AAAI 2025 论文列表

Randomized search heuristics have been applied successfully to a plethora of problems. This success is complemented by a large body of theoretical results. Unfortunately, the vast majority of these results regard problems with binary or continuous decision variables -- the theoretical analysis of randomized search heuristics for unbounded integer domains is almost nonexistent. To resolve this shortcoming, we start the runtime analysis of multi-objective evolutionary algorithms, which are among the most successful randomized search heuristics, for unbounded integer search spaces. We analyze single- and full-dimensional mutation operators with three different mutation strengths, namely changes by plus/minus one (unit strength), random changes following a law with exponential tails, and random changes following a power-law. The performance guarantees we prove on a recently proposed natural benchmark problem suggest that unit mutation strengths can be slow when the initial solutions are far from the Pareto front. When setting the expected change right (depending on the benchmark parameter and the distance of the initial solutions), the mutation strength with exponential tails yields the best runtime guarantees in our results -- however, with a wrong choice of this expectation, the performance guarantees quickly become highly uninteresting. With power-law mutation, which is an essentially parameter-less mutation operator, we obtain good results uniformly over all problem parameters and starting points. We complement our mathematical findings with experimental results that suggest that our bounds are not always tight. Most prominently, our experiments indicate that power-law mutation outperforms the one with exponential tails even when the latter uses a near-optimal parametrization. Hence, we suggest to favor power-law mutation for unknown problems in integer spaces.

The connections between symbolic rules and neural networks have been explored in various directions, including rule mining through neural networks and rule-based explanation for neural networks. These approaches allow symbolic rules to be extracted from neural network models, which offers explainability to the models. However, the plausibility of the extracted rules is rarely analysed. In this paper, we show that the confidence degrees of extracted rules are generally not high, and we propose a new family of Graph Neural Networks that can be trained with the guidance of rules. Hence, the inference of our model simulates the rule reasoning. Moreover, rules with high confidence degrees can be extracted from the trained model that aligns with the inference of the model, which verifies the effectiveness of the rule guidance. Experimental evaluation of knowledge graph reasoning tasks further demonstrates the effectiveness of our model.

Information retrieval methods often rely on a single embedding model trained on large, general-domain datasets like MSMARCO. While this approach can produce a retriever with reasonable overall performance, they often underperform models trained on domain-specific data when testing on their respective domains. Prior work in information retrieval has tackled this through multi-task training, but the idea of routing over a mixture of domain-specific expert retrievers remains unexplored despite the popularity of such ideas in language model generation research. In this work, we introduce RouterRetriever, a retrieval model that leverages a mixture of domain-specific experts by using a routing mechanism to select the most appropriate expert for each query. RouterRetriever is lightweight and allows easy addition or removal of experts without additional training. Evaluation on the BEIR benchmark demonstrates that RouterRetriever outperforms both models trained on MSMARCO (+2.1 absolute nDCG@10) and multi-task models (+3.2). This is achieved by employing our routing mechanism, which surpasses other routing techniques (+1.8 on average) commonly used in language modeling. Furthermore, the benefit generalizes well to other datasets, even in the absence of a specific expert on the dataset. RouterRetriever is the first work to demonstrate the advantages of routing over a mixture of domain-specific expert embedding models as an alternative to a single, general-purpose embedding model, especially when retrieving from diverse, specialized domains.

Recent advancements in online Video Instance Segmentation (VIS) methods show notable performance improvements across benchmarks. However, the leading methods in the tracking-by-detection paradigm often result in temporally inconsistent predictions at both instance-level and pixel-level that lead to visually unsatisfactory outcomes. To address these challenges, we propose RoCoVIS, a simple yet effective approach that integrates segmentation and tracking to provide consistent online VIS. Our approach is an end-to-end sequential learning where object queries are propagated through mask predictions, improving the accuracy of temporal instance mapping at the pixel level. Additionally, we propose a new label assignment criterion in harmony with our approach. We also examine the limitations and challenges presented by the current standard evaluation protocol (AP) and suggest adopting additional metrics, Tube-Boundary AP and AP_Pool. RoCoVIS demonstrates superior performance on challenging VIS benchmarks with a Swin-L backbone and shows competitive results when employing a ResNet-50 backbone. By employing Tube-Boundary AP and AP_Pool as metrics to measure mask accuracy and consistency, RoCoVIS outperforms its counterpart, GenVIS, on the HQ-YTVIS and VIPSeg.

