AAAI 2025 Papers

We introduce Neural Conjugate Flows (NCF), a class of neural-network architectures equipped with exact flow structure. By leveraging topological conjugation, we prove that these networks are not only naturally isomorphic to a continuous group, but are also universal approximators for flows of ordinary differential equation (ODEs). Furthermore, topological properties of these flows can be enforced by the architecture in an interpretable manner. We demonstrate in numerical experiments how this topological group structure leads to concrete computational gains over other physics informed neural networks in estimating and extrapolating latent dynamics of ODEs, while training up to five times faster than other flow-based architectures.

We introduce a neural network conformal prediction method for time series that enhances adaptivity in non-stationary environments. Our approach acts as a neural controller designed to achieve desired target coverage, leveraging auxiliary multi-view data with neural network encoders in an end-to-end manner to further enhance adaptivity. Additionally, our model is designed to enhance the consistency of prediction intervals in different quantiles by integrating monotonicity constraints and leverages data from related tasks to boost few-shot learning performance. Using real-world datasets from epidemics, electric demand, weather, and others, we empirically demonstrate significant improvements in coverage and probabilistic accuracy, and find that our method is the only one that combines good calibration with consistency in prediction intervals.

Neural combinatorial optimization (NCO) has gained significant attention due to the potential of deep learning to efficiently solve combinatorial optimization problems. NCO has been widely applied to job shop scheduling problems (JSPs) with the current focus predominantly on deterministic problems. In this paper, we propose a novel attention-based scenario processing module (SPM) to extend NCO methods for solving stochastic JSPs. Our approach explicitly incorporates stochastic information by an attention mechanism that captures the embedding of sampled scenarios (i.e., an approximation of stochasticity). Fed with the embedding, the base neural network is intervened by the attended scenarios, which accordingly learns an effective policy under stochasticity. We also propose a training paradigm that works harmoniously with either the expected makespan or Value-at-Risk objective. Results demonstrate that our approach outperforms existing learning and non-learning methods for the flexible JSP problem with stochastic processing times on a variety of instances. In addition, our approach holds significant generalizability to varied numbers of scenarios and disparate distributions.

We consider the contextual combinatorial bandit setting where in each round, the learning agent, e.g., a recommender system, selects a subset of "arms,'' e.g., products, and observes rewards for both the individual base arms, which are a function of known features (called "context''), and the super arm (the subset of arms), which is a function of the base arm rewards. The agent's goal is to simultaneously learn the unknown reward functions and choose the highest-reward arms. For example, the "reward'' may represent a user's probability of clicking on one of the recommended products. Conventional bandit models, however, employ restrictive reward function models in order to obtain performance guarantees. We make use of deep neural networks to estimate and learn the unknown reward functions and propose Neural UCB Clustering (NeUClust), which adopts a clustering approach to select the super arm in every round by exploiting underlying structure in the context space. Unlike prior neural bandit works, NeUClust uses a neural network to estimate the super arm reward and select the super arm, thus eliminating the need for a known optimization oracle. We non-trivially extend prior neural combinatorial bandit works to prove that NeUClust achieves sublinear regret in the number of rounds. Experiments on real world recommendation datasets show that NeUClust achieves better regret and reward than other contextual combinatorial and neural bandit algorithms.

Existing knowledge distillation (KD) methods have demonstrated their ability to achieve student network performance on par with their teachers. However, the knowledge gap between the teacher and student remains significant and may hinder the effectiveness of the distillation process. In this work, we introduce the structure of Neural Collapse (NC) into the KD framework. NC typically occurs in the final phase of training, resulting in a graceful geometric structure where the last-layer features form a simplex equiangular tight frame. We hypothesize that NC can alleviate the knowledge gap in distillation, thereby enhancing student performance. This paper begins with an empirical analysis to bridge the connection between KD and NC. Through this analysis, we establish that transferring the teacher's NC structure to the student benefits the distillation process. Therefore, instead of merely transferring instance-level logits or features, as done by existing distillation methods, we encourage students to learn the teacher's NC structure. We propose the new distillation paradigm termed Neural Collapse-inspired Knowledge Distillation (NCKD). Comprehensive experiments demonstrate that NCKD is simple yet effective, improving the generalization of all distilled student models and achieving state-of-the-art accuracy performance.

