Yang Gao

Jinkun Cao, Ruiqian Nai, Qing Yang et al. · cmu

Negative-free contrastive learning methods have attracted a lot of attention with simplicity and impressive performances for large-scale pretraining. However, its disentanglement property remains unexplored. In this paper, we examine negative-free contrastive learning methods to study the disentanglement property empirically. We find that existing disentanglement metrics fail to make meaningful measurements for high-dimensional representation models, so we propose a new disentanglement metric based on Mutual Information between latent representations and data factors. With this proposed metric, we benchmark the disentanglement property of negative-free contrastive learning on both popular synthetic datasets and a real-world dataset CelebA. Our study shows that the investigated methods can learn a well-disentangled subset of representation. As far as we know, we are the first to extend the study of disentangled representation learning to high-dimensional representation space and introduce negative-free contrastive learning methods into this area. The source code of this paper is available at \url{https://github.com/noahcao/disentanglement_lib_med}.

Wenbin Li, Zhichen Fan, Jing Huo et al.

Novel class discovery (NCD) aims at learning a model that transfers the common knowledge from a class-disjoint labelled dataset to another unlabelled dataset and discovers new classes (clusters) within it. Many methods, as well as elaborate training pipelines and appropriate objectives, have been proposed and considerably boosted performance on NCD tasks. Despite all this, we find that the existing methods do not sufficiently take advantage of the essence of the NCD setting. To this end, in this paper, we propose to model both inter-class and intra-class constraints in NCD based on the symmetric Kullback-Leibler divergence (sKLD). Specifically, we propose an inter-class sKLD constraint to effectively exploit the disjoint relationship between labelled and unlabelled classes, enforcing the separability for different classes in the embedding space. In addition, we present an intra-class sKLD constraint to explicitly constrain the intra-relationship between a sample and its augmentations and ensure the stability of the training process at the same time. We conduct extensive experiments on the popular CIFAR10, CIFAR100 and ImageNet benchmarks and successfully demonstrate that our method can establish a new state of the art and can achieve significant performance improvements, e.g., 3.5%/3.7% clustering accuracy improvements on CIFAR100-50 dataset split under the task-aware/-agnostic evaluation protocol, over previous state-of-the-art methods. Code is available at https://github.com/FanZhichen/NCD-IIC.

Yingdong Hu, Renhao Wang, Li Erran Li et al.

In recent years, increasing attention has been directed to leveraging pre-trained vision models for motor control. While existing works mainly emphasize the importance of this pre-training phase, the arguably equally important role played by downstream policy learning during control-specific fine-tuning is often neglected. It thus remains unclear if pre-trained vision models are consistent in their effectiveness under different control policies. To bridge this gap in understanding, we conduct a comprehensive study on 14 pre-trained vision models using 3 distinct classes of policy learning methods, including reinforcement learning (RL), imitation learning through behavior cloning (BC), and imitation learning with a visual reward function (VRF). Our study yields a series of intriguing results, including the discovery that the effectiveness of pre-training is highly dependent on the choice of the downstream policy learning algorithm. We show that conventionally accepted evaluation based on RL methods is highly variable and therefore unreliable, and further advocate for using more robust methods like VRF and BC. To facilitate more universal evaluations of pre-trained models and their policy learning methods in the future, we also release a benchmark of 21 tasks across 3 different environments alongside our work.

Zhecheng Yuan, Zhengrong Xue, Bo Yuan et al.

Learning generalizable policies that can adapt to unseen environments remains challenging in visual Reinforcement Learning (RL). Existing approaches try to acquire a robust representation via diversifying the appearances of in-domain observations for better generalization. Limited by the specific observations of the environment, these methods ignore the possibility of exploring diverse real-world image datasets. In this paper, we investigate how a visual RL agent would benefit from the off-the-shelf visual representations. Surprisingly, we find that the early layers in an ImageNet pre-trained ResNet model could provide rather generalizable representations for visual RL. Hence, we propose Pre-trained Image Encoder for Generalizable visual reinforcement learning (PIE-G), a simple yet effective framework that can generalize to the unseen visual scenarios in a zero-shot manner. Extensive experiments are conducted on DMControl Generalization Benchmark, DMControl Manipulation Tasks, Drawer World, and CARLA to verify the effectiveness of PIE-G. Empirical evidence suggests PIE-G improves sample efficiency and significantly outperforms previous state-of-the-art methods in terms of generalization performance. In particular, PIE-G boasts a 55% generalization performance gain on average in the challenging video background setting. Project Page: https://sites.google.com/view/pie-g/home.

Jialei Huang, Zhaoheng Yin, Yingdong Hu et al.

Adversarial imitation learning (AIL) is a popular method that has recently achieved much success. However, the performance of AIL is still unsatisfactory on the more challenging tasks. We find that one of the major reasons is due to the low quality of AIL discriminator representation. Since the AIL discriminator is trained via binary classification that does not necessarily discriminate the policy from the expert in a meaningful way, the resulting reward might not be meaningful either. We propose a new method called Policy Contrastive Imitation Learning (PCIL) to resolve this issue. PCIL learns a contrastive representation space by anchoring on different policies and generates a smooth cosine-similarity-based reward. Our proposed representation learning objective can be viewed as a stronger version of the AIL objective and provide a more meaningful comparison between the agent and the policy. From a theoretical perspective, we show the validity of our method using the apprenticeship learning framework. Furthermore, our empirical evaluation on the DeepMind Control suite demonstrates that PCIL can achieve state-of-the-art performance. Finally, qualitative results suggest that PCIL builds a smoother and more meaningful representation space for imitation learning.

Chuan Wen, Jianing Qian, Jierui Lin et al.

Across applications spanning supervised classification and sequential control, deep learning has been reported to find "shortcut" solutions that fail catastrophically under minor changes in the data distribution. In this paper, we show empirically that DNNs can be coaxed to avoid poor shortcuts by providing an additional "priming" feature computed from key input features, usually a coarse output estimate. Priming relies on approximate domain knowledge of these task-relevant key input features, which is often easy to obtain in practical settings. For example, one might prioritize recent frames over past frames in a video input for visual imitation learning, or salient foreground over background pixels for image classification. On NICO image classification, MuJoCo continuous control, and CARLA autonomous driving, our priming strategy works significantly better than several popular state-of-the-art approaches for feature selection and data augmentation. We connect these empirical findings to recent theoretical results on DNN optimization, and argue theoretically that priming distracts the optimizer away from poor shortcuts by creating better, simpler shortcuts.

Kaifeng Zhang, Rui Zhao, Ziming Zhang et al.

Reinforcement learning (RL) provides a powerful framework for decision-making, but its application in practice often requires a carefully designed reward function. Adversarial Imitation Learning (AIL) sheds light on automatic policy acquisition without access to the reward signal from the environment. In this work, we propose Auto-Encoding Adversarial Imitation Learning (AEAIL), a robust and scalable AIL framework. To induce expert policies from demonstrations, AEAIL utilizes the reconstruction error of an auto-encoder as a reward signal, which provides more information for optimizing policies than the prior discriminator-based ones. Subsequently, we use the derived objective functions to train the auto-encoder and the agent policy. Experiments show that our AEAIL performs superior compared to state-of-the-art methods on both state and image based environments. More importantly, AEAIL shows much better robustness when the expert demonstrations are noisy.