Accurate forecasting of medication usage and ICD-9/10 code streams is critical for optimizing medical logistics, especially during periods of high demand, such as pandemics, disease outbreaks, wartime, or natural disasters. In this study, we develop a novel and robust forecasting framework using unsupervised learning techniques and Natural Language Processing (NLP) methods to build vector representations of daily ICD-9/10 codes and medication daily usage from Electronic Health Record (EHR) data. Multiple forecasting models, including Linear Drift Model, Vector Autoregression (VAR), Temporal Fusion Transformer (TFT), and Autoregressive Long Short-Term Memory (AR-LSTM) are trained, tested and evaluated. Finally multiple TFT and AR-LSTM models with different lookback horizon are trained and ensembled together to achieve better forecasting accuracy in near further (10 days). The AI framework is validated using MIMIC-IV ER and MIMIC-III datasets, resulting in the average forecasting error 5.2% at 5-th day and 18.1% at the 10-th day. The results demonstrate the ensemble model’s superior performance on near-future medication usage forecasting and ICD code progression, offering valuable insights for healthcare logistics and decision making. The framework also provides the mechanism to detect the model drift and finetune the model if necessary, which offers a robust tool for managing healthcare logistics under extreme and fluctuating conditions.

How to make a good trade-off between performance and computational cost is crucial for a tracker. However, current famous methods typically focus on complicated and time-consuming learning that combining temporal and appearance information by input more and more images (or features). Consequently, these methods not only increase the model's computational source and learning burden but also introduce much useless and potentially interfering information. To alleviate the above issues, we propose a simple yet robust tracker that separates temporal information learning from appearance modeling and extracts temporal relations from a set of representative tokens rather than several images (or features). Specifically, we introduce one track token for each frame to collect the target's appearance information in the backbone. Then, we design a mamba-based Temporal Module for track tokens to be aware of context by interacting with other track tokens within a sliding window. This module consists of a mamba layer with autoregressive characteristic and a cross-attention layer with strong global perception ability, ensuring sufficient interaction for track tokens to perceive the appearance changes and movement trends of the target. Finally, track tokens serve as a guidance to adjust the appearance feature for the final prediction in the head. Experiments show our method is effective and achieves competitive performance on multiple benchmarks at a real-time speed.

Cross-modal hashing (CMH) has appeared as a popular technique for cross-modal retrieval due to its low storage cost and high computational efficiency in large-scale data. Most existing methods implicitly assume that multi-modal data is correctly labeled, which is expensive and even unattainable due to the inevitable imperfect annotations (i.e., noisy labels) in real-world scenarios. Inspired by human cognitive learning, a few methods introduce self-paced learning to gradually train the model from easy to hard samples, which is often used to mitigate the effects of feature noise or outliers. It is a less-touched problem that how to utilize SPL to alleviate the misleading of noisy labels on the hash model. To tackle this problem, we propose a new cognitive cross-modal retrieval method called Robust Self-paced Hashing with Noisy Labels (RSHNL), which can mimic the human cognitive process to identify the noise while embracing robustness against noisy labels. Specifically, we first propose a contrastive hashing learning (CHL) scheme to improve multi-modal consistency, thereby reducing the inherent semantic gap. Afterward, we propose center aggregation learning (CAL) to mitigate the intra-class variations. Finally, we propose Noise-tolerance Self-paced Hashing (NSH) that dynamically estimates the learning difficulty for each instance and distinguishes noisy labels through the difficulty level. For all estimated clean pairs, we further adopt a self-paced regularizer to gradually learn hash codes from easy to hard. Extensive experiments demonstrate that the proposed RSHNL performs remarkably well over the state-of-the-art CMH methods.

