chu, Zhi Tian, Yuqing Wang et al.

Very recently, a variety of vision transformer architectures for dense prediction tasks have been proposed and they show that the design of spatial attention is critical to their success in these tasks. In this work, we revisit the design of the spatial attention and demonstrate that a carefully-devised yet simple spatial attention mechanism performs favourably against the state-of-the-art schemes. As a result, we propose two vision transformer architectures, namely, Twins-PCPVT and Twins-SVT. Our proposed architectures are highly-efficient and easy to implement, only involving matrix multiplications that are highly optimized in modern deep learning frameworks. More importantly, the proposed architectures achieve excellent performance on a wide range of visual tasks, including image level classification as well as dense detection and segmentation. The simplicity and strong performance suggest that our proposed architectures may serve as stronger backbones for many vision tasks. Our code is released at https://github.com/Meituan-AutoML/Twins .

Xiangxiang Chu, Zhi Tian, Bo Zhang et al.

We propose a conditional positional encoding (CPE) scheme for vision Transformers. Unlike previous fixed or learnable positional encodings, which are pre-defined and independent of input tokens, CPE is dynamically generated and conditioned on the local neighborhood of the input tokens. As a result, CPE can easily generalize to the input sequences that are longer than what the model has ever seen during training. Besides, CPE can keep the desired translation-invariance in the image classification task, resulting in improved performance. We implement CPE with a simple Position Encoding Generator (PEG) to get seamlessly incorporated into the current Transformer framework. Built on PEG, we present Conditional Position encoding Vision Transformer (CPVT). We demonstrate that CPVT has visually similar attention maps compared to those with learned positional encodings and delivers outperforming results. Our code is available at https://github.com/Meituan-AutoML/CPVT .

Xiangxiang Chu, Xiaoxing Wang, Bo Zhang et al.

Despite the fast development of differentiable architecture search (DARTS), it suffers from long-standing performance instability, which extremely limits its application. Existing robustifying methods draw clues from the resulting deteriorated behavior instead of finding out its causing factor. Various indicators such as Hessian eigenvalues are proposed as a signal to stop searching before the performance collapses. However, these indicator-based methods tend to easily reject good architectures if the thresholds are inappropriately set, let alone the searching is intrinsically noisy. In this paper, we undertake a more subtle and direct approach to resolve the collapse. We first demonstrate that skip connections have a clear advantage over other candidate operations, where it can easily recover from a disadvantageous state and become dominant. We conjecture that this privilege is causing degenerated performance. Therefore, we propose to factor out this benefit with an auxiliary skip connection, ensuring a fairer competition for all operations. We call this approach DARTS-. Extensive experiments on various datasets verify that it can substantially improve robustness. Our code is available at https://github.com/Meituan-AutoML/DARTS- .

Xiangxiang Chu, Bo Zhang

Simplicity is the ultimate sophistication. Differentiable Architecture Search (DARTS) has now become one of the mainstream paradigms of neural architecture search. However, it largely suffers from the well-known performance collapse issue due to the aggregation of skip connections. It is thought to have overly benefited from the residual structure which accelerates the information flow. To weaken this impact, we propose to inject unbiased random noise to impede the flow. We name this novel approach NoisyDARTS. In effect, a network optimizer should perceive this difficulty at each training step and refrain from overshooting, especially on skip connections. In the long run, since we add no bias to the gradient in terms of expectation, it is still likely to converge to the right solution area. We also prove that the injected noise plays a role in smoothing the loss landscape, which makes the optimization easier. Our method features extreme simplicity and acts as a new strong baseline. We perform extensive experiments across various search spaces, datasets, and tasks, where we robustly achieve state-of-the-art results. Our code is available at https://github.com/xiaomi-automl/NoisyDARTS.

Xiangxiang Chu, Tianbao Zhou, Bo Zhang et al.

Differentiable Architecture Search (DARTS) is now a widely disseminated weight-sharing neural architecture search method. However, it suffers from well-known performance collapse due to an inevitable aggregation of skip connections. In this paper, we first disclose that its root cause lies in an unfair advantage in exclusive competition. Through experiments, we show that if either of two conditions is broken, the collapse disappears. Thereby, we present a novel approach called Fair DARTS where the exclusive competition is relaxed to be collaborative. Specifically, we let each operation's architectural weight be independent of others. Yet there is still an important issue of discretization discrepancy. We then propose a zero-one loss to push architectural weights towards zero or one, which approximates an expected multi-hot solution. Our experiments are performed on two mainstream search spaces, and we derive new state-of-the-art results on CIFAR-10 and ImageNet. Our code is available on https://github.com/xiaomi-automl/fairdarts .

