Yihao Chen

Xianbiao Qi, Jianan Wang, Yihao Chen et al.

We present a Lipschitz continuous Transformer, called LipsFormer, to pursue training stability both theoretically and empirically for Transformer-based models. In contrast to previous practical tricks that address training instability by learning rate warmup, layer normalization, attention formulation, and weight initialization, we show that Lipschitz continuity is a more essential property to ensure training stability. In LipsFormer, we replace unstable Transformer component modules with Lipschitz continuous counterparts: CenterNorm instead of LayerNorm, spectral initialization instead of Xavier initialization, scaled cosine similarity attention instead of dot-product attention, and weighted residual shortcut. We prove that these introduced modules are Lipschitz continuous and derive an upper bound on the Lipschitz constant of LipsFormer. Our experiments show that LipsFormer allows stable training of deep Transformer architectures without the need of careful learning rate tuning such as warmup, yielding a faster convergence and better generalization. As a result, on the ImageNet 1K dataset, LipsFormer-Swin-Tiny based on Swin Transformer training for 300 epochs can obtain 82.7\% without any learning rate warmup. Moreover, LipsFormer-CSwin-Tiny, based on CSwin, training for 300 epochs achieves a top-1 accuracy of 83.5\% with 4.7G FLOPs and 24M parameters. The code will be released at \url{https://github.com/IDEA-Research/LipsFormer}.

Yihao Chen, Xianbiao Qi, Jianan Wang et al.

We propose DisCo-CLIP, a distributed memory-efficient CLIP training approach, to reduce the memory consumption of contrastive loss when training contrastive learning models. Our approach decomposes the contrastive loss and its gradient computation into two parts, one to calculate the intra-GPU gradients and the other to compute the inter-GPU gradients. According to our decomposition, only the intra-GPU gradients are computed on the current GPU, while the inter-GPU gradients are collected via all_reduce from other GPUs instead of being repeatedly computed on every GPU. In this way, we can reduce the GPU memory consumption of contrastive loss computation from $\bigO(B^2)$ to $\bigO(\frac{B^2}{N})$, where $B$ and $N$ are the batch size and the number of GPUs used for training. Such a distributed solution is mathematically equivalent to the original non-distributed contrastive loss computation, without sacrificing any computation accuracy. It is particularly efficient for large-batch CLIP training. For instance, DisCo-CLIP can enable contrastive training of a ViT-B/32 model with a batch size of 32K or 196K using 8 or 64 A100 40GB GPUs, compared with the original CLIP solution which requires 128 A100 40GB GPUs to train a ViT-B/32 model with a batch size of 32K. The code will be released at https://github.com/IDEA-Research/DisCo-CLIP

Meituan LongCat Team, Anchun Gui, Bei Li et al.

We introduce LongCat-Flash-Thinking-2601, a 560-billion-parameter open-source Mixture-of-Experts (MoE) reasoning model with superior agentic reasoning capability. LongCat-Flash-Thinking-2601 achieves state-of-the-art performance among open-source models on a wide range of agentic benchmarks, including agentic search, agentic tool use, and tool-integrated reasoning. Beyond benchmark performance, the model demonstrates strong generalization to complex tool interactions and robust behavior under noisy real-world environments. Its advanced capability stems from a unified training framework that combines domain-parallel expert training with subsequent fusion, together with an end-to-end co-design of data construction, environments, algorithms, and infrastructure spanning from pre-training to post-training. In particular, the model's strong generalization capability in complex tool-use are driven by our in-depth exploration of environment scaling and principled task construction. To optimize long-tailed, skewed generation and multi-turn agentic interactions, and to enable stable training across over 10,000 environments spanning more than 20 domains, we systematically extend our asynchronous reinforcement learning framework, DORA, for stable and efficient large-scale multi-environment training. Furthermore, recognizing that real-world tasks are inherently noisy, we conduct a systematic analysis and decomposition of real-world noise patterns, and design targeted training procedures to explicitly incorporate such imperfections into the training process, resulting in improved robustness for real-world applications. To further enhance performance on complex reasoning tasks, we introduce a Heavy Thinking mode that enables effective test-time scaling by jointly expanding reasoning depth and width through intensive parallel thinking.

Yu Zheng, Yihao Chen, Jinghan Peng et al.

This report describes our speaker verification systems for the tasks of the CN-Celeb Speaker Recognition Challenge 2022 (CNSRC 2022). This challenge includes two tasks, namely speaker verification(SV) and speaker retrieval(SR). The SV task involves two tracks: fixed track and open track. In the fixed track, we only used CN-Celeb.T as the training set. For the open track of the SV task and SR task, we added our open-source audio data. The ResNet-based, RepVGG-based, and TDNN-based architectures were developed for this challenge. Global statistic pooling structure and MQMHA pooling structure were used to aggregate the frame-level features across time to obtain utterance-level representation. We adopted AM-Softmax and AAM-Softmax combined with the Sub-Center method to classify the resulting embeddings. We also used the Large-Margin Fine-Tuning strategy to further improve the model performance. In the backend, Sub-Mean and AS-Norm were used. In the SV task fixed track, our system was a fusion of five models, and two models were fused in the SV task open track. And we used a single system in the SR task. Our approach leads to superior performance and comes the 1st place in the open track of the SV task, the 2nd place in the fixed track of the SV task, and the 3rd place in the SR task.

He Cao, Jianan Wang, Tianhe Ren et al.