Huashan Sun, Shengyi Liao, Yansen Han et al.

Despite advances in pretraining with extended context sizes, large language models (LLMs) still face challenges in effectively utilizing real-world long-context information, primarily due to insufficient long-context alignment caused by data quality issues, training inefficiencies, and the lack of well-designed optimization objectives. To address these limitations, we propose a framework named \textbf{S}h\textbf{o}rt-to-\textbf{Lo}ng \textbf{P}reference \textbf{O}ptimization (\textbf{SoLoPO}), decoupling long-context preference optimization (PO) into two components: short-context PO and short-to-long reward alignment (SoLo-RA), supported by both theoretical and empirical evidence. Specifically, short-context PO leverages preference pairs sampled from short contexts to enhance the model's contextual knowledge utilization ability. Meanwhile, SoLo-RA explicitly encourages reward score consistency for the responses when conditioned on both short and long contexts that contain identical task-relevant information. This facilitates transferring the model's ability to handle short contexts into long-context scenarios. SoLoPO is compatible with mainstream preference optimization algorithms, while substantially improving the efficiency of data construction and training processes. Experimental results show that SoLoPO enhances all these algorithms with respect to stronger length and domain generalization abilities across various long-context benchmarks, while achieving notable improvements in both computational and memory efficiency.

Chia-Chi Chuang, Donglin Yang, Chuan Wen et al.

Imitation learning is a widely used policy learning method that enables intelligent agents to acquire complex skills from expert demonstrations. The input to the imitation learning algorithm is usually composed of both the current observation and historical observations since the most recent observation might not contain enough information. This is especially the case with image observations, where a single image only includes one view of the scene, and it suffers from a lack of motion information and object occlusions. In theory, providing multiple observations to the imitation learning agent will lead to better performance. However, surprisingly people find that sometimes imitation from observation histories performs worse than imitation from the most recent observation. In this paper, we explain this phenomenon from the information flow within the neural network perspective. We also propose a novel imitation learning neural network architecture that does not suffer from this issue by design. Furthermore, our method scales to high-dimensional image observations. Finally, we benchmark our approach on two widely used simulators, CARLA and MuJoCo, and it successfully alleviates the copycat problem and surpasses the existing solutions.

Kaizhe Hu, Ray Chen Zheng, Yang Gao et al.

Controlling agents remotely with deep reinforcement learning~(DRL) in the real world is yet to come. One crucial stepping stone is to devise RL algorithms that are robust in the face of dropped information from corrupted communication or malfunctioning sensors. Typical RL methods usually require considerable online interaction data that are costly and unsafe to collect in the real world. Furthermore, when applying to the frame dropping scenarios, they perform unsatisfactorily even with moderate drop rates. To address these issues, we propose Decision Transformer under Random Frame Dropping~(DeFog), an offline RL algorithm that enables agents to act robustly in frame dropping scenarios without online interaction. DeFog first randomly masks out data in the offline datasets and explicitly adds the time span of frame dropping as inputs. After that, a finetuning stage on the same offline dataset with a higher mask rate would further boost the performance. Empirical results show that DeFog outperforms strong baselines under severe frame drop rates like 90\%, while maintaining similar returns under non-frame-dropping conditions in the regular MuJoCo control benchmarks and the Atari environments. Our approach offers a robust and deployable solution for controlling agents in real-world environments with limited or unreliable data.

Wenbin Li, Meihao Kong, Xuesong Yang et al.

In recent years, a variety of contrastive learning based unsupervised visual representation learning methods have been designed and achieved great success in many visual tasks. Generally, these methods can be roughly classified into four categories: (1) standard contrastive methods with an InfoNCE like loss, such as MoCo and SimCLR; (2) non-contrastive methods with only positive pairs, such as BYOL and SimSiam; (3) whitening regularization based methods, such as W-MSE and VICReg; and (4) consistency regularization based methods, such as CO2. In this study, we present a new unified contrastive learning representation framework (named UniCLR) suitable for all the above four kinds of methods from a novel perspective of basic affinity matrix. Moreover, three variants, i.e., SimAffinity, SimWhitening and SimTrace, are presented based on UniCLR. In addition, a simple symmetric loss, as a new consistency regularization term, is proposed based on this framework. By symmetrizing the affinity matrix, we can effectively accelerate the convergence of the training process. Extensive experiments have been conducted to show that (1) the proposed UniCLR framework can achieve superior results on par with and even be better than the state of the art, (2) the proposed symmetric loss can significantly accelerate the convergence of models, and (3) SimTrace can avoid the mode collapse problem by maximizing the trace of a whitened affinity matrix without relying on asymmetry designs or stop-gradients.

Xiaochen Liu, Yang Gao, Yu Bai et al.

Few-shot abstractive summarization has become a challenging task in natural language generation. To support it, we designed a novel soft prompts architecture coupled with a prompt pre-training plus fine-tuning paradigm that is effective and tunes only extremely light parameters. The soft prompts include continuous input embeddings across an encoder and a decoder to fit the structure of the generation models. Importantly, a novel inner-prompt placed in the text is introduced to capture document-level information. The aim is to devote attention to understanding the document that better prompts the model to generate document-related content. The first step in the summarization procedure is to conduct prompt pre-training with self-supervised pseudo-data. This teaches the model basic summarizing capabilities. The model is then fine-tuned with few-shot examples. Experimental results on the CNN/DailyMail and XSum datasets show that our method, with only 0.1% of the parameters, outperforms full-model tuning where all model parameters are tuned. It also surpasses Prompt Tuning by a large margin and delivers competitive results against Prefix-Tuning with 3% of the parameters.

Zichuan Liu, Yuanyang Zhu, Zhi Wang et al.

While achieving tremendous success in various fields, existing multi-agent reinforcement learning (MARL) with a black-box neural network makes decisions in an opaque manner that hinders humans from understanding the learned knowledge and how input observations influence decisions. In contrast, existing interpretable approaches usually suffer from weak expressivity and low performance. To bridge this gap, we propose MIXing Recurrent soft decision Trees (MIXRTs), a novel interpretable architecture that can represent explicit decision processes via the root-to-leaf path and reflect each agent's contribution to the team. Specifically, we construct a novel soft decision tree using a recurrent structure and demonstrate which features influence the decision-making process. Then, based on the value decomposition framework, we linearly assign credit to each agent by explicitly mixing individual action values to estimate the joint action value using only local observations, providing new insights into interpreting the cooperation mechanism. Theoretical analysis confirms that MIXRTs guarantee additivity and monotonicity in the factorization of joint action values. Evaluations on complex tasks like Spread and StarCraft II demonstrate that MIXRTs compete with existing methods while providing clear explanations, paving the way for interpretable and high-performing MARL systems.

Weirui Ye, Yunsheng Zhang, Haoyang Weng et al.