Xiangxiang Chu, Bo Zhang, Qingyuan Li et al.

To discover powerful yet compact models is an important goal of neural architecture search. Previous two-stage one-shot approaches are limited by search space with a fixed depth. It seems handy to include an additional skip connection in the search space to make depths variable. However, it creates a large range of perturbation during supernet training and it has difficulty giving a confident ranking for subnetworks. In this paper, we discover that skip connections bring about significant feature inconsistency compared with other operations, which potentially degrades the supernet performance. Based on this observation, we tackle the problem by imposing an equivariant learnable stabilizer to homogenize such disparities. Experiments show that our proposed stabilizer helps to improve the supernet's convergence as well as ranking performance. With an evolutionary search backend that incorporates the stabilized supernet as an evaluator, we derive a family of state-of-the-art architectures, the SCARLET series of several depths, especially SCARLET-A obtains 76.9% top-1 accuracy on ImageNet. Code is available at https://github.com/xiaomi-automl/ScarletNAS.

Xiangxiang Chu, Bo Zhang, Ruijun Xu

The evolution of MobileNets has laid a solid foundation for neural network applications on mobile end. With the latest MobileNetV3, neural architecture search again claimed its supremacy in network design. Unfortunately, till today all mobile methods mainly focus on CPU latencies instead of GPU, the latter, however, is much preferred in practice for it has faster speed, lower overhead and less interference. Bearing the target hardware in mind, we propose the first Mobile GPU-Aware (MoGA) neural architecture search in order to be precisely tailored for real-world applications. Further, the ultimate objective to devise a mobile network lies in achieving better performance by maximizing the utilization of bounded resources. Urging higher capability while restraining time consumption is not reconcilable. We alleviate the tension by weighted evolution techniques. Moreover, we encourage increasing the number of parameters for higher representational power. With 200x fewer GPU days than MnasNet, we obtain a series of models that outperform MobileNetV3 under the similar latency constraints, i.e., MoGA-A achieves 75.9% top-1 accuracy on ImageNet, MoGA-B meets 75.5% which costs only 0.5 ms more on mobile GPU. MoGA-C best attests GPU-awareness by reaching 75.3% and being slower on CPU but faster on GPU.The models and test code is made available here https://github.com/xiaomi-automl/MoGA.

Xiangxiang Chu, Bo Zhang, Ruijun Xu

One of the most critical problems in weight-sharing neural architecture search is the evaluation of candidate models within a predefined search space. In practice, a one-shot supernet is trained to serve as an evaluator. A faithful ranking certainly leads to more accurate searching results. However, current methods are prone to making misjudgments. In this paper, we prove that their biased evaluation is due to inherent unfairness in the supernet training. In view of this, we propose two levels of constraints: expectation fairness and strict fairness. Particularly, strict fairness ensures equal optimization opportunities for all choice blocks throughout the training, which neither overestimates nor underestimates their capacity. We demonstrate that this is crucial for improving the confidence of models' ranking. Incorporating the one-shot supernet trained under the proposed fairness constraints with a multi-objective evolutionary search algorithm, we obtain various state-of-the-art models, e.g., FairNAS-A attains 77.5% top-1 validation accuracy on ImageNet. The models and their evaluation codes are made publicly available online http://github.com/fairnas/FairNAS .

Xiangxiang Chu

As the most successful variant and improvement for Trust Region Policy Optimization (TRPO), proximal policy optimization (PPO) has been widely applied across various domains with several advantages: efficient data utilization, easy implementation, and good parallelism. In this paper, a first-order gradient reinforcement learning algorithm called Policy Optimization with Penalized Point Probability Distance (POP3D), which is a lower bound to the square of total variance divergence is proposed as another powerful variant. Firstly, we talk about the shortcomings of several commonly used algorithms, by which our method is partly motivated. Secondly, we address to overcome these shortcomings by applying POP3D. Thirdly, we dive into its mechanism from the perspective of solution manifold. Finally, we make quantitative comparisons among several state-of-the-art algorithms based on common benchmarks. Simulation results show that POP3D is highly competitive compared with PPO. Besides, our code is released in https://github.com/paperwithcode/pop3d.