Score-based diffusion models have captured widespread attention and funded fast progress of recent vision generative tasks. In this paper, we focus on diffusion model backbone which has been much neglected before. We systematically explore vision Transformers as diffusion learners for various generative tasks. With our improvements the performance of vanilla ViT-based backbone (IU-ViT) is boosted to be on par with traditional U-Net-based methods. We further provide a hypothesis on the implication of disentangling the generative backbone as an encoder-decoder structure and show proof-of-concept experiments verifying the effectiveness of a stronger encoder for generative tasks with ASymmetriC ENcoder Decoder (ASCEND). Our improvements achieve competitive results on CIFAR-10, CelebA, LSUN, CUB Bird and large-resolution text-to-image tasks. To the best of our knowledge, we are the first to successfully train a single diffusion model on text-to-image task beyond 64x64 resolution. We hope this will motivate people to rethink the modeling choices and the training pipelines for diffusion-based generative models.

Jianye Pang, Cheng Jiang, Yihao Chen et al.

Dense prediction in medical volume provides enriched guidance for clinical analysis. CNN backbones have met bottleneck due to lack of long-range dependencies and global context modeling power. Recent works proposed to combine vision transformer with CNN, due to its strong global capture ability and learning capability. However, most works are limited to simply applying pure transformer with several fatal flaws (i.e., lack of inductive bias, heavy computation and little consideration for 3D data). Therefore, designing an elegant and efficient vision transformer learner for dense prediction in medical volume is promising and challenging. In this paper, we propose a novel 3D Shuffle-Mixer network of a new Local Vision Transformer-MLP paradigm for medical dense prediction. In our network, a local vision transformer block is utilized to shuffle and learn spatial context from full-view slices of rearranged volume, a residual axial-MLP is designed to mix and capture remaining volume context in a slice-aware manner, and a MLP view aggregator is employed to project the learned full-view rich context to the volume feature in a view-aware manner. Moreover, an Adaptive Scaled Enhanced Shortcut is proposed for local vision transformer to enhance feature along spatial and channel dimensions adaptively, and a CrossMerge is proposed to skip-connects the multi-scale feature appropriately in the pyramid architecture. Extensive experiments demonstrate the proposed model outperforms other state-of-the-art medical dense prediction methods.

Yu Zheng, Jinghan Peng, Yihao Chen et al.

This paper describes speaker verification (SV) systems submitted by the SpeakIn team to the Task 1 and Task 2 of the Far-Field Speaker Verification Challenge 2022 (FFSVC2022). SV tasks of the challenge focus on the problem of fully supervised far-field speaker verification (Task 1) and semi-supervised far-field speaker verification (Task 2). In Task 1, we used the VoxCeleb and FFSVC2020 datasets as train datasets. And for Task 2, we only used the VoxCeleb dataset as train set. The ResNet-based and RepVGG-based architectures were developed for this challenge. Global statistic pooling structure and MQMHA pooling structure were used to aggregate the frame-level features across time to obtain utterance-level representation. We adopted AM-Softmax and AAM-Softmax to classify the resulting embeddings. We innovatively propose a staged transfer learning method. In the pre-training stage we reserve the speaker weights, and there are no positive samples to train them in this stage. Then we fine-tune these weights with both positive and negative samples in the second stage. Compared with the traditional transfer learning strategy, this strategy can better improve the model performance. The Sub-Mean and AS-Norm backend methods were used to solve the problem of domain mismatch. In the fusion stage, three models were fused in Task1 and two models were fused in Task2. On the FFSVC2022 leaderboard, the EER of our submission is 3.0049% and the corresponding minDCF is 0.2938 in Task1. In Task2, EER and minDCF are 6.2060% and 0.5232 respectively. Our approach leads to excellent performance and ranks 1st in both challenge tasks.

Shufa Wei, Xiaolong Xu, Xianbiao Qi et al.

Large Language Models (LLMs) have demonstrated exceptional capabilities across various natural language processing tasks. Yet, many of these advanced LLMs are tailored for broad, general-purpose applications. In this technical report, we introduce AcademicGPT, designed specifically to empower academic research. AcademicGPT is a continual training model derived from LLaMA2-70B. Our training corpus mainly consists of academic papers, thesis, content from some academic domain, high-quality Chinese data and others. While it may not be extensive in data scale, AcademicGPT marks our initial venture into a domain-specific GPT tailored for research area. We evaluate AcademicGPT on several established public benchmarks such as MMLU and CEval, as well as on some specialized academic benchmarks like PubMedQA, SCIEval, and our newly-created ComputerScienceQA, to demonstrate its ability from general knowledge ability, to Chinese ability, and to academic ability. Building upon AcademicGPT's foundation model, we also developed several applications catered to the academic area, including General Academic Question Answering, AI-assisted Paper Reading, Paper Review, and AI-assisted Title and Abstract Generation.

Tianhe Ren, Qing Jiang, Shilong Liu et al.

This paper introduces Grounding DINO 1.5, a suite of advanced open-set object detection models developed by IDEA Research, which aims to advance the "Edge" of open-set object detection. The suite encompasses two models: Grounding DINO 1.5 Pro, a high-performance model designed for stronger generalization capability across a wide range of scenarios, and Grounding DINO 1.5 Edge, an efficient model optimized for faster speed demanded in many applications requiring edge deployment. The Grounding DINO 1.5 Pro model advances its predecessor by scaling up the model architecture, integrating an enhanced vision backbone, and expanding the training dataset to over 20 million images with grounding annotations, thereby achieving a richer semantic understanding. The Grounding DINO 1.5 Edge model, while designed for efficiency with reduced feature scales, maintains robust detection capabilities by being trained on the same comprehensive dataset. Empirical results demonstrate the effectiveness of Grounding DINO 1.5, with the Grounding DINO 1.5 Pro model attaining a 54.3 AP on the COCO detection benchmark and a 55.7 AP on the LVIS-minival zero-shot transfer benchmark, setting new records for open-set object detection. Furthermore, the Grounding DINO 1.5 Edge model, when optimized with TensorRT, achieves a speed of 75.2 FPS while attaining a zero-shot performance of 36.2 AP on the LVIS-minival benchmark, making it more suitable for edge computing scenarios. Model examples and demos with API will be released at https://github.com/IDEA-Research/Grounding-DINO-1.5-API

Shunming Tao, Yihao Chen, Jingxing Wu et al.