Reinforcement learning (RL) is a promising approach for solving robotic manipulation tasks. However, it is challenging to apply the RL algorithms directly in the real world. For one thing, RL is data-intensive and typically requires millions of interactions with environments, which are impractical in real scenarios. For another, it is necessary to make heavy engineering efforts to design reward functions manually. To address these issues, we leverage foundation models in this paper. We propose Reinforcement Learning with Foundation Priors (RLFP) to utilize guidance and feedback from policy, value, and success-reward foundation models. Within this framework, we introduce the Foundation-guided Actor-Critic (FAC) algorithm, which enables embodied agents to explore more efficiently with automatic reward functions. The benefits of our framework are threefold: (1) \textit{sample efficient}; (2) \textit{minimal and effective reward engineering}; (3) \textit{agnostic to foundation model forms and robust to noisy priors}. Our method achieves remarkable performances in various manipulation tasks on both real robots and in simulation. Across 5 dexterous tasks with real robots, FAC achieves an average success rate of 86\% after one hour of real-time learning. Across 8 tasks in the simulated Meta-world, FAC achieves 100\% success rates in 7/8 tasks under less than 100k frames (about 1-hour training), outperforming baseline methods with manual-designed rewards in 1M frames. We believe the RLFP framework can enable future robots to explore and learn autonomously in the physical world for more tasks. Visualizations and code are available at \url{https://yewr.github.io/rlfp}.

Kun Lei, Zhengmao He, Chenhao Lu et al.

Combining offline and online reinforcement learning (RL) is crucial for efficient and safe learning. However, previous approaches treat offline and online learning as separate procedures, resulting in redundant designs and limited performance. We ask: Can we achieve straightforward yet effective offline and online learning without introducing extra conservatism or regularization? In this study, we propose Uni-o4, which utilizes an on-policy objective for both offline and online learning. Owning to the alignment of objectives in two phases, the RL agent can transfer between offline and online learning seamlessly. This property enhances the flexibility of the learning paradigm, allowing for arbitrary combinations of pretraining, fine-tuning, offline, and online learning. In the offline phase, specifically, Uni-o4 leverages diverse ensemble policies to address the mismatch issues between the estimated behavior policy and the offline dataset. Through a simple offline policy evaluation (OPE) approach, Uni-o4 can achieve multi-step policy improvement safely. We demonstrate that by employing the method above, the fusion of these two paradigms can yield superior offline initialization as well as stable and rapid online fine-tuning capabilities. Through real-world robot tasks, we highlight the benefits of this paradigm for rapid deployment in challenging, previously unseen real-world environments. Additionally, through comprehensive evaluations using numerous simulated benchmarks, we substantiate that our method achieves state-of-the-art performance in both offline and offline-to-online fine-tuning learning. Our website: https://lei-kun.github.io/uni-o4/ .

Jiaheng Liu, Dawei Zhu, Zhiqi Bai et al. · pku

Efficient processing of long contexts has been a persistent pursuit in Natural Language Processing. With the growing number of long documents, dialogues, and other textual data, it is important to develop Long Context Language Models (LCLMs) that can process and analyze extensive inputs in an effective and efficient way. In this paper, we present a comprehensive survey on recent advances in long-context modeling for large language models. Our survey is structured around three key aspects: how to obtain effective and efficient LCLMs, how to train and deploy LCLMs efficiently, and how to evaluate and analyze LCLMs comprehensively. For the first aspect, we discuss data strategies, architectural designs, and workflow approaches oriented with long context processing. For the second aspect, we provide a detailed examination of the infrastructure required for LCLM training and inference. For the third aspect, we present evaluation paradigms for long-context comprehension and long-form generation, as well as behavioral analysis and mechanism interpretability of LCLMs. Beyond these three key aspects, we thoroughly explore the diverse application scenarios where existing LCLMs have been deployed and outline promising future development directions. This survey provides an up-to-date review of the literature on long-context LLMs, which we wish to serve as a valuable resource for both researchers and engineers. An associated GitHub repository collecting the latest papers and repos is available at: \href{https://github.com/LCLM-Horizon/A-Comprehensive-Survey-For-Long-Context-Language-Modeling}{\color[RGB]{175,36,67}{LCLM-Horizon}}.

Hongye Cao, Zhixin Bai, Ziyue Peng et al.

Reinforcement learning with verifiable rewards (RLVR) has demonstrated superior performance in enhancing the reasoning capability of large language models (LLMs). However, this accuracy-oriented learning paradigm often suffers from entropy collapse, which reduces policy exploration and limits reasoning capabilities. To address this challenge, we propose an efficient reinforcement learning framework that leverages entropy signals at both the semantic and token levels to improve reasoning. From the data perspective, we introduce semantic entropy-guided curriculum learning, organizing training data from low to high semantic entropy to guide progressive optimization from easier to more challenging tasks. For the algorithmic design, we adopt non-uniform token treatment by imposing KL regularization on low-entropy tokens that critically impact policy exploration and applying stronger constraints on high-covariance portions within these tokens. By jointly optimizing data organization and algorithmic design, our method effectively mitigates entropy collapse and enhances LLM reasoning. Experimental results across 6 benchmarks with 3 different parameter-scale base models demonstrate that our method outperforms other entropy-based approaches in improving reasoning.

Saeed Saadatnejad, Yang Gao, Kaouther Messaoud et al.

Accurate human trajectory prediction is crucial for applications such as autonomous vehicles, robotics, and surveillance systems. Yet, existing models often fail to fully leverage the non-verbal social cues human subconsciously communicate when navigating the space. To address this, we introduce Social-Transmotion, a generic Transformer-based model that exploits diverse and numerous visual cues to predict human behavior. We translate the idea of a prompt from Natural Language Processing (NLP) to the task of human trajectory prediction, where a prompt can be a sequence of x-y coordinates on the ground, bounding boxes in the image plane, or body pose keypoints in either 2D or 3D. This, in turn, augments trajectory data, leading to enhanced human trajectory prediction. Using masking technique, our model exhibits flexibility and adaptability by capturing spatiotemporal interactions between agents based on the available visual cues. We delve into the merits of using 2D versus 3D poses, and a limited set of poses. Additionally, we investigate the spatial and temporal attention map to identify which keypoints and time-steps in the sequence are vital for optimizing human trajectory prediction. Our approach is validated on multiple datasets, including JTA, JRDB, Pedestrians and Cyclists in Road Traffic, and ETH-UCY. The code is publicly available: https://github.com/vita-epfl/social-transmotion.

Shengjie Wang, Fengbo Lan, Xiang Zheng et al.

In current model-free reinforcement learning (RL) algorithms, stability criteria based on sampling methods are commonly utilized to guide policy optimization. However, these criteria only guarantee the infinite-time convergence of the system's state to an equilibrium point, which leads to sub-optimality of the policy. In this paper, we propose a policy optimization technique incorporating sampling-based Lyapunov stability. Our approach enables the system's state to reach an equilibrium point within an optimal time and maintain stability thereafter, referred to as "optimal-time stability". To achieve this, we integrate the optimization method into the Actor-Critic framework, resulting in the development of the Adaptive Lyapunov-based Actor-Critic (ALAC) algorithm. Through evaluations conducted on ten robotic tasks, our approach outperforms previous studies significantly, effectively guiding the system to generate stable patterns.

