AAAI 2025 Papers

Multi-object tracking faces a major challenge in handling the variations of tracked targets within complex scenes. In existing transformer-based tracking methods, typically each tracked target is only associated with one track query. However, trajectories in crowded scenes often experience varying levels of occlusion, making the association brittle for using a single track query to identify the tracked target. Therefore, we argue that relying on a single track query to track a target in complex scenes is inadequate. In this paper, we introduce TGFormer, with the core idea of designing a Track Query Group for each tracked target. Each group encompasses track queries that handle the same tracked target across different levels of occlusion scenes. To achieve long-term robust association, we propose a novel updater that integrates temporal memories and occlusion-aware features to update the Track Query Group, ensuring the tracked target can be consistently captured in complex scenes. Additionally, we introduce a Position Predictor that allows TGFormer to forecast motion trends, helping the model accurately locate moving tracklets. Experimental results show that our method achieves competitive performance on the MOT Challenge and DanceTrack datasets.

Video object detection has made significant progress in recent years thanks to convolutional neural networks (CNNs) and vision transformers (ViTs). Typically, CNNs excel at capturing local features but struggle to model global representations. Conversely, ViTs are adept at capturing long-range global features but face challenges in representing local feature details. Off-the-shelf video object detection methods solely rely on CNNs or ViTs to conduct feature aggregation, which hampers their capability to simultaneously leverage global and local information, thereby resulting in limited detection performance. In this paper, we propose a Transformer-GraphFormer Blender Network (TGBFormer) for video object detection, with three key technical improvements to fully exploit the advantages of transformers and graph convolutional networks while compensating for their limitations. First, we develop a spatial-temporal transformer module to aggregate global contextual information, constituting global representations with long-range feature dependencies. Second, we introduce a spatial-temporal GraphFormer module that utilizes local spatial and temporal relationships to aggregate features, generating new local representations that are complementary to the transformer outputs. Third, we design a global-local feature blender module to adaptively couple transformer-based global representations and GraphFormer-based local representations. Extensive experiments demonstrate that our TGBFormer establishes new state-of-the-art results on the ImageNet VID dataset. Particularly, our TGBFormer achieves 86.5% mAP while running at around 41.0 FPS on a single Tesla A100 GPU.

Currently, inspired by the success of vision-language models (VLMs), an increasing number of researchers are focusing on improving VLMs and have achieved promising results. However, most existing methods concentrate on optimizing the connector and enhancing the language model component, while neglecting improvements to the vision encoder itself. In contrast, we propose Text Guided LLaVA (TG-LLaVA) in this paper, which optimizes VLMs by guiding the vision encoder with text, offering a new and orthogonal optimization direction. Specifically, inspired by the purpose-driven logic inherent in human behavior, we use learnable latent embeddings as a bridge to analyze textual instruction and add the analysis results to the vision encoder as guidance, refining it. Subsequently, another set of latent embeddings extracts additional detailed text-guided information from high-resolution local patches as auxiliary information. Finally, with the guidance of text, the vision encoder can extract text-related features, similar to how humans focus on the most relevant parts of an image when considering a question. This results in generating better answers. Experiments on various datasets validate the effectiveness of the proposed method. Remarkably, without the need for additional training data, our proposed method can bring more benefits to the baseline (LLaVA-1.5) compared with other concurrent methods. Furthermore, the proposed method consistently brings improvement in different settings.

This paper presents the Text Encoding Diffusion Model (TEncDM), a novel approach to diffusion modeling that operates in the space of pre-trained language model encodings. In contrast to traditionally used embeddings, encodings integrate contextual information. In our approach, we also employ a transformer-based decoder, specifically designed to incorporate context in the token prediction process. We conduct a comprehensive examination of the influence of the encoder, decoder, noise scheduler, and self-conditioning on zero-shot generation. Furthermore, we compare TEncDM with previous approaches on three conditional text generation tasks: QQP, XSum, and Wiki-Auto. The results show that TEncDM exhibits superior performance compared to existing non-autoregressive