The Segment Anything Model (SAM) is a widely used vision foundation model with diverse applications, including image segmentation, detection, and tracking. Given SAM's wide applications, understanding its robustness against adversarial attacks is crucial for real-world deployment. However, research on SAM's robustness is still in its early stages. Existing attacks often overlook the role of prompts in evaluating SAM's robustness, and there has been insufficient exploration of defense methods to balance the robustness and accuracy. To address these gaps, this paper proposes an adversarial robustness framework designed to evaluate and enhance the robustness of SAM. Specifically, we introduce a cross-prompt attack method to enhance the attack transferability across different prompt types. Besides attacking, we propose a few-parameter adaptation strategy to defend SAM against various adversarial attacks. To balance robustness and accuracy, we use the singular value decomposition (SVD) to constrain the space of trainable parameters, where only singular values are adaptable. Experiments demonstrate that our cross-prompt attack method outperforms previous approaches in terms of attack success rate on both SAM and SAM 2. By adapting only 512 parameters, we achieve at least a 15% improvement in mean intersection over union (mIoU) against various adversarial attacks. Compared to previous defense methods, our approach enhances the robustness of SAM while maximally maintaining its original performance.

Motivated by applications such as online labor markets we consider a variant of the stochastic multi-armed bandit problem where we have a collection of arms representing strategic agents with different performance characteristics. The platform (principal) chooses an agent in each round to complete a task. Unlike the standard setting, when an arm is pulled it can modify its reward by absorbing it or improving it at the expense of a higher cost. The principle has to solve a mechanism design problem to incentivize the arms to give their best performance. However, since even with an effective mechanism agents may still deviate from rational behavior, the principal wants a robust algorithm that also gives a non-vacuous guarantee on the total accumulated rewards under non-equilibrium behavior. In this paper, we introduce a class of bandit algorithms that meet the two objectives of performance incentivization and robustness simultaneously. We do this by identifying a collection of intuitive properties that a bandit algorithm has to satisfy to achieve these objectives. Finally, we show that settings where the principal has no information about the arms' performance characteristics can be handled by combining ideas from second price auctions with our algorithms.

Multi-objective preference alignment of large language models (LLMs) is critical for developing AI systems that are more configurable, personalizable, helpful, and safe. However, optimizing model outputs to satisfy diverse objectives with variable weights at inference time for truly personalized models presents a significant challenge. Existing approaches are either computationally expensive to train or do not sufficiently steer model behaviors. This paper introduces the Multi-Objective Online DPO (MO-ODPO) algorithm, designed to robustly and efficiently align model behaviors with multiple, potentially conflicting human preferences. Our approach incorporates a prompt conditioning mechanism, allowing us to train a single preference-conditional policy, that can adapt to new preference combinations at inference. Experiments on two popular benchmarks show that MO-ODPO Pareto-dominates existing baselines while providing excellent inference-time steerability between diverse objectives.

Learning with noisy labels (LNL) methods have enabled the deployment of machine learning systems with imperfectly labeled data. However, these methods often struggle to identify noise in the presence of long-tailed (LT) class distributions, where the memorization effect becomes class-dependent. Conversely, LT methods are suboptimal under label noise, as it hinders access to accurate label frequency statistics. This study aims to address the long-tailed noisy data by bridging the methodological gap between LNL and LT approaches. We propose a direct solution, termed Robust Logit Adjustment, which estimates ground-truth labels through label refurbishment, thereby mitigating the impact of label noise. Simultaneously, our method incorporates the distribution of training-time corrected target labels into the LT method logit adjustment, providing class-rebalanced supervision. Extensive experiments on both synthetic and real-world long-tailed noisy datasets demonstrate the superior performance of our method.