Yuquan Xie, Shengtao Peng, Wanqi Yang et al.

Knowledge Tracing (KT) is a critical task in online education systems, aiming to monitor students' knowledge states throughout a learning period. Common KT approaches involve predicting the probability of a student correctly answering the next question based on their exercise history. However, these methods often suffer from performance degradation when faced with the scarcity of student interactions in new education systems. To address this, we leverage student interactions from existing education systems to mitigate performance degradation caused by limited training data. Nevertheless, these interactions exhibit significant differences since they are derived from different education systems. To address this issue, we propose a domain generalization approach for knowledge tracing, where existing education systems are considered source domains, and new education systems with limited data are considered target domains. Additionally, we design a domain-generalizable knowledge tracing framework (DGKT) that can be applied to any KT model. Specifically, we present a concept aggregation approach designed to reduce conceptual disparities within sequences of student interactions from diverse domains. To further mitigate domain discrepancies, we introduce a novel normalization module called Sequence Instance Normalization (SeqIN). Moreover, to fully leverage exercise information, we propose a new knowledge tracing model tailored for the domain generalization KT task, named Domain-Generalizable Relation-based Knowledge Tracing (DGRKT). Extensive experiments across five benchmark datasets demonstrate that the proposed method performs well despite limited training data.

Xiao Liu, Shuyang Liu, Wenbin Li et al.

Although deep reinforcement learning has become a universal solution for complex control tasks, its real-world applicability is still limited because lacking security guarantees for policies. To address this problem, we propose Boundary Characterization via the Minimum Experience Retention (BCMER), an end-to-end Interpretable Policy Distillation (IPD) framework. Unlike previous IPD approaches, BCMER distinguishes the importance of experiences and keeps a minimal but critical experience pool with almost no loss of policy similarity. Specifically, the proposed BCMER contains two basic steps. Firstly, we propose a novel multidimensional hyperspheres intersection (MHI) approach to divide experience points into boundary points and internal points, and reserve the crucial boundary points. Secondly, we develop a nearest-neighbor-based model to generate robust and interpretable decision rules based on the boundary points. Extensive experiments show that the proposed BCMER is able to reduce the amount of experience to 1.4%~19.1% (when the count of the naive experiences is 10k) and maintain high IPD performance. In general, the proposed BCMER is more suitable for the experience storage limited regime because it discovers the critical experience and eliminates redundant experience.

Yang Gao, Po-Chien Luan, Alexandre Alahi

The ability of intelligent systems to predict human behaviors is crucial, particularly in fields such as autonomous vehicle navigation and social robotics. However, the complexity of human motion have prevented the development of a standardized dataset for human motion prediction, thereby hindering the establishment of pre-trained models. In this paper, we address these limitations by integrating multiple datasets, encompassing both trajectory and 3D pose keypoints, to propose a pre-trained model for human motion prediction. We merge seven distinct datasets across varying modalities and standardize their formats. To facilitate multimodal pre-training, we introduce Multi-Transmotion, an innovative transformer-based model designed for cross-modality pre-training. Additionally, we present a novel masking strategy to capture rich representations. Our methodology demonstrates competitive performance across various datasets on several downstream tasks, including trajectory prediction in the NBA and JTA datasets, as well as pose prediction in the AMASS and 3DPW datasets. The code is publicly available: https://github.com/vita-epfl/multi-transmotion

Qinghe Ma, Jian Zhang, Lei Qi et al.

Both limited annotation and domain shift are prevalent challenges in medical image segmentation. Traditional semi-supervised segmentation and unsupervised domain adaptation methods address one of these issues separately. However, the coexistence of limited annotation and domain shift is quite common, which motivates us to introduce a novel and challenging scenario: Mixed Domain Semi-supervised medical image Segmentation (MiDSS). In this scenario, we handle data from multiple medical centers, with limited annotations available for a single domain and a large amount of unlabeled data from multiple domains. We found that the key to solving the problem lies in how to generate reliable pseudo labels for the unlabeled data in the presence of domain shift with labeled data. To tackle this issue, we employ Unified Copy-Paste (UCP) between images to construct intermediate domains, facilitating the knowledge transfer from the domain of labeled data to the domains of unlabeled data. To fully utilize the information within the intermediate domain, we propose a symmetric Guidance training strategy (SymGD), which additionally offers direct guidance to unlabeled data by merging pseudo labels from intermediate samples. Subsequently, we introduce a Training Process aware Random Amplitude MixUp (TP-RAM) to progressively incorporate style-transition components into intermediate samples. Compared with existing state-of-the-art approaches, our method achieves a notable 13.57% improvement in Dice score on Prostate dataset, as demonstrated on three public datasets. Our code is available at https://github.com/MQinghe/MiDSS .

Guang Yang, Tianpei Yang, Jingwen Qiao et al.

Communication is one of the effective means to improve the learning of cooperative policy in multi-agent systems. However, in most real-world scenarios, lossy communication is a prevalent issue. Existing multi-agent reinforcement learning with communication, due to their limited scalability and robustness, struggles to apply to complex and dynamic real-world environments. To address these challenges, we propose a generalized communication-constrained model to uniformly characterize communication conditions across different scenarios. Based on this, we utilize it as a learning prior to distinguish between lossy and lossless messages for specific scenarios. Additionally, we decouple the impact of lossy and lossless messages on distributed decision-making, drawing on a dual mutual information estimatior, and introduce a communication-constrained multi-agent reinforcement learning framework, quantifying the impact of communication messages into the global reward. Finally, we validate the effectiveness of our approach across several communication-constrained benchmarks.

Zihan Cheng, Jintao Guo, Jian Zhang et al.

To segment medical images with distribution shifts, domain generalization (DG) has emerged as a promising setting to train models on source domains that can generalize to unseen target domains. Existing DG methods are mainly based on CNN or ViT architectures. Recently, advanced state space models, represented by Mamba, have shown promising results in various supervised medical image segmentation. The success of Mamba is primarily owing to its ability to capture long-range dependencies while keeping linear complexity with input sequence length, making it a promising alternative to CNNs and ViTs. Inspired by the success, in the paper, we explore the potential of the Mamba architecture to address distribution shifts in DG for medical image segmentation. Specifically, we propose a novel Mamba-based framework, Mamba-Sea, incorporating global-to-local sequence augmentation to improve the model's generalizability under domain shift issues. Our Mamba-Sea introduces a global augmentation mechanism designed to simulate potential variations in appearance across different sites, aiming to suppress the model's learning of domain-specific information. At the local level, we propose a sequence-wise augmentation along input sequences, which perturbs the style of tokens within random continuous sub-sequences by modeling and resampling style statistics associated with domain shifts. To our best knowledge, Mamba-Sea is the first work to explore the generalization of Mamba for medical image segmentation, providing an advanced and promising Mamba-based architecture with strong robustness to domain shifts. Remarkably, our proposed method is the first to surpass a Dice coefficient of 90% on the Prostate dataset, which exceeds previous SOTA of 88.61%. The code is available at https://github.com/orange-czh/Mamba-Sea.

Rui Ye, Keduan Huang, Qimin Wu et al.