The imperative for compression of material textures emerges from the critical demand for high-quality rendering, which necessitates sophisticated textures that, in turn, require substantial storage and memory resources. Thus, low-bitrate compression is crucial, especially in modern games demanding higher texture resolutions. Concurrent methodologies in texture compression predominantly employ a block-based paradigm based on color space, which inevitably leads to representational redundancies and a limited compression scope, particularly at lower bitrates. In the context of mobile devices, bandwidth during texture loading and runtime memory are major bottlenecks, making existing compression algorithms inadequate for high-resolution textures. To mitigate these limitations, we propose a novel multi-resolution texture compression scheme, Neural Block Compression (NBC), developed within the neural feature domain. Our encoding scheme is constructed on a hierarchy of multi-resolution neural feature blocks, and the key ingredient is the variable bitrates quantization scheme. It allocates higher bitrates to higher feature mip-levels and lower bitrates to lower feature mip-levels, thereby extending the concept of block compression from color domain into neural feature domain. Extensive experiments demonstrate the superior texture compression quality achieved by the proposed scheme, especially at low bitrates.

Image-guided object assembly represents a burgeoning research topic in computer vision. This paper introduces a novel task: translating multi-view images of a structural 3D model (for example, one constructed with building blocks drawn from a 3D-object library) into a detailed sequence of assembly instructions executable by a robotic arm. Fed with multi-view images of the target 3D model for replication, the model designed for this task must address several sub-tasks, including recognizing individual components used in constructing the 3D model, estimating the geometric pose of each component, and deducing a feasible assembly order adhering to physical rules. Establishing accurate 2D-3D correspondence between multi-view images and 3D objects is technically challenging. To tackle this, we propose an end-to-end model known as the Neural Assembler. This model learns an object graph where each vertex represents recognized components from the images, and the edges specify the topology of the 3D model, enabling the derivation of an assembly plan. We establish benchmarks for this task and conduct comprehensive empirical evaluations of Neural Assembler and alternative solutions. Our experiments clearly demonstrate the superiority of Neural Assembler.

Neural networks have emerged as powerful tools across various applications, yet their decision-making process often remains opaque, leading to them being perceived as "black boxes." This opacity raises concerns about their interpretability and reliability, especially in safety-critical scenarios. Network inversion techniques offer a solution by allowing us to peek inside these black boxes, revealing the features and patterns learned by the networks behind their decision-making processes and thereby provide valuable insights into how neural networks arrive at their conclusions, making them more interpretable and trustworthy. This paper presents a simple yet effective approach to network inversion using a meticulously conditioned generator that learns the data distribution in the input space of the trained neural network, enabling the reconstruction of inputs that would most likely lead to the desired outputs. To capture the diversity in the input space for a given output, instead of simply revealing the conditioning labels to the generator, we encode the conditioning label information into vectors and intermediate matrices and further minimize the cosine similarity between features of the generated images.

Simple random negative sampling is a technique used to enhance decision-making in sequential models with numerous potential negative instances, like recommender systems. However, it ignores the patterns that can be discovered in complex sequences to select the most informative negative samples. In this paper, we address this challenge by introducing a Neighborhood-Aware Negative Sampling (NANS) technique in the context of student knowledge modeling (KM) and behavior modeling (BM). In the education domain, KM quantifies student knowledge based on past performance, while BM focuses on behaviors like student preferences of questions. With the vast number of problems to choose from and the intricate relationship between student knowledge and behavior, selecting the proper negative samples becomes a notable challenge in this problem. NANS, along with our proposed multi-objective, multi-task sequential model for KM and BM, NANS-KoBeM frames the simultaneous modeling of student knowledge and question selection as a multi-task learning problem with dual objectives: predicting students’ performance and their question selections.

In medical image analysis, multi-organ semi-supervised segmentation faces challenges such as insufficient labels and low contrast in soft tissues. To address these issues, existing studies typically employ semi-supervised segmentation techniques using pseudo-labeling and consistency regularization. However, these methods mainly rely on individual data samples for training, ignoring the rich neighborhood information present in the feature space. In this work, we argue that supervisory information can be directly extracted from the geometry of the feature space. Inspired by the density-based clustering hypothesis, we propose using feature density to locate sparse regions within feature clusters. Our goal is to increase intra-class compactness by addressing sparsity issues. To achieve this, we propose a Density-Aware Contrastive Learning (DACL) strategy, pushing anchored features in sparse regions towards cluster centers approximated by high-density positive samples, resulting in more compact clusters. Specifically, our method constructs density-aware neighbor graphs using labeled and unlabeled data samples to estimate feature density and locate sparse regions. We also combine label-guided co-training with density-guided geometric regularization to form complementary supervision for unlabeled data. Experiments on the Multi-Organ Segmentation Challenge dataset demonstrate that our proposed method outperforms state-of-the-art methods, highlighting its efficacy in medical image segmentation tasks.