Honghao Chen, Xiangxiang Chu, Yongjian Ren et al.

Recently, some large kernel convnets strike back with appealing performance and efficiency. However, given the square complexity of convolution, scaling up kernels can bring about an enormous amount of parameters and the proliferated parameters can induce severe optimization problem. Due to these issues, current CNNs compromise to scale up to 51x51 in the form of stripe convolution (i.e., 51x5 + 5x51) and start to saturate as the kernel size continues growing. In this paper, we delve into addressing these vital issues and explore whether we can continue scaling up kernels for more performance gains. Inspired by human vision, we propose a human-like peripheral convolution that efficiently reduces over 90% parameter count of dense grid convolution through parameter sharing, and manage to scale up kernel size to extremely large. Our peripheral convolution behaves highly similar to human, reducing the complexity of convolution from O(K^2) to O(logK) without backfiring performance. Built on this, we propose Parameter-efficient Large Kernel Network (PeLK). Our PeLK outperforms modern vision Transformers and ConvNet architectures like Swin, ConvNeXt, RepLKNet and SLaK on various vision tasks including ImageNet classification, semantic segmentation on ADE20K and object detection on MS COCO. For the first time, we successfully scale up the kernel size of CNNs to an unprecedented 101x101 and demonstrate consistent improvements.

Sule Bai, Mingxing Li, Yong Liu et al.

Traditional visual grounding methods primarily focus on single-image scenarios with simple textual references. However, extending these methods to real-world scenarios that involve implicit and complex instructions, particularly in conjunction with multiple images, poses significant challenges, which is mainly due to the lack of advanced reasoning ability across diverse multi-modal contexts. In this work, we aim to address the more practical universal grounding task, and propose UniVG-R1, a reasoning guided multimodal large language model (MLLM) for universal visual grounding, which enhances reasoning capabilities through reinforcement learning (RL) combined with cold-start data. Specifically, we first construct a high-quality Chain-of-Thought (CoT) grounding dataset, annotated with detailed reasoning chains, to guide the model towards correct reasoning paths via supervised fine-tuning. Subsequently, we perform rule-based reinforcement learning to encourage the model to identify correct reasoning chains, thereby incentivizing its reasoning capabilities. In addition, we identify a difficulty bias arising from the prevalence of easy samples as RL training progresses, and we propose a difficulty-aware weight adjustment strategy to further strengthen the performance. Experimental results demonstrate the effectiveness of UniVG-R1, which achieves state-of-the-art performance on MIG-Bench with a 9.1% improvement over the previous method. Furthermore, our model exhibits strong generalizability, achieving an average improvement of 23.4% in zero-shot performance across four image and video reasoning grounding benchmarks. The project page can be accessed at https://amap-ml.github.io/UniVG-R1-page/.

Zhenlong Yuan, Jing Tang, Jinguo Luo et al.

Vision-Language-Action (VLA) models in autonomous driving systems have recently demonstrated transformative potential by integrating multimodal perception with decision-making capabilities. However, the interpretability and coherence of the decision process and the plausibility of action sequences remain largely underexplored. To address these issues, we propose AutoDrive-R$^2$, a novel VLA framework that enhances both reasoning and self-reflection capabilities of autonomous driving systems through chain-of-thought (CoT) processing and reinforcement learning (RL). Specifically, we first propose an innovative CoT dataset named nuScenesR$^2$-6K for supervised fine-tuning, which effectively builds cognitive bridges between input information and output trajectories through a four-step logical chain with self-reflection for validation. Moreover, to maximize both reasoning and self-reflection during the RL stage, we further employ the Group Relative Policy Optimization (GRPO) algorithm within a physics-grounded reward framework that incorporates spatial alignment, vehicle dynamic, and temporal smoothness criteria to ensure reliable and realistic trajectory planning. Extensive evaluation results across both nuScenes and Waymo datasets demonstrates the state-of-the-art performance and robust generalization capacity of our proposed method.

Yifei Wang, Feng Xiong, Yong Wang et al.