LLM-based multi-agent systems (MAS) have demonstrated significant potential in enhancing single LLMs to address complex and diverse tasks in practical applications. Despite considerable advancements, the field lacks a unified codebase that consolidates existing methods, resulting in redundant re-implementation efforts, unfair comparisons, and high entry barriers for researchers. To address these challenges, we introduce MASLab, a unified, comprehensive, and research-friendly codebase for LLM-based MAS. (1) MASLab integrates over 20 established methods across multiple domains, each rigorously validated by comparing step-by-step outputs with its official implementation. (2) MASLab provides a unified environment with various benchmarks for fair comparisons among methods, ensuring consistent inputs and standardized evaluation protocols. (3) MASLab implements methods within a shared streamlined structure, lowering the barriers for understanding and extension. Building on MASLab, we conduct extensive experiments covering 10+ benchmarks and 8 models, offering researchers a clear and comprehensive view of the current landscape of MAS methods. MASLab will continue to evolve, tracking the latest developments in the field, and invite contributions from the broader open-source community.

Jintao Guo, Lei Qi, Yinghuan Shi et al.

Domain generalization (DG) aims to enhance the model robustness against domain shifts without accessing target domains. A prevalent category of methods for DG is data augmentation, which focuses on generating virtual samples to simulate domain shifts. However, existing augmentation techniques in DG are mainly tailored for convolutional neural networks (CNNs), with limited exploration in token-based architectures, i.e., vision transformer (ViT) and multi-layer perceptrons (MLP) models. In this paper, we study the impact of prior CNN-based augmentation methods on token-based models, revealing their performance is suboptimal due to the lack of incentivizing the model to learn holistic shape information. To tackle the issue, we propose the SEmantic-aware Token Augmentation (SETA) method. SETA transforms token features by perturbing local edge cues while preserving global shape features, thereby enhancing the model learning of shape information. To further enhance the generalization ability of the model, we introduce two stylized variants of our method combined with two state-of-the-art style augmentation methods in DG. We provide a theoretical insight into our method, demonstrating its effectiveness in reducing the generalization risk bound. Comprehensive experiments on five benchmarks prove that our method achieves SOTA performances across various ViT and MLP architectures. Our code is available at https://github.com/lingeringlight/SETA.

Saeed Saadatnejad, Yang Gao, Hamid Rezatofighi et al.

Predicting future trajectories is critical in autonomous navigation, especially in preventing accidents involving humans, where a predictive agent's ability to anticipate in advance is of utmost importance. Trajectory forecasting models, employed in fields such as robotics, autonomous vehicles, and navigation, face challenges in real-world scenarios, often due to the isolation of model components. To address this, we introduce a novel dataset for end-to-end trajectory forecasting, facilitating the evaluation of models in scenarios involving less-than-ideal preceding modules such as tracking. This dataset, an extension of the JRDB dataset, provides comprehensive data, including the locations of all agents, scene images, and point clouds, all from the robot's perspective. The objective is to predict the future positions of agents relative to the robot using raw sensory input data. It bridges the gap between isolated models and practical applications, promoting a deeper understanding of navigation dynamics. Additionally, we introduce a novel metric for assessing trajectory forecasting models in real-world scenarios where ground-truth identities are inaccessible, addressing issues related to undetected or over-detected agents. Researchers are encouraged to use our benchmark for model evaluation and benchmarking.

Haishuai Wang, Yang Gao, Xin Zheng et al.

Graph Neural Architecture Search (GNAS) has shown promising results in finding the best graph neural network architecture on a given graph dataset. However, existing GNAS methods still require intensive human labor and rich domain knowledge when designing the search space and search strategy. To this end, we integrate Large Language Models (LLMs) into GNAS and present a new GNAS model based on LLMs (GNAS-LLM for short). The basic idea of GNAS-LLM is to design a new class of GNAS prompts for LLMs to guide LLMs towards understanding the generative task of graph neural architectures. The prompts consist of descriptions of the search space, search strategy, and search feedback of GNAS. By iteratively running LLMs with the prompts, GNAS-LLM generates more accurate graph neural network architectures with fast convergence. Experimental results show that GNAS-LLM outperforms the state-of-the-art GNAS methods on four benchmark graph datasets, with an average improvement of 0.7% on the validation sets and 0.3% on the test sets. Besides, GNAS-LLM achieves an average improvement of 1.0% on the test sets based on the search space from AutoGEL.

Linjian Meng, Youzhi Zhang, Zhenxing Ge et al.

To establish last-iterate convergence for Counterfactual Regret Minimization (CFR) algorithms in learning a Nash equilibrium (NE) of extensive-form games (EFGs), recent studies reformulate learning an NE of the original EFG as learning the NEs of a sequence of (perturbed) regularized EFGs. Consequently, proving last-iterate convergence in solving the original EFG reduces to proving last-iterate convergence in solving (perturbed) regularized EFGs. However, the empirical convergence rates of the algorithms in these studies are suboptimal, since they do not utilize Regret Matching (RM)-based CFR algorithms to solve perturbed EFGs, which are known the exceptionally fast empirical convergence rates. Additionally, since solving multiple perturbed regularized EFGs is required, fine-tuning across all such games is infeasible, making parameter-free algorithms highly desirable. In this paper, we prove that CFR$^+$, a classical parameter-free RM-based CFR algorithm, achieves last-iterate convergence in learning an NE of perturbed regularized EFGs. Leveraging CFR$^+$ to solve perturbed regularized EFGs, we get Reward Transformation CFR$^+$ (RTCFR$^+$). Importantly, we extend prior work on the parameter-free property of CFR$^+$, enhancing its stability, which is crucial for the empirical convergence of RTCFR$^+$. Experiments show that RTCFR$^+$ significantly outperforms existing algorithms with theoretical last-iterate convergence guarantees.

Chuan Wen, Xingyu Lin, John So et al.

Learning from demonstration is a powerful method for teaching robots new skills, and having more demonstration data often improves policy learning. However, the high cost of collecting demonstration data is a significant bottleneck. Videos, as a rich data source, contain knowledge of behaviors, physics, and semantics, but extracting control-specific information from them is challenging due to the lack of action labels. In this work, we introduce a novel framework, Any-point Trajectory Modeling (ATM), that utilizes video demonstrations by pre-training a trajectory model to predict future trajectories of arbitrary points within a video frame. Once trained, these trajectories provide detailed control guidance, enabling the learning of robust visuomotor policies with minimal action-labeled data. Across over 130 language-conditioned tasks we evaluated in both simulation and the real world, ATM outperforms strong video pre-training baselines by 80% on average. Furthermore, we show effective transfer learning of manipulation skills from human videos and videos from a different robot morphology. Visualizations and code are available at: \url{https://xingyu-lin.github.io/atm}.

Weirui Ye, Shaohuai Liu, Thanard Kurutach et al.