In Hotelling's model of spatial competition, a unit mass of voters is distributed in the interval [0,1] (with their location corresponding to their political persuasion), and each of m candidates selects as a strategy their distinct position in this interval. Each voter votes for the nearest candidate, and candidates choose their strategy to maximize their votes. It is known that if there are more than two candidates, equilibria may not exist in this model. It was unknown, however, how close to an equilibrium one could get. Our work studies approximate equilibria in this model, where a strategy profile is an (additive) ϵ-equilibria if no candidate can increase their votes by ϵ, and provides tight or nearly-tight bounds on the approximation ϵ achievable. We show that for 3 candidates, for any distribution of the voters, ϵ ≥ 1/12. Thus, somewhat surprisingly, for any distribution of the voters and any strategy profile of the candidates, at least 1/12th of the total votes is always left ``on the table.'' Extending this, we show that in the worst case, there exist voter distributions for which ϵ ≥ 1/6, and this is tight: one can always compute a 1/6-approximate equilibria in polynomial time. We then study the general case of m candidates, and show that as m grows large, we get closer to an exact equilibrium: one can always obtain a 1/(m+1)-approximate equilibria in polynomial time. We show this bound is asymptotically tight, by giving voter distributions for which ϵ ≥ 1/(m+3).

We investigate the sample-memory-pass trade-offs for pure exploration in multi-pass streaming multi-armed bandits (MABs) with the *a priori* knowledge of the optimality gap ?_[2]. Here, and throughout, the optimality gap ?_[i] is defined as the mean reward gap between the best and the i-th best arms. A recent line of results have shown that if there is no known ?_[2], a pass complexity of ̃?(log(1/?_[2])) is necessary and sufficient to obtain the *worst-case optimal* O(n/?²_[2]) sample complexity with a single-arm memory. However, our understanding of multi-pass algorithms with known ?_[2] is still limited. Here, the key open problem is how many passes are required to achieve the complexity, i.e., O( ∑ᵢ₌₂ⁿ1/?²_[i] log{n}) arm pulls, with a sublinear memory size. In this work, we show that the ``right answer'' for the question is ̃?(log{n}) passes. We first present a lower bound, showing that any algorithm that finds the best arm with slightly sublinear memory -- a memory of o(n/polylog(n)) arms -- and O( ∑ᵢ₌₂ⁿ1/?²_[i] log n) arm pulls has to make ?(log n/loglog n) passes over the stream. We then show a nearly-matching algorithm that assuming the knowledge of ?_[2], finds the best arm with O( ∑ᵢ₌₂ⁿ1/?²_[i] log n) arm pulls and a *single arm* memory.

Compositional generalization is crucial for artificial intelligence agents to solve complex vision-language reasoning tasks. Neuro-symbolic approaches have demonstrated promise in capturing compositional structures, but they face critical challenges: (a) reliance on predefined predicates for symbolic representations that limit adaptability, (b) difficulty in extracting predicates from raw data, and (c) using non-differentiable operations for combining primitive concepts. To address these issues, we propose NeSyCoCo, a neuro-symbolic framework that leverages large language models (LLMs) to generate symbolic representations and map them to differentiable neural computations. NeSyCoCo introduces three innovations: (a) augmenting natural language inputs with dependency structures to enhance the alignment with symbolic representations, (b) employing distributed word representations to link diverse, linguistically motivated logical predicates to neural modules, and (c) using the soft composition of normalized predicate scores to align symbolic and differentiable reasoning. Our framework achieves state-of-the-art results on the ReaSCAN and CLEVR-CoGenT compositional generalization benchmarks and demonstrates robust performance with novel concepts in the CLEVR-SYN benchmark.

Machine learning algorithms often struggle to eliminate inherent data biases, particularly those arising from unreliable labels, which poses a significant challenge in ensuring fairness. Existing fairness techniques that address label bias typically involve modifying models and intervening in the training process, but these lack flexibility for large-scale datasets. To address this limitation, we introduce a data selection method designed to efficiently and flexibly mitigate label bias, tailored to more practical needs. Our approach utilizes a zero-shot predictor as a proxy model that simulates training on a clean holdout set. This strategy, supported by peer predictions, ensures the fairness of the proxy model and eliminates the need for an additional holdout set, which is a common requirement in previous methods. Without altering the classifier's architecture, our modality-agnostic method effectively selects appropriate training data and has proven efficient and effective in handling label bias and improving fairness across diverse datasets in experimental evaluations.