Recently, numerous robust image hashing schemes have been developed for content identification. However, many of these schemes face the challenges of maintaining discrimination while simultaneously resisting large-scale attacks. In this paper, we propose a robust image hashing scheme based on Contrastive Masked Autoencoder with weak-strong augmentation Alignment (CMAA). Leveraging contrastive learning, CMAA is designed to learn features that are robust to large-scale and hybrid attacks while maintaining the discrimination of those features. Specifically, it utilizes distribution divergence to align weak attack augmented features with strong attack augmented features, namely weak-strong augmentation alignment, to enhance the robustness to strong attacks. In addition, a masked vision transformer is incorporated to further enhance content identification performance. CMAA also includes a parameter-free quantization layer to mitigate the loss induced by binarization. Experimental results demonstrate that our method exhibits remarkable robustness against various attacks, including challenging ones such as rotation and hybrid attacks, and delivers excellent identification performance with a F1 score close to 1.0. Our code and supplementary materials are available on Github.

Heterogeneous Graph Neural Networks (HGNNs) have achieved state-of-the-art performance in classifying molecular graphs, capitalizing on their ability to capture rich semantics. However, HGNNs for molecule property prediction exhibit significant susceptibility to adversarial attacks—a challenge that prior research has entirely overlooked. To fill this gap, this paper introduces the first study focused on robust graph-level representation learning tailored for heterogeneous molecular graphs. To achieve this goal, we propose a comprehensive Robust Heterogeneous Graph Classification (RHGC) framework grounded in the Information Bottleneck principle, which aims to identify the most informative and least noisy heterogeneous subgraphs to derive robust, holistic representations. This is specifically accomplished through a dedicated Node Semantic Purifier, which enhances node-level and semantic-level robustness by eliminating label-irrelevant interference using graph stochastic attention and the Hilbert-Schmidt Independence Criterion, along with a Global Graph Disentanglement method, which improves graph-level robustness by addressing information leak. Experiments on three molecular benchmarks demonstrate that RHGC enhances accuracy by an average of 5.06% under all three attack settings and meanwhile by 4.33% on clean data.

Social recommendation leverages the social connections between users to mitigate the issue of data sparsity and enhance recommendation quality. Although existing related works show their effectiveness, there remain two critical questions: i) The patterns of preference interactions among users are varied and heterogeneous. Current models struggle to accurately capture preference shifts from user interactions in noisy social environments. ii) Existing methods handle the integration of auxiliary information coarsely, potentially introducing noise and leading to biases in user preferences. To address the limitations above, we introduce a novel framework named Robust Graph Based Social Recommendation through Contrastive Multi-view Learning (RGCML). This framework leverages denoised social relations and global intents as dual auxiliary information sources to provide comprehensive characterization of users. Firstly, RGCML employs the concept of opinion dynamics to simulate how user preferences evolve due to noisy social relations. Then, it utilizes a specifically designed information fusion module to extract critical contextual information from multiple semantic perspectives, thereby achieving efficient personalized information fusion. Finally, it adopts the designed global-local contrastive learning paradigm that untangles and discriminates user preferences from global intents, further addressing the noise problem and enhancing the quality of user representations. Extensive experiments conducted on three real-world datasets demonstrate the superior performance of RGCML compared to several state-of-the-art (SOTA) baselines.

Artificial Intelligence (AI) has revolutionized fields like computer vision and natural language processing, yet its impact on robotics remains limited by challenges in long-horizon decision-making and complex physical interactions. My research pioneers robot learning algorithms that exploit (predict, perceive, plan, and reason about) physical interaction as a core component of artificial intelligence, pushing beyond passive solutions in domains such as perception, navigation, and manipulation. By leveraging techniques in imitation learning and hierarchical reinforcement learning, my work empowers robots to learn from human demonstrations, navigate interactively in real-world environments, and gather information through purposeful interactions. In my talk, I will explain how these advances are critical for robots to become useful helpers in human environments, opening the door to the next generation of household robots. I will present several AI algorithmic innovations to integrate physical interactions in computation procedures and outline the path toward developing continually learning robots capable of operating autonomously in unstructured human environments, enhancing their utility as adaptable and intelligent assistants.