Reinforcement learning has achieved great success in many applications. However, sample efficiency remains a key challenge, with prominent methods requiring millions (or even billions) of environment steps to train. Recently, there has been significant progress in sample efficient image-based RL algorithms; however, consistent human-level performance on the Atari game benchmark remains an elusive goal. We propose a sample efficient model-based visual RL algorithm built on MuZero, which we name EfficientZero. Our method achieves 194.3% mean human performance and 109.0% median performance on the Atari 100k benchmark with only two hours of real-time game experience and outperforms the state SAC in some tasks on the DMControl 100k benchmark. This is the first time an algorithm achieves super-human performance on Atari games with such little data. EfficientZero's performance is also close to DQN's performance at 200 million frames while we consume 500 times less data. EfficientZero's low sample complexity and high performance can bring RL closer to real-world applicability. We implement our algorithm in an easy-to-understand manner and it is available at https://github.com/YeWR/EfficientZero. We hope it will accelerate the research of MCTS-based RL algorithms in the wider community.

Wenbin Li, Ziyi, Wang et al.

Few-shot learning, especially few-shot image classification, has received increasing attention and witnessed significant advances in recent years. Some recent studies implicitly show that many generic techniques or ``tricks'', such as data augmentation, pre-training, knowledge distillation, and self-supervision, may greatly boost the performance of a few-shot learning method. Moreover, different works may employ different software platforms, backbone architectures and input image sizes, making fair comparisons difficult and practitioners struggle with reproducibility. To address these situations, we propose a comprehensive library for few-shot learning (LibFewShot) by re-implementing eighteen state-of-the-art few-shot learning methods in a unified framework with the same single codebase in PyTorch. Furthermore, based on LibFewShot, we provide comprehensive evaluations on multiple benchmarks with various backbone architectures to evaluate common pitfalls and effects of different training tricks. In addition, with respect to the recent doubts on the necessity of meta- or episodic-training mechanism, our evaluation results confirm that such a mechanism is still necessary especially when combined with pre-training. We hope our work can not only lower the barriers for beginners to enter the area of few-shot learning but also elucidate the effects of nontrivial tricks to facilitate intrinsic research on few-shot learning. The source code is available from https://github.com/RL-VIG/LibFewShot.

Wenbin Li, Xuesong Yang, Meihao Kong et al.

Contrastive self-supervised learning (SSL) methods, such as MoCo and SimCLR, have achieved great success in unsupervised visual representation learning. They rely on a large number of negative pairs and thus require either large memory banks or large batches. Some recent non-contrastive SSL methods, such as BYOL and SimSiam, attempt to discard negative pairs and have also shown remarkable performance. To avoid collapsed solutions caused by not using negative pairs, these methods require non-trivial asymmetry designs. However, in small data regimes, we can not obtain a sufficient number of negative pairs or effectively avoid the over-fitting problem when negatives are not used at all. To address this situation, we argue that negative pairs are still important but one is generally sufficient for each positive pair. We show that a simple Triplet-based loss (Trip) can achieve surprisingly good performance without requiring large batches or asymmetry designs. Moreover, to alleviate the over-fitting problem in small data regimes and further enhance the effect of Trip, we propose a simple plug-and-play RandOm MApping (ROMA) strategy by randomly mapping samples into other spaces and requiring these randomly projected samples to satisfy the same relationship indicated by the triplets. Integrating the triplet-based loss with random mapping, we obtain the proposed method Trip-ROMA. Extensive experiments, including unsupervised representation learning and unsupervised few-shot learning, have been conducted on ImageNet-1K and seven small datasets. They successfully demonstrate the effectiveness of Trip-ROMA and consistently show that ROMA can further effectively boost other SSL methods. Code is available at https://github.com/WenbinLee/Trip-ROMA.

Wen Ji, Kelei He, Jing Huo et al.

Caricature attributes provide distinctive facial features to help research in Psychology and Neuroscience. However, unlike the facial photo attribute datasets that have a quantity of annotated images, the annotations of caricature attributes are rare. To facility the research in attribute learning of caricatures, we propose a caricature attribute dataset, namely WebCariA. Moreover, to utilize models that trained by face attributes, we propose a novel unsupervised domain adaptation framework for cross-modality (i.e., photos to caricatures) attribute recognition, with an integrated inter- and intra-domain consistency learning scheme. Specifically, the inter-domain consistency learning scheme consisting an image-to-image translator to first fill the domain gap between photos and caricatures by generating intermediate image samples, and a label consistency learning module to align their semantic information. The intra-domain consistency learning scheme integrates the common feature consistency learning module with a novel attribute-aware attention-consistency learning module for a more efficient alignment. We did an extensive ablation study to show the effectiveness of the proposed method. And the proposed method also outperforms the state-of-the-art methods by a margin. The implementation of the proposed method is available at https://github.com/KeleiHe/DAAN.

Tiexin Qin, Wenbin Li, Yinghuan Shi et al.

Few-shot learning aims to learn a new concept when only a few training examples are available, which has been extensively explored in recent years. However, most of the current works heavily rely on a large-scale labeled auxiliary set to train their models in an episodic-training paradigm. Such a kind of supervised setting basically limits the widespread use of few-shot learning algorithms. Instead, in this paper, we develop a novel framework called Unsupervised Few-shot Learning via Distribution Shift-based Data Augmentation (ULDA), which pays attention to the distribution diversity inside each constructed pretext few-shot task when using data augmentation. Importantly, we highlight the value and importance of the distribution diversity in the augmentation-based pretext few-shot tasks, which can effectively alleviate the overfitting problem and make the few-shot model learn more robust feature representations. In ULDA, we systemically investigate the effects of different augmentation techniques and propose to strengthen the distribution diversity (or difference) between the query set and support set in each few-shot task, by augmenting these two sets diversely (i.e., distribution shifting). In this way, even incorporated with simple augmentation techniques (e.g., random crop, color jittering, or rotation), our ULDA can produce a significant improvement. In the experiments, few-shot models learned by ULDA can achieve superior generalization performance and obtain state-of-the-art results in a variety of established few-shot learning tasks on Omniglot and miniImageNet. The source code is available in https://github.com/WonderSeven/ULDA.

Wenbin Li, Lei Wang, Xingxing Zhang et al.

This paper investigates a new challenging problem called defensive few-shot learning in order to learn a robust few-shot model against adversarial attacks. Simply applying the existing adversarial defense methods to few-shot learning cannot effectively solve this problem. This is because the commonly assumed sample-level distribution consistency between the training and test sets can no longer be met in the few-shot setting. To address this situation, we develop a general defensive few-shot learning (DFSL) framework to answer the following two key questions: (1) how to transfer adversarial defense knowledge from one sample distribution to another? (2) how to narrow the distribution gap between clean and adversarial examples under the few-shot setting? To answer the first question, we propose an episode-based adversarial training mechanism by assuming a task-level distribution consistency to better transfer the adversarial defense knowledge. As for the second question, within each few-shot task, we design two kinds of distribution consistency criteria to narrow the distribution gap between clean and adversarial examples from the feature-wise and prediction-wise perspectives, respectively. Extensive experiments demonstrate that the proposed framework can effectively make the existing few-shot models robust against adversarial attacks. Code is available at https://github.com/WenbinLee/DefensiveFSL.git.

Florian Böhm, Yang Gao, Christian M. Meyer et al.