Positional bias (PB), manifesting as non-uniform sensitivity across different contextual locations, significantly impairs long-context comprehension and processing capabilities. Previous studies have addressed PB either by modifying the underlying architectures or by employing extensive contextual awareness training. However, the former approach fails to effectively eliminate the substantial performance disparities, while the latter imposes significant data and computational overhead. To address PB effectively, we introduce \textbf{Pos2Distill}, a position to position knowledge distillation framework. Pos2Distill transfers the superior capabilities from advantageous positions to less favorable ones, thereby reducing the huge performance gaps. The conceptual principle is to leverage the inherent, position-induced disparity to counteract the PB itself. We identify distinct manifestations of PB under \textbf{\textsc{r}}etrieval and \textbf{\textsc{r}}easoning paradigms, thereby designing two specialized instantiations: \emph{Pos2Distill-R\textsuperscript{1}} and \emph{Pos2Distill-R\textsuperscript{2}} respectively, both grounded in this core principle. By employing the Pos2Distill approach, we achieve enhanced uniformity and significant performance gains across all contextual positions in long-context retrieval and reasoning tasks. Crucially, both specialized systems exhibit strong cross-task generalization mutually, while achieving superior performance on their respective tasks.

Zhenlong Yuan, Xiangyan Qu, Chengxuan Qian et al.

Multimodal large language models (MLLMs) have demonstrated remarkable potential in bridging visual and textual reasoning, yet their reliance on text-centric priors often limits their ability to disentangle semantically similar actions in open-vocabulary scenarios. To address this, we propose Video-STAR, a framework that harmonizes contextual sub-motion decomposition with tool-augmented reinforcement learning for open-vocabulary action recognition (OVAR). Unlike prior methods that treat actions as monolithic entities, our approach innovatively decomposes actions into discriminative sub-motions for fine-grained matching while dynamically invoking domain-specific tools for cross-modal interleaving, thereby enabling category-specific reasoning capacity and reducing cross-modal hallucination. Moreover, by designing a hierarchical reward that balances tool-usage efficiency, sub-motion relevance, and structural coherence in reasoning, our method autonomously leverages external tools to prioritize sub-motion patterns without explicit supervision, transmitting from text-centric reasoning to visually grounded inference. Extensive evaluations on HMDB-51, UCF-101, SSv2, Kinetics-400, and Kinetics-600 datasets demonstrate our state-of-the-art performance, outperforming existing methods in distinguishing fine-grained actions and handling cross-modal hallucination, validating our excellent robustness and generalization.

Dingyuan Shi, Yong Wang, Hangyu Li et al.

Diffusion models have shown remarkable success in text-to-image generation, making preference alignment for these models increasingly important. The preference labels are typically available only at the terminal of denoising trajectories, which poses challenges in optimizing the intermediate denoising steps. In this paper, we propose to conduct Denoised Distribution Estimation (DDE) that explicitly connects intermediate steps to the terminal denoised distribution. Therefore, preference labels can be used for the entire trajectory optimization. To this end, we design two estimation strategies for our DDE. The first is stepwise estimation, which utilizes the conditional denoised distribution to estimate the model denoised distribution. The second is single-shot estimation, which converts the model output into the terminal denoised distribution via DDIM modeling. Analytically and empirically, we reveal that DDE equipped with two estimation strategies naturally derives a novel credit assignment scheme that prioritizes optimizing the middle part of the denoising trajectory. Extensive experiments demonstrate that our approach achieves superior performance, both quantitatively and qualitatively.

Hailang Huang, Yong Wang, Zixuan Huang et al.

Large Multimodal Models (LMMs) demonstrate impressive capabilities. However, current benchmarks predominantly focus on image comprehension in specific domains, and these benchmarks are labor-intensive to construct. Moreover, their answers tend to be brief, making it difficult to assess the ability of LMMs to generate detailed descriptions of images. To address these limitations, we propose the MMGenBench-Pipeline, a straightforward and fully automated evaluation pipeline. This involves generating textual descriptions from input images, using these descriptions to create auxiliary images via text-to-image generative models, and then comparing the original and generated images. Furthermore, to ensure the effectiveness of MMGenBench-Pipeline, we design MMGenBench-Test, evaluating LMMs across 13 distinct image patterns, and MMGenBench-Domain, focusing on generative image performance. A thorough evaluation involving over 50 popular LMMs demonstrates the effectiveness and reliability of both the pipeline and benchmark. Our observations indicate that numerous LMMs excelling in existing benchmarks fail to adequately complete the basic tasks related to image understanding and description. This finding highlights the substantial potential for performance improvement in current LMMs and suggests avenues for future model optimization. Concurrently, MMGenBench-Pipeline can efficiently assess the performance of LMMs across diverse domains using only image inputs.