Large vision-language models (LVLMs) often fail to align with human preferences, leading to issues like generating misleading content without proper visual context (also known as hallucination). A promising solution to this problem is using human-preference alignment techniques, such as best-of-n sampling and reinforcement learning. However, these techniques face the difficulty arising from the scarcity of visual preference data, which is required to train a visual reward model (VRM). In this work, we continue the line of research. We present a Robust Visual Reward Model (RoVRM) which improves human-preference alignment for LVLMs. RoVRM leverages auxiliary textual preference data through a three-phase progressive training and optimal transport-based preference data selection to effectively mitigate the scarcity of visual preference data. We experiment with RoVRM on the commonly used vision-language tasks based on the LLaVA-1.5-7B and -13B models. Experimental results demonstrate that RoVRM consistently outperforms traditional VRMs. Furthermore, our three-phase progressive training and preference data selection approaches can yield consistent performance gains over ranking-based alignment techniques, such as direct preference optimization.

Screen-shooting robust watermarking is an effective means of preventing screen content leakage from unauthorized camera shooting, as it can trace the leaked source through the watermark extraction thereby providing an effective deterrent. However, current screen-shooting resilient watermarking schemes rely on the image's contours to synchronize and then extract the watermark. While in practical applications, it's common for only a portion of the image to be captured, resulting in a limited performance of the previous watermarking schemes. To address this problem, we propose the RoPaSS: a robust watermarking scheme for partial screen-shooting scenarios, which effectively constructs symmetric characteristics on the embedding watermark to handle the sticky re-synchronization issue. Specifically, RoPaSS consists of a watermark encoder, a decoder, and three estimators, which are trained in two stages. In the first training stage, RoPaSS integrates the flipping operation into the watermark encoder and decoder training to increase the redundancy of watermark messages and artificially guide the generation of symmetric watermarks. In the second stage, estimators utilize the watermark symmetry as an additional reference to estimate the restoration parameters to resynchronize the partially captured watermarked image. Experiments have demonstrated the excellent performance of RoPaSS in partial screen-shooting traceability, with extraction accuracy of above 93% in frontal shooting and above 86% in 30° shooting even if only 50% of the image content is captured.

This paper studies the complex challenge of cross-domain image retrieval under the condition of noisy labels (NCIR), a scenario that not only includes the inherent obstacles of traditional cross-domain image retrieval (CIR) but also requires alleviating the adverse effects of label noise. To address this challenge, this paper introduces a novel Robust Domain Alignment framework (RoDA), specifically designed for the NCIR task. At the heart of RoDA is the Selective Division and Adaptive Learning mechanism (SDAL), a key component crafted to shield the model from overfitting the noisy labels. SDAL effectively learns discriminative knowledge by dividing the dataset into clean and noisy parts, subsequently rectifying the labels for the latter based on information drawn from the clean one. This process involves adaptively weighting the relabeled samples and leveraging both the clean and relabeled data to bootstrap model training. Moreover, to bridge the domain gap further, we introduce the Accumulative Class Center Alignment (ACCA), a novel approach that fosters domain alignment through an accumulative domain loss mechanism.Thanks to SDAL and ACCA, our RoDA demonstrates its superiority in overcoming label noise and domain discrepancies within the NCIR paradigm. The effectiveness and robustness of our RoDA framework are comprehensively validated through extensive experiments across three multi-domain benchmarks.

Optimizing risk-averse objectives in discounted MDPs is challenging because most models do not admit direct dynamic programming equations and require complex history-dependent policies. In this paper, we show that the risk-averse total reward criterion, under the Entropic Risk Measure (ERM) and Entropic Value at Risk (EVaR) risk measures, can be optimized by a stationary policy, making it simple to analyze, interpret, and deploy. We propose exponential value iteration, policy iteration, and linear programming to compute optimal policies. Compared with prior work, our results only require the relatively mild condition of transient MDPs and allow for both positive and negative rewards. Our results indicate that the total reward criterion may be preferable to the discounted criterion in a broad range of risk-averse reinforcement learning domains.