Reinforcement Learning (RL) based document summarisation systems yield state-of-the-art performance in terms of ROUGE scores, because they directly use ROUGE as the rewards during training. However, summaries with high ROUGE scores often receive low human judgement. To find a better reward function that can guide RL to generate human-appealing summaries, we learn a reward function from human ratings on 2,500 summaries. Our reward function only takes the document and system summary as input. Hence, once trained, it can be used to train RL-based summarisation systems without using any reference summaries. We show that our learned rewards have significantly higher correlation with human ratings than previous approaches. Human evaluation experiments show that, compared to the state-of-the-art supervised-learning systems and ROUGE-as-rewards RL summarisation systems, the RL systems using our learned rewards during training generate summarieswith higher human ratings. The learned reward function and our source code are available at https://github.com/yg211/summary-reward-no-reference.

Wenbin Li, Lei Wang, Jinglin Xu et al.

Few-shot learning in image classification aims to learn a classifier to classify images when only few training examples are available for each class. Recent work has achieved promising classification performance, where an image-level feature based measure is usually used. In this paper, we argue that a measure at such a level may not be effective enough in light of the scarcity of examples in few-shot learning. Instead, we think a local descriptor based image-to-class measure should be taken, inspired by its surprising success in the heydays of local invariant features. Specifically, building upon the recent episodic training mechanism, we propose a Deep Nearest Neighbor Neural Network (DN4 in short) and train it in an end-to-end manner. Its key difference from the literature is the replacement of the image-level feature based measure in the final layer by a local descriptor based image-to-class measure. This measure is conducted online via a $k$-nearest neighbor search over the deep local descriptors of convolutional feature maps. The proposed DN4 not only learns the optimal deep local descriptors for the image-to-class measure, but also utilizes the higher efficiency of such a measure in the case of example scarcity, thanks to the exchangeability of visual patterns across the images in the same class. Our work leads to a simple, effective, and computationally efficient framework for few-shot learning. Experimental study on benchmark datasets consistently shows its superiority over the related state-of-the-art, with the largest absolute improvement of $17\%$ over the next best. The source code can be available from \UrlFont{https://github.com/WenbinLee/DN4.git}.

Jinquan Sun, Yinghuan Shi, Yang Gao et al.

Due to low tissue contrast, irregular object appearance, and unpredictable location variation, segmenting the objects from different medical imaging modalities (e.g., CT, MR) is considered as an important yet challenging task. In this paper, we present a novel method for interactive medical image segmentation with the following merits. (1) Our design is fundamentally different from previous pure patch-based and image-based segmentation methods. We observe that during delineation, the physician repeatedly check the inside-outside intensity changing to determine the boundary, which indicates that comparison in an inside-outside manner is extremely important. Thus, we innovatively model our segmentation task as learning the representation of the bi-directional sequential patches, starting from (or ending in) the given central point of the object. This can be realized by our proposed ConvRNN network embedded with a gated memory propagation unit. (2) Unlike previous interactive methods (requiring bounding box or seed points), we only ask the physician to merely click on the rough central point of the object before segmentation, which could simultaneously enhance the performance and reduce the segmentation time. (3) We utilize our method in a multi-level framework for better performance. We systematically evaluate our method in three different segmentation tasks including CT kidney tumor, MR prostate, and PROMISE12 challenge, showing promising results compared with state-of-the-art methods. The code is available here: \href{https://github.com/sunalbert/Sequential-patch-based-segmentation}{Sequential-patch-based-segmentation}.

Haoyuan Dong, Yang Gao, Haishuai Wang et al.

Heterogeneous graph neural architecture search (HGNAS) represents a powerful tool for automatically designing effective heterogeneous graph neural networks. However, existing HGNAS algorithms suffer from inefficient searches and unstable results. In this paper, we present a new GPT-4 based HGNAS model to improve the search efficiency and search accuracy of HGNAS. Specifically, we present a new GPT-4 enhanced Heterogeneous Graph Neural Architecture Search (GHGNAS for short). The basic idea of GHGNAS is to design a set of prompts that can guide GPT-4 toward the task of generating new heterogeneous graph neural architectures. By iteratively asking GPT-4 with the prompts, GHGNAS continually validates the accuracy of the generated HGNNs and uses the feedback to further optimize the prompts. Experimental results show that GHGNAS can design new HGNNs by leveraging the powerful generalization capability of GPT-4. Moreover, GHGNAS runs more effectively and stably than previous HGNAS models based on reinforcement learning and differentiable search algorithms.

Na Wang, Lei Qi, Jintao Guo et al.

Domain generalization (DG) intends to train a model on multiple source domains to ensure that it can generalize well to an arbitrary unseen target domain. The acquisition of domain-invariant representations is pivotal for DG as they possess the ability to capture the inherent semantic information of the data, mitigate the influence of domain shift, and enhance the generalization capability of the model. Adopting multiple perspectives, such as the sample and the feature, proves to be effective. The sample perspective facilitates data augmentation through data manipulation techniques, whereas the feature perspective enables the extraction of meaningful generalization features. In this paper, we focus on improving the generalization ability of the model by compelling it to acquire domain-invariant representations from both the sample and feature perspectives by disentangling spurious correlations and enhancing potential correlations. 1) From the sample perspective, we develop a frequency restriction module, guiding the model to focus on the relevant correlations between object features and labels, thereby disentangling spurious correlations. 2) From the feature perspective, the simple Tail Interaction module implicitly enhances potential correlations among all samples from all source domains, facilitating the acquisition of domain-invariant representations across multiple domains for the model. The experimental results show that Convolutional Neural Networks (CNNs) or Multi-Layer Perceptrons (MLPs) with a strong baseline embedded with these two modules can achieve superior results, e.g., an average accuracy of 92.30% on Digits-DG.

Muleilan Pei, Jiayao Shan, Peiliang Li et al.

Online scene perception and topology reasoning are critical for autonomous vehicles to understand their driving environments, particularly for mapless driving systems that endeavor to reduce reliance on costly High-Definition (HD) maps. However, recent advances in online scene understanding still face limitations, especially in long-range or occluded scenarios, due to the inherent constraints of onboard sensors. To address this challenge, we propose a Standard-Definition (SD) Map Enhanced scene Perception and Topology reasoning (SEPT) framework, which explores how to effectively incorporate the SD map as prior knowledge into existing perception and reasoning pipelines. Specifically, we introduce a novel hybrid feature fusion strategy that combines SD maps with Bird's-Eye-View (BEV) features, considering both rasterized and vectorized representations, while mitigating potential misalignment between SD maps and BEV feature spaces. Additionally, we leverage the SD map characteristics to design an auxiliary intersection-aware keypoint detection task, which further enhances the overall scene understanding performance. Experimental results on the large-scale OpenLane-V2 dataset demonstrate that by effectively integrating SD map priors, our framework significantly improves both scene perception and topology reasoning, outperforming existing methods by a substantial margin.

Jiawei Li, Xinyue Liang, Junlong Zhang et al.