Facial expression recognition (FER) remains a challenging task due to label ambiguity caused by the subjective nature of facial expressions and noisy samples. Additionally, class imbalance, which is common in real-world datasets, further complicates FER. Although many studies have shown impressive improvements, they typically address only one of these issues, leading to suboptimal results. To tackle both challenges simultaneously, we propose a novel framework called Navigating Label Ambiguity (NLA), which is robust under real-world conditions. The motivation behind NLA is that dynamically estimating and emphasizing ambiguous samples at each iteration helps mitigate noise and class imbalance by reducing the model's bias toward majority classes. To achieve this, NLA consists of two main components: Noise-aware Adaptive Weighting (NAW) and consistency regularization. Specifically, NAW adaptively assigns higher importance to ambiguous samples and lower importance to noisy ones, based on the correlation between the intermediate prediction scores for the ground truth and the nearest negative. Moreover, we incorporate a regularization term to ensure consistent latent distributions. Consequently, NLA enables the model to progressively focus on more challenging ambiguous samples, which primarily belong to the minority class, in the later stages of training. Extensive experiments demonstrate that NLA outperforms existing methods in both overall and mean accuracy, confirming its robustness against noise and class imbalance. To the best of our knowledge, this is the first framework to address both problems simultaneously.

Learning discriminative state representations of agents, encompassing the spatial layout and temporal pose trajectory, is essential for effective navigation decisions. However, existing approaches often rely on simplistic plain networks for navigation information fusion, overlooking the complex long-range dependencies across spatio-temporal cues, which leads to suboptimal state perception and potential decision failures. In this paper, we introduce NaviFormer, an effective encoder-decoder navigation transformer, to aggregate discriminative spatio-temporal context information for object navigation. Our navigation encoder not only encodes spatial layouts and temporal agent poses but also innovatively constructs and encodes a passable frontier map, enriching the original state encoding with cues of potential exploration regions. Furthermore, our navigation decoder employs spatio-temporal self-attention and cross-attention mechanisms to model the dependencies among spatial layout encoding, temporal pose encoding, and passable frontier encoding, thereby facilitating comprehensive contextual state feature aggregation. Finally, we leverage these learned spatio-temporal contextual state representations for PPO-based navigation decisions. Extensive experiments on the Gibson, Habitat-Matterport3D (HM3D) and Matterport3D (MP3D) datasets demonstrate the superiority of our approach.

Standards, or expert-defined preferences, are documented guidelines describing strict specifications for text-based content such as books, manuals, and reports. These guidelines are curated, defined, and continuously improved by domain experts in various fields, such as education, policy, and healthcare, and are used for maintaining quality. In my dissertation, I focus on evaluating and teaching large language models (LLMs) to capture standards to improve generation quality across diverse language generation tasks. I draw motivation from my preliminary published works, where I explored how open and commercial LLMs can learn complex constraints from standards in education and language assessment to produce classroom-ready narrative content. In this proposal, I also discuss the technical novelty, impact, and target contributions and highlight how this line of work can be scaled and generalized for other domains where standards are also used as a reference of quality.

Breast cancer remains a leading cause of mortality among women, with millions of new cases diagnosed annually. Early detection through screening is crucial. Using neural networks to improve the accuracy of breast cancer screening has become increasingly important. In accordance with radiologists' practices, we proposed using images from the unaffected side to create adversarial samples with critical medical implications in our adversarial learning process. By introducing beneficial perturbations, this method aims to reduce overconfidence and improve the precision and robustness of breast cancer classification. Our proposed framework is an adversarial quadruple-view classification network (NaFV-Net) incorporating images from both affected and unaffected perspectives. By comprehensively capturing local and global information and implementing adversarial learning from four mammography views, this framework allows for the fusion of features and the integration of medical principles and radiologist evaluation techniques, thus facilitating the accurate identification and characterization of breast tissues. Extensive experiments have shown the high effectiveness of our model in accurately distinguishing between benign and malignant findings, demonstrating state-of-the-art classification performance on both internal and public datasets.

In Explainable AI (XAI), counterfactual explanations (CEs) are a well-studied method to communicate feature relevance through contrastive reasoning of ``what if'' to explain AI models' predictions. However, they only focus on important (i.e., relevant) features and largely disregard less important (i.e., irrelevant) ones. Such irrelevant features can be crucial in many applications, especially when users need to ensure that an AI model's decisions are not affected or biased against specific attributes such as gender, race, religion, or political affiliation. To address this gap, the concept of alterfactual explanations (AEs) has been proposed. AEs explore an alternative reality of ``no matter what'', where irrelevant features are substituted with alternative features (e.g., ``republicans'' -> ``democrats'') within the same attribute (e.g., ``politics'') while maintaining a similar prediction output. This serves to validate whether the specified attributes influence AI model predictions. Despite the promise of AEs, there is a lack of computational approaches to systematically generate them, particularly in the text domain, where creating AEs for AI text classifiers presents unique challenges. This paper addresses this challenge by formulating AE generation as an optimization problem and introducing NoMatterXAI, a novel algorithm that generates AEs for text classification tasks. Our approach achieves high fidelity of up to 95% while preserving context similarity of over 90% across multiple models and datasets. A human study further validates the effectiveness of AEs in explaining AI text classifiers to end users.