Virus infection is one of the major diseases that seriously threaten human health. To meet the growing demand for mining and sharing data resources related to antiviral drugs and to accelerate the design and discovery of new antiviral drugs, we presented an open-access antiviral drug resource and machine learning platform (VDDB), which, to the best of our knowledge, is the first comprehensive dedicated resource for experimentally verified potential drugs/molecules based on manually curated data. Currently, VDDB highlights 848 clinical vaccines, 199 clinical antibodies, as well as over 710,000 small molecules targeting 39 medically important viruses including SARS-CoV-2. Furthermore, VDDB stores approximately 3 million records of pharmacological data for these collected potential antiviral drugs/molecules, involving 314 cell infection-based phenotypic and 234 target-based genotypic assays. Based on these annotated pharmacological data, VDDB allows users to browse, search and download reliable information about these collects for various viruses of interest. In particular, VDDB also integrates 57 cell infection- and 117 target-based associated high-accuracy machine learning models to support various antivirals identification-related tasks, such as compound activity prediction, virtual screening, drug repositioning and target fishing. VDDB is freely accessible at http://vddb.idruglab.cn.

Han Shu, Wenshuo Li, Yehui Tang et al.

Recently segment anything model (SAM) has shown powerful segmentation capability and has drawn great attention in computer vision fields. Massive following works have developed various applications based on the pre-trained SAM and achieved impressive performance on downstream vision tasks. However, SAM consists of heavy architectures and requires massive computational capacity, which hinders the further application of SAM on computation constrained edge devices. To this end, in this paper we propose a framework to obtain a tiny segment anything model (TinySAM) while maintaining the strong zero-shot performance. We first propose a full-stage knowledge distillation method with hard prompt sampling and hard mask weighting strategy to distill a lightweight student model. We also adapt the post-training quantization to the prompt-based segmentation task and further reduce the computational cost. Moreover, a hierarchical segmenting everything strategy is proposed to accelerate the everything inference by $2\times$ with almost no performance degradation. With all these proposed methods, our TinySAM leads to orders of magnitude computational reduction and pushes the envelope for efficient segment anything task. Extensive experiments on various zero-shot transfer tasks demonstrate the significantly advantageous performance of our TinySAM against counterpart methods. Codes are available at https://github.com/xinghaochen/TinySAM and https://gitee.com/mindspore/models/tree/master/research/cv/TinySAM.

Qing Jiang, Gen Luo, Yuqin Yang et al.

Perception and understanding are two pillars of computer vision. While multimodal large language models (MLLM) have demonstrated remarkable visual understanding capabilities, they arguably lack accurate perception abilities, e.g. the stage-of-the-art model Qwen2-VL only achieves a 43.9 recall rate on the COCO dataset, limiting many tasks requiring the combination of perception and understanding. In this work, we aim to bridge this perception gap from both model designing and data development perspectives. We first introduce ChatRex, an MLLM with a decoupled perception design. Instead of having the LLM directly predict box coordinates, we feed the output boxes from a universal proposal network into the LLM, allowing it to output the corresponding box indices to represent its detection results, turning the regression task into a retrieval-based task that LLM handles more proficiently. From the data perspective, we build a fully automated data engine and construct the Rexverse-2M dataset which possesses multiple granularities to support the joint training of perception and understanding. After a three-stage training approach, ChatRex demonstrates strong perception and understanding performance, and the combination of these two capabilities also unlocks many attractive applications, demonstrating their complementary roles in MLLM. Code is available at https://github.com/IDEA-Research/ChatRex.

Yihao Chen, Haochen Wu, Nan Jiang et al.

This paper describes the USTC-KXDIGIT system submitted to the ASVspoof5 Challenge for Track 1 (speech deepfake detection) and Track 2 (spoofing-robust automatic speaker verification, SASV). Track 1 showcases a diverse range of technical qualities from potential processing algorithms and includes both open and closed conditions. For these conditions, our system consists of a cascade of a frontend feature extractor and a back-end classifier. We focus on extensive embedding engineering and enhancing the generalization of the back-end classifier model. Specifically, the embedding engineering is based on hand-crafted features and speech representations from a self-supervised model, used for closed and open conditions, respectively. To detect spoof attacks under various adversarial conditions, we trained multiple systems on an augmented training set. Additionally, we used voice conversion technology to synthesize fake audio from genuine audio in the training set to enrich the synthesis algorithms. To leverage the complementary information learned by different model architectures, we employed activation ensemble and fused scores from different systems to obtain the final decision score for spoof detection. During the evaluation phase, the proposed methods achieved 0.3948 minDCF and 14.33% EER in the close condition, and 0.0750 minDCF and 2.59% EER in the open condition, demonstrating the robustness of our submitted systems under adversarial conditions. In Track 2, we continued using the CM system from Track 1 and fused it with a CNN-based ASV system. This approach achieved 0.2814 min-aDCF in the closed condition and 0.0756 min-aDCF in the open condition, showcasing superior performance in the SASV system.

Liye Chen, Biaoshun Li, Yihao Chen et al.

Antibody-drug conjugate (ADC) has revolutionized the field of cancer treatment in the era of precision medicine due to their ability to precisely target cancer cells and release highly effective drug. Nevertheless, the realization of rational design of ADC is very difficult because the relationship between their structures and activities is difficult to understand. In the present study, we introduce a unified deep learning framework called ADCNet to help design potential ADCs. The ADCNet highly integrates the protein representation learning language model ESM-2 and small-molecule representation learning language model FG-BERT models to achieve activity prediction through learning meaningful features from antigen and antibody protein sequences of ADC, SMILES strings of linker and payload, and drug-antibody ratio (DAR) value. Based on a carefully designed and manually tailored ADC data set, extensive evaluation results reveal that ADCNet performs best on the test set compared to baseline machine learning models across all evaluation metrics. For example, it achieves an average prediction accuracy of 87.12%, a balanced accuracy of 0.8689, and an area under receiver operating characteristic curve of 0.9293 on the test set. In addition, cross-validation, ablation experiments, and external independent testing results further prove the stability, advancement, and robustness of the ADCNet architecture. For the convenience of the community, we develop the first online platform (https://ADCNet.idruglab.cn) for the prediction of ADCs activity based on the optimal ADCNet model, and the source code is publicly available at https://github.com/idrugLab/ADCNet.

Qing Jiang, Lin Wu, Zhaoyang Zeng et al.

Humans are undoubtedly the most important participants in computer vision, and the ability to detect any individual given a natural language description, a task we define as referring to any person, holds substantial practical value. However, we find that existing models generally fail to achieve real-world usability, and current benchmarks are limited by their focus on one-to-one referring, that hinder progress in this area. In this work, we revisit this task from three critical perspectives: task definition, dataset design, and model architecture. We first identify five aspects of referable entities and three distinctive characteristics of this task. Next, we introduce HumanRef, a novel dataset designed to tackle these challenges and better reflect real-world applications. From a model design perspective, we integrate a multimodal large language model with an object detection framework, constructing a robust referring model named RexSeek. Experimental results reveal that state-of-the-art models, which perform well on commonly used benchmarks like RefCOCO/+/g, struggle with HumanRef due to their inability to detect multiple individuals. In contrast, RexSeek not only excels in human referring but also generalizes effectively to common object referring, making it broadly applicable across various perception tasks. Code is available at https://github.com/IDEA-Research/RexSeek

Yiheng Cao, Yihao Chen, Xin Hu et al.

Detecting vulnerabilities in source code remains critical yet challenging, as conventional static analysis tools construct inaccurate program representations, while existing LLM-based approaches often miss essential vulnerability context and lack grounded reasoning. To mitigate these challenges, we introduce VulWeaver, a novel LLM-based approach that weaves broken program semantics into accurate representations and extracts holistic vulnerability context for grounded vulnerability detection. Specifically, VulWeaver first constructs an enhanced unified dependency graph (UDG) by integrating deterministic rules with LLM-based semantic inference to address static analysis inaccuracies. It then extracts holistic vulnerability context by combining explicit contexts from program slicing with implicit contexts, including usage, definition, and declaration information. Finally, VulWeaver employs meta-prompting with vulnerability type specific expert guidelines to steer LLMs through systematic reasoning, aggregated via majority voting for robustness. Extensive experiments on PrimeVul4J dataset have demonstrated that VulWeaver achieves a F1-score of 0.75, outperforming state-of-the-art learning-based, LLM-based, and agent-based baselines by 23%, 15%, and 60% in F1-score, respectively. VulWeaver has also detected 26 true vulnerabilities across 9 realworld Java projects, with 15 confirmed by developers and 5 CVE identifiers assigned. In industrial deployment, VulWeaver identified 40 confirmed vulnerabilities in an internal repository.

Ming Xu, Yihao Chen, Ji Guan

Matrix product states (MPS) are a standard tensor-network representation for ground states of one-dimensional quantum many-body systems, and they underpin widely used simulation tools such as DMRG. However, while quantum model checking has been developed mainly for quantum programs and communication protocols (with properties expressed along a time axis), there is still no comparable framework for systematically verifying \emph{spatial} and \emph{size-dependent} properties of physical many-body states, where the key parameter is the system size. This paper takes a step toward bridging the gap. We propose \emph{Linear Chain Logic} (LCL), a spatial logic designed to specify physically meaningful properties of periodic MPS families as the system size grows, such as nontriviality on rings and large-size asymptotic patterns. Our approach builds on a simple but powerful connection: every periodic MPS naturally induces a completely positive map (a quantum operation) on its virtual space, so many quantitative features of the MPS can be analysed through the repeated application of the operation. Using this perspective, we derive an effective procedure to compute the inner products of an MPS at a given size and to support richer LCL specifications, without relying on brute-force state expansion. We then develop approximate model-checking algorithms that combine sound bounding with asymptotic structural analysis, enabling scalable reasoning about large system sizes. Experiments on representative MPS families illustrate that our method can automatically verify nontriviality and detect asymptotic spatial regimes in a way that complements traditional numerical techniques.

Yihao Chen, Zefang Wang

Channel pruning is a promising method for accelerating and compressing convolutional neural networks. However, current pruning algorithms still remain unsolved problems that how to assign layer-wise pruning ratios properly and discard the least important channels with a convincing criterion. In this paper, we present a novel channel pruning approach via information theory and interpretability of neural networks. Specifically, we regard information entropy as the expected amount of information for convolutional layers. In addition, if we suppose a matrix as a system of linear equations, a higher-rank matrix represents there exist more solutions to it, which indicates more uncertainty. From the point of view of information theory, the rank can also describe the amount of information. In a neural network, considering the rank and entropy as two information indicators of convolutional layers, we propose a fusion function to reach a compromise of them, where the fusion results are defined as ``information concentration''. When pre-defining layer-wise pruning ratios, we employ the information concentration as a reference instead of heuristic and engineering tuning to provide a more interpretable solution. Moreover, we leverage Shapley values, which are a potent tool in the interpretability of neural networks, to evaluate the channel contributions and discard the least important channels for model compression while maintaining its performance. Extensive experiments demonstrate the effectiveness and promising performance of our method. For example, our method improves the accuracy by 0.21% when reducing 45.5% FLOPs and removing 40.3% parameters for ResNet-56 on CIFAR-10. Moreover, our method obtains loss in Top-1/Top-5 accuracies of 0.43%/0.11% by reducing 41.6% FLOPs and removing 35.0% parameters for ResNet-50 on ImageNet.

Gustavo A. Oliva, Gopi Krishnan Rajbahadur, Aaditya Bhatia et al.

High-quality labeled datasets are crucial for training and evaluating foundation models in software engineering, but creating them is often prohibitively expensive and labor-intensive. We introduce SPICE, a scalable, automated pipeline for labeling SWE-bench-style datasets with annotations for issue clarity, test coverage, and effort estimation. SPICE combines context-aware code navigation, rationale-driven prompting, and multi-pass consensus to produce labels that closely approximate expert annotations. SPICE's design was informed by our own experience and frustration in labeling more than 800 instances from SWE-Gym. SPICE achieves strong agreement with human-labeled SWE-bench Verified data while reducing the cost of labeling 1,000 instances from around \$100,000 (manual annotation) to just \$5.10. These results demonstrate SPICE's potential to enable cost-effective, large-scale dataset creation for SE-focused FMs. To support the community, we release both SPICE tool and SPICE Bench, a new dataset of 6,802 SPICE-labeled instances curated from 291 open-source projects in SWE-Gym (over 13x larger than SWE-bench Verified).

Dunwei Tu, Hongyan Hao, Hansi Yang et al.

Training generalist agents capable of adapting to diverse scenarios requires interactive environments for self-exploration. However, interactive environments remain critically scarce, and existing synthesis methods suffer from significant limitations regarding environmental diversity and scalability. To address these challenges, we introduce ScaleEnv, a framework that constructs fully interactive environments and verifiable tasks entirely from scratch. Specifically, ScaleEnv ensures environment reliability through procedural testing, and guarantees task completeness and solvability via tool dependency graph expansion and executable action verification. By enabling agents to learn through exploration within ScaleEnv, we demonstrate significant performance improvements on unseen, multi-turn tool-use benchmarks such as $τ^2$-Bench and VitaBench, highlighting strong generalization capabilities. Furthermore, we investigate the relationship between increasing number of domains and model generalization performance, providing empirical evidence that scaling environmental diversity is critical for robust agent learning.

Yihao Chen, Xin Tang, Xianbiao Qi et al.

Graph Neural Networks (GNNs) have attracted considerable attention and have emerged as a new promising paradigm to process graph-structured data. GNNs are usually stacked to multiple layers and the node representations in each layer are computed through propagating and aggregating the neighboring node features with respect to the graph. By stacking to multiple layers, GNNs are able to capture the long-range dependencies among the data on the graph and thus bring performance improvements. To train a GNN with multiple layers effectively, some normalization techniques (e.g., node-wise normalization, batch-wise normalization) are necessary. However, the normalization techniques for GNNs are highly task-relevant and different application tasks prefer to different normalization techniques, which is hard to know in advance. To tackle this deficiency, in this paper, we propose to learn graph normalization by optimizing a weighted combination of normalization techniques at four different levels, including node-wise normalization, adjacency-wise normalization, graph-wise normalization, and batch-wise normalization, in which the adjacency-wise normalization and the graph-wise normalization are newly proposed in this paper to take into account the local structure and the global structure on the graph, respectively. By learning the optimal weights, we are able to automatically select a single best or a best combination of multiple normalizations for a specific task. We conduct extensive experiments on benchmark datasets for different tasks, including node classification, link prediction, graph classification and graph regression, and confirm that the learned graph normalization leads to competitive results and that the learned weights suggest the appropriate normalization techniques for the specific task. Source code is released here https://github.com/cyh1112/GraphNormalization.

Ning Lu, Wenwen Yu, Xianbiao Qi et al.

Attention-based scene text recognizers have gained huge success, which leverages a more compact intermediate representation to learn 1d- or 2d- attention by a RNN-based encoder-decoder architecture. However, such methods suffer from attention-drift problem because high similarity among encoded features leads to attention confusion under the RNN-based local attention mechanism. Moreover, RNN-based methods have low efficiency due to poor parallelization. To overcome these problems, we propose the MASTER, a self-attention based scene text recognizer that (1) not only encodes the input-output attention but also learns self-attention which encodes feature-feature and target-target relationships inside the encoder and decoder and (2) learns a more powerful and robust intermediate representation to spatial distortion, and (3) owns a great training efficiency because of high training parallelization and a high-speed inference because of an efficient memory-cache mechanism. Extensive experiments on various benchmarks demonstrate the superior performance of our MASTER on both regular and irregular scene text. Pytorch code can be found at https://github.com/wenwenyu/MASTER-pytorch, and Tensorflow code can be found at https://github.com/jiangxiluning/MASTER-TF.

Yihao Chen, Yue Yang

We develop a deep reinforcement learning method for training a jellyfish-like swimmer to effectively track a moving target in a two-dimensional flow. This swimmer is a flexible object equipped with a muscle model based on torsional springs. We employ a deep Q-network (DQN) that takes the swimmer's geometry and dynamic parameters as inputs, and outputs actions which are the forces applied to the swimmer. In particular, we introduce an action regulation to mitigate the interference from complex fluid-structure interactions. The goal of these actions is to navigate the swimmer to a target point in the shortest possible time. In the DQN training, the data on the swimmer's motions are obtained from simulations conducted using the immersed boundary method. During tracking a moving target, there is an inherent delay between the application of forces and the corresponding response of the swimmer's body due to hydrodynamic interactions between the shedding vortices and the swimmer's own locomotion. Our tests demonstrate that the swimmer, with the DQN agent and action regulation, is able to dynamically adjust its course based on its instantaneous state. This work extends the application scope of machine learning in controlling flexible objects within fluid environments.

Tianhe Ren, Yihao Chen, Qing Jiang et al.

In this paper, we introduce DINO-X, which is a unified object-centric vision model developed by IDEA Research with the best open-world object detection performance to date. DINO-X employs the same Transformer-based encoder-decoder architecture as Grounding DINO 1.5 to pursue an object-level representation for open-world object understanding. To make long-tailed object detection easy, DINO-X extends its input options to support text prompt, visual prompt, and customized prompt. With such flexible prompt options, we develop a universal object prompt to support prompt-free open-world detection, making it possible to detect anything in an image without requiring users to provide any prompt. To enhance the model's core grounding capability, we have constructed a large-scale dataset with over 100 million high-quality grounding samples, referred to as Grounding-100M, for advancing the model's open-vocabulary detection performance. Pre-training on such a large-scale grounding dataset leads to a foundational object-level representation, which enables DINO-X to integrate multiple perception heads to simultaneously support multiple object perception and understanding tasks, including detection, segmentation, pose estimation, object captioning, object-based QA, etc. Experimental results demonstrate the superior performance of DINO-X. Specifically, the DINO-X Pro model achieves 56.0 AP, 59.8 AP, and 52.4 AP on the COCO, LVIS-minival, and LVIS-val zero-shot object detection benchmarks, respectively. Notably, it scores 63.3 AP and 56.5 AP on the rare classes of LVIS-minival and LVIS-val benchmarks, improving the previous SOTA performance by 5.8 AP and 5.0 AP. Such a result underscores its significantly improved capacity for recognizing long-tailed objects.

Yuqi Li, Yao Lu, Junhao Dong et al.

Layer pruning has emerged as a potent approach to remove redundant layers in the pre-trained network on the purpose of reducing network size and improve computational efficiency. However, existing layer pruning methods mostly overlook the intrinsic connections and inter-dependencies between different layers within complicated deep neural networks. This oversight can result in pruned models that do not preserve the essential characteristics of the pre-trained network as effectively as desired. To address these limitations, we propose a Similarity-Guided Layer Partition (SGLP) Pruning, a novel pruning framework that exploits representation similarity to guide efficient and informed layer removal for compressing large deep models. Our method begins by employing Centered Kernel Alignment (CKA) to quantify representational similarity between layers, uncovering structural patterns within the network. We then apply Fisher Optimal Segmentation on the similarity matrix to partition the network into semantically coherent layer segments. This segmentation allows pruning decisions to respect layer interdependencies and preserve essential knowledge. Within each segment, we introduce a fine-tuning-free importance evaluation using GradNorm, identifying and removing redundant layers in a targeted, segment-wise manner. Experimental results on both image classification tasks and large language models (LLMs) demonstrate that our proposed SGLP outperforms the state-of-the-art methods in accuracy and efficiency. Our approach achieves significant model compression with minimal performance degradation, making it well-suited for deployment in resource-limited environments.

Zhimin Zhao, Yihao Chen, Abdul Ali Bangash et al.

In machine learning (ML), efficient asset management, including ML models, datasets, algorithms, and tools, is vital for resource optimization, consistent performance, and a streamlined development lifecycle. This enables quicker iterations, adaptability, reduced development-to-deployment time, and reliable outputs. Despite existing research, a significant knowledge gap remains in operational challenges like model versioning, data traceability, and collaboration, which are crucial for the success of ML projects. Our study aims to address this gap by analyzing 15,065 posts from developer forums and platforms, employing a mixed-method approach to classify inquiries, extract challenges using BERTopic, and identify solutions through open card sorting and BERTopic clustering. We uncover 133 topics related to asset management challenges, grouped into 16 macro-topics, with software dependency, model deployment, and model training being the most discussed. We also find 79 solution topics, categorized under 18 macro-topics, highlighting software dependency, feature development, and file management as key solutions. This research underscores the need for further exploration of identified pain points and the importance of collaborative efforts across academia, industry, and the research community.

Anton Obukhov, Matteo Poggi, Fabio Tosi et al.

This paper presents the results of the fourth edition of the Monocular Depth Estimation Challenge (MDEC), which focuses on zero-shot generalization to the SYNS-Patches benchmark, a dataset featuring challenging environments in both natural and indoor settings. In this edition, we revised the evaluation protocol to use least-squares alignment with two degrees of freedom to support disparity and affine-invariant predictions. We also revised the baselines and included popular off-the-shelf methods: Depth Anything v2 and Marigold. The challenge received a total of 24 submissions that outperformed the baselines on the test set; 10 of these included a report describing their approach, with most leading methods relying on affine-invariant predictions. The challenge winners improved the 3D F-Score over the previous edition's best result, raising it from 22.58% to 23.05%.

Qing Jiang, Junan Huo, Xingyu Chen et al.

Object detection has long been dominated by traditional coordinate regression-based models, such as YOLO, DETR, and Grounding DINO. Although recent efforts have attempted to leverage MLLMs to tackle this task, they face challenges like low recall rate, duplicate predictions, coordinate misalignment, etc. In this work, we bridge this gap and propose Rex-Omni, a 3B-scale MLLM that achieves state-of-the-art object perception performance. On benchmarks like COCO and LVIS, Rex-Omni attains performance comparable to or exceeding regression-based models (e.g., DINO, Grounding DINO) in a zero-shot setting. This is enabled by three key designs: 1) Task Formulation: we use special tokens to represent quantized coordinates from 0 to 999, reducing the model's learning difficulty and improving token efficiency for coordinate prediction; 2) Data Engines: we construct multiple data engines to generate high-quality grounding, referring, and pointing data, providing semantically rich supervision for training; \3) Training Pipelines: we employ a two-stage training process, combining supervised fine-tuning on 22 million data with GRPO-based reinforcement post-training. This RL post-training leverages geometry-aware rewards to effectively bridge the discrete-to-continuous coordinate prediction gap, improve box accuracy, and mitigate undesirable behaviors like duplicate predictions that stem from the teacher-guided nature of the initial SFT stage. Beyond conventional detection, Rex-Omni's inherent language understanding enables versatile capabilities such as object referring, pointing, visual prompting, GUI grounding, spatial referring, OCR and key-pointing, all systematically evaluated on dedicated benchmarks. We believe that Rex-Omni paves the way for more versatile and language-aware visual perception systems.

Zhiyu Fan, Kirill Vasilevski, Dayi Lin et al.

The advancement of large language models (LLMs) and code agents has demonstrated significant potential to assist software engineering (SWE) tasks, such as autonomous issue resolution and feature addition. Existing AI for software engineering leaderboards (e.g., SWE-bench) focus solely on solution accuracy, ignoring the crucial factor of effectiveness in a resource-constrained world. This is a universal problem that also exists beyond software engineering tasks: any AI system should be more than correct - it must also be cost-effective. To address this gap, we introduce SWE-Effi, a set of new metrics to re-evaluate AI systems in terms of holistic effectiveness scores. We define effectiveness as the balance between the accuracy of outcome (e.g., issue resolve rate) and the resources consumed (e.g., token and time). In this paper, we specifically focus on the software engineering scenario by re-ranking popular AI systems for issue resolution on a subset of the SWE-bench benchmark using our new multi-dimensional metrics. We found that AI system's effectiveness depends not just on the scaffold itself, but on how well it integrates with the base model, which is key to achieving strong performance in a resource-efficient manner. We also identified systematic challenges such as the "token snowball" effect and, more significantly, a pattern of "expensive failures". In these cases, agents consume excessive resources while stuck on unsolvable tasks - an issue that not only limits practical deployment but also drives up the cost of failed rollouts during RL training. Lastly, we observed a clear trade-off between effectiveness under the token budget and effectiveness under the time budget, which plays a crucial role in managing project budgets and enabling scalable reinforcement learning, where fast responses are essential.

Yucun Hou, Fenglin Zhan, Chenxi Li et al.

Deep learning-based reconstruction of positron emission tomography(PET) data has gained increasing attention in recent years. While these methods achieve fast reconstruction,concerns remain regarding quantitative accuracy and the presence of artifacts,stemming from limited model interpretability,data driven dependence, and overfitting risks.These challenges have hindered clinical adoption.To address them,we propose a conditional diffusion model with posterior physical correction (PET-DPC) for PET image reconstruction. An innovative normalization procedure generates the input Geometric TOF Probabilistic Image (GTP-image),while physical information is incorporated during the diffusion sampling process to perform posterior scatter,attenuation,and random corrections. The model was trained and validated on 300 brain and 50 whole-body PET datasets,a physical phantom,and 20 simulated brain datasets. PET-DPC produced reconstructions closely aligned with fully corrected OSEM images,outperforming end-to-end deep learning models in quantitative metrics and,in some cases, surpassing traditional iterative methods. The model also generalized well to out-of-distribution(OOD) data. Compared to iterative methods,PET-DPC reduced reconstruction time by 50% for brain scans and 85% for whole-body scans. Ablation studies confirmed the critical role of posterior correction in implementing scatter and attenuation corrections,enhancing reconstruction accuracy. Experiments with physical phantoms further demonstrated PET-DPC's ability to preserve background uniformity and accurately reproduce tumor-to-background intensity ratios. Overall,these results highlight PET-DPC as a promising approach for rapid, quantitatively accurate PET reconstruction,with strong potential to improve clinical imaging workflows.

Zeyan Li, Junjie Chen, Yihao Chen et al.

Localizing root causes for multi-dimensional data is critical to ensure online service systems' reliability. When a fault occurs, only the measure values within specific attribute combinations are abnormal. Such attribute combinations are substantial clues to the underlying root causes and thus are called root causes of multidimensional data. This paper proposes a generic and robust root cause localization approach for multi-dimensional data, PSqueeze. We propose a generic property of root cause for multi-dimensional data, generalized ripple effect (GRE). Based on it, we propose a novel probabilistic cluster method and a robust heuristic search method. Moreover, we identify the importance of determining external root causes and propose an effective method for the first time in literature. Our experiments on two real-world datasets with 5400 faults show that the F1-score of PSqueeze outperforms baselines by 32.89%, while the localization time is around 10 seconds across all cases. The F1-score in determining external root causes of PSqueeze achieves 0.90. Furthermore, case studies in several production systems demonstrate that PSqueeze is helpful to fault diagnosis in the real world.

Cheng Jiang, Yihao Chen, Jianbo Chang et al.

Doctors often make diagonostic decisions based on patient's image scans, such as magnetic resonance imaging (MRI), and patient's electronic health records (EHR) such as age, gender, blood pressure and so on. Despite a lot of automatic methods have been proposed for either image or text analysis in computer vision or natural language research areas, much fewer studies have been developed for the fusion of medical image and EHR data for medical problems. Among existing early or intermediate fusion methods, concatenation of features from both modalities is still a mainstream. For a better exploiting of image and EHR data, we propose a multi-modal attention module which use EHR data to help the selection of important regions during image feature extraction process conducted by traditional CNN. Moreover, we propose to incorporate multi-head machnism to gated multimodal unit (GMU) to make it able to parallelly fuse image and EHR features in different subspaces. With the help of the two modules, existing CNN architecture can be enhanced using both modalities. Experiments on predicting Glasgow outcome scale (GOS) of intracerebral hemorrhage patients and classifying Alzheimer's Disease showed the proposed method can automatically focus on task-related areas and achieve better results by making better use of image and EHR features.

Yuxiang Zhong, Xianbiao Qi, Shanjun Li et al.

In this technical report, we present our 1st place solution for the ICDAR 2021 competition on mathematical formula detection (MFD). The MFD task has three key challenges including a large scale span, large variation of the ratio between height and width, and rich character set and mathematical expressions. Considering these challenges, we used Generalized Focal Loss (GFL), an anchor-free method, instead of the anchor-based method, and prove the Adaptive Training Sampling Strategy (ATSS) and proper Feature Pyramid Network (FPN) can well solve the important issue of scale variation. Meanwhile, we also found some tricks, e.g., Deformable Convolution Network (DCN), SyncBN, and Weighted Box Fusion (WBF), were effective in MFD task. Our proposed method ranked 1st in the final 15 teams.

Yixuan Qiao, Hao Chen, Jun Wang et al.

TextVQA requires models to read and reason about text in images to answer questions about them. Specifically, models need to incorporate a new modality of text present in the images and reason over it to answer TextVQA questions. In this challenge, we use generative model T5 for TextVQA task. Based on pre-trained checkpoint T5-3B from HuggingFace repository, two other pre-training tasks including masked language modeling(MLM) and relative position prediction(RPP) are designed to better align object feature and scene text. In the stage of pre-training, encoder is dedicate to handle the fusion among multiple modalities: question text, object text labels, scene text labels, object visual features, scene visual features. After that decoder generates the text sequence step-by-step, cross entropy loss is required by default. We use a large-scale scene text dataset in pre-training and then fine-tune the T5-3B with the TextVQA dataset only.

Jiaquan Ye, Xianbiao Qi, Yelin He et al.

This paper presents our solution for ICDAR 2021 competition on scientific literature parsing taskB: table recognition to HTML. In our method, we divide the table content recognition task into foursub-tasks: table structure recognition, text line detection, text line recognition, and box assignment.Our table structure recognition algorithm is customized based on MASTER [1], a robust image textrecognition algorithm. PSENet [2] is used to detect each text line in the table image. For text linerecognition, our model is also built on MASTER. Finally, in the box assignment phase, we associatedthe text boxes detected by PSENet with the structure item reconstructed by table structure prediction,and fill the recognized content of the text line into the corresponding item. Our proposed methodachieves a 96.84% TEDS score on 9,115 validation samples in the development phase, and a 96.32%TEDS score on 9,064 samples in the final evaluation phase.

Yelin He, Xianbiao Qi, Jiaquan Ye et al.

This paper presents our solution for the ICDAR 2021 Competition on Scientific Table Image Recognition to LaTeX. This competition has two sub-tasks: Table Structure Reconstruction (TSR) and Table Content Reconstruction (TCR). We treat both sub-tasks as two individual image-to-sequence recognition problems. We leverage our previously proposed algorithm MASTER \cite{lu2019master}, which is originally proposed for scene text recognition. We optimize the MASTER model from several perspectives: network structure, optimizer, normalization method, pre-trained model, resolution of input image, data augmentation, and model ensemble. Our method achieves 0.7444 Exact Match and 0.8765 Exact Match @95\% on the TSR task, and obtains 0.5586 Exact Match and 0.7386 Exact Match 95\% on the TCR task.

Yihao Chen, Alexander Lerch

Automatic lyrics generation has received attention from both music and AI communities for years. Early rule-based approaches have~---due to increases in computational power and evolution in data-driven models---~mostly been replaced with deep-learning-based systems. Many existing approaches, however, either rely heavily on prior knowledge in music and lyrics writing or oversimplify the task by largely discarding melodic information and its relationship with the text. We propose an end-to-end melody-conditioned lyrics generation system based on Sequence Generative Adversarial Networks (SeqGAN), which generates a line of lyrics given the corresponding melody as the input. Furthermore, we investigate the performance of the generator with an additional input condition: the theme or overarching topic of the lyrics to be generated. We show that the input conditions have no negative impact on the evaluation metrics while enabling the network to produce more meaningful results.

Bingcong Li, Xin Tang, Xianbiao Qi et al.

Recently, inspired by Transformer, self-attention-based scene text recognition approaches have achieved outstanding performance. However, we find that the size of model expands rapidly with the lexicon increasing. Specifically, the number of parameters for softmax classification layer and output embedding layer are proportional to the vocabulary size. It hinders the development of a lightweight text recognition model especially applied for Chinese and multiple languages. Thus, we propose a lightweight scene text recognition model named Hamming OCR. In this model, a novel Hamming classifier, which adopts locality sensitive hashing (LSH) algorithm to encode each character, is proposed to replace the softmax regression and the generated LSH code is directly employed to replace the output embedding. We also present a simplified transformer decoder to reduce the number of parameters by removing the feed-forward network and using cross-layer parameter sharing technique. Compared with traditional methods, the number of parameters in both classification and embedding layers is independent on the size of vocabulary, which significantly reduces the storage requirement without loss of accuracy. Experimental results on several datasets, including four public benchmaks and a Chinese text dataset synthesized by SynthText with more than 20,000 characters, shows that Hamming OCR achieves competitive results.

Di Wu, Yihao Chen, Xianbiao Qi et al.

How can we effectively utilise the 2D monocular image information for recovering the 6D pose (6-DoF) of the visual objects? Deep learning has shown to be effective for robust and real-time monocular pose estimation. Oftentimes, the network learns to regress the 6-DoF pose using a naive loss function. However, due to a lack of geometrical scene understanding from the directly regressed pose estimation, there are misalignments between the rendered mesh from the 3D object and the 2D instance segmentation result, e.g., bounding boxes and masks prediction. This paper bridges the gap between 2D mask generation and 3D location prediction via a differentiable neural mesh renderer. We utilise the overlay between the accurate mask prediction and less accurate mesh prediction to iteratively optimise the direct regressed 6D pose information with a focus on translation estimation. By leveraging geometry, we demonstrate that our technique significantly improves direct regression performance on the difficult task of translation estimation and achieve the state of the art results on Peking University/Baidu - Autonomous Driving dataset and the ApolloScape 3D Car Instance dataset. The code can be found at \url{https://bit.ly/2IRihfU}.