Process supervision enhances the performance of large language models in reasoning tasks by providing feedback at each step of chain-of-thought reasoning. However, due to the lack of effective process supervision methods, even advanced large language models are prone to logical errors and redundant reasoning. We claim that the effectiveness of process supervision significantly depends on both the accuracy and the length of reasoning chains. Moreover, we identify that these factors exhibit a nonlinear relationship with the overall reward score of the reasoning process. Inspired by these insights, we propose a novel process supervision paradigm, PSPO*, which systematically outlines the workflow from reward model training to policy optimization, and highlights the importance of nonlinear rewards in process supervision. Based on PSPO*, we develop the PSPO-WRS, which considers the number of reasoning steps in determining reward scores and utilizes an adjusted Weibull distribution for nonlinear reward shaping. Experimental results on six mathematical reasoning datasets demonstrate that PSPO-WRS consistently outperforms current mainstream models.

Dongdong Ren, Wenbin Li, Tianyu Ding et al.

Recent advancements in model pruning have focused on developing new algorithms and improving upon benchmarks. However, the practical application of these algorithms across various models and platforms remains a significant challenge. To address this challenge, we propose ONNXPruner, a versatile pruning adapter designed for the ONNX format models. ONNXPruner streamlines the adaptation process across diverse deep learning frameworks and hardware platforms. A novel aspect of ONNXPruner is its use of node association trees, which automatically adapt to various model architectures. These trees clarify the structural relationships between nodes, guiding the pruning process, particularly highlighting the impact on interconnected nodes. Furthermore, we introduce a tree-level evaluation method. By leveraging node association trees, this method allows for a comprehensive analysis beyond traditional single-node evaluations, enhancing pruning performance without the need for extra operations. Experiments across multiple models and datasets confirm ONNXPruner's strong adaptability and increased efficacy. Our work aims to advance the practical application of model pruning.

Hongye Cao, Fan Feng, Tianpei Yang et al.

Current Reinforcement Learning (RL) methods often suffer from sample-inefficiency, resulting from blind exploration strategies that neglect causal relationships among states, actions, and rewards. Although recent causal approaches aim to address this problem, they lack grounded modeling of reward-guided causal understanding of states and actions for goal-orientation, thus impairing learning efficiency. To tackle this issue, we propose a novel method named Causal Information Prioritization (CIP) that improves sample efficiency by leveraging factored MDPs to infer causal relationships between different dimensions of states and actions with respect to rewards, enabling the prioritization of causal information. Specifically, CIP identifies and leverages causal relationships between states and rewards to execute counterfactual data augmentation to prioritize high-impact state features under the causal understanding of the environments. Moreover, CIP integrates a causality-aware empowerment learning objective, which significantly enhances the agent's execution of reward-guided actions for more efficient exploration in complex environments. To fully assess the effectiveness of CIP, we conduct extensive experiments across 39 tasks in 5 diverse continuous control environments, encompassing both locomotion and manipulation skills learning with pixel-based and sparse reward settings. Experimental results demonstrate that CIP consistently outperforms existing RL methods across a wide range of scenarios.

You Wang, Zekun Li, Lei Qi et al.

Despite the promising performance achieved by current semi-supervised models in segmenting individual medical targets, many of these models suffer a notable decrease in performance when tasked with the simultaneous segmentation of multiple targets. A vital factor could be attributed to the imbalanced scales among different targets: during simultaneously segmenting multiple targets, large targets dominate the loss, leading to small targets being misclassified as larger ones. To this end, we propose a novel method, which consists of a Collaborative Generalist and several Specialists, termed CGS. It is centered around the idea of employing a specialist for each target class, thus avoiding the dominance of larger targets. The generalist performs conventional multi-target segmentation, while each specialist is dedicated to distinguishing a specific target class from the remaining target classes and the background. Based on a theoretical insight, we demonstrate that CGS can achieve a more balanced training. Moreover, we develop cross-consistency losses to foster collaborative learning between the generalist and the specialists. Lastly, regarding their intrinsic relation that the target class of any specialized head should belong to the remaining classes of the other heads, we introduce an inter-head error detection module to further enhance the quality of pseudo-labels. Experimental results on three popular benchmarks showcase its superior performance compared to state-of-the-art methods.

Shaoqi Dong, Chaoyou Fu, Haihan Gao et al.

Vision-Language Action (VLA) models significantly advance robotic manipulation by leveraging the strong perception capabilities of pretrained vision-language models (VLMs). By integrating action modules into these pretrained models, VLA methods exhibit improved generalization. However, training them from scratch is costly. In this work, we propose a simple yet effective distillation-based framework that equips VLMs with action-execution capability by transferring knowledge from pretrained small action models. Our architecture retains the original VLM structure, adding only an action token and a state encoder to incorporate physical inputs. To distill action knowledge, we adopt a two-stage training strategy. First, we perform lightweight alignment by mapping VLM hidden states into the action space of the small action model, enabling effective reuse of its pretrained action decoder and avoiding expensive pretraining. Second, we selectively fine-tune the language model, state encoder, and action modules, enabling the system to integrate multimodal inputs with precise action generation. Specifically, the action token provides the VLM with a direct handle for predicting future actions, while the state encoder allows the model to incorporate robot dynamics not captured by vision alone. This design yields substantial efficiency gains over training large VLA models from scratch. Compared with previous state-of-the-art methods, our method achieves 97.3% average success rate on LIBERO (11.8% improvement) and 93.5% on LIBERO-LONG (24.5% improvement). In real-world experiments across five manipulation tasks, our method consistently outperforms the teacher model, achieving 82.0% success rate (17% improvement), which demonstrate that action distillation effectively enables VLMs to generate precise actions while substantially reducing training costs.

Yizhe Yang, Heyan Huang, Yihang Liu et al.

Knowledge-grounded dialogue is a task of generating an informative response based on both the dialogue history and external knowledge source. In general, there are two forms of knowledge: manually annotated knowledge graphs and knowledge text from website. From various evaluation viewpoints, each type of knowledge has advantages and downsides. To further distinguish the principles and determinants from the intricate factors, we conduct a thorough experiment and study on the task to answer three essential questions. The questions involve the choice of appropriate knowledge form, the degree of mutual effects between knowledge and the model selection, and the few-shot performance of knowledge. Supported by statistical shreds of evidence, we offer conclusive solutions and sensible suggestions for directions and standards of future research.

Ze Peng, Jian Zhang, Jintao Guo et al.

Continual learning seeks the human-like ability to accumulate new skills in machine intelligence. Its central challenge is catastrophic forgetting, whose underlying cause has not been fully understood for deep networks. In this paper, we demystify catastrophic forgetting by revealing that the new-task training is implicitly an adversarial attack against the old-task knowledge. Specifically, the new-task gradients automatically and accurately align with the sharp directions of the old-task loss landscape, rapidly increasing the old-task loss. This adversarial alignment is intriguingly counter-intuitive because the sharp directions are too sparsely distributed to align with by chance. To understand it, we theoretically show that it arises from training's low-rank bias, which, through forward and backward propagation, confines the two directions into the same low-dimensional subspace, facilitating alignment. Gradient projection (GP) methods, a representative family of forgetting-mitigating methods, reduce adversarial alignment caused by forward propagation, but cannot address the alignment due to backward propagation. We propose backGP to address it, which reduces forgetting by 10.8% and improves accuracy by 12.7% on average over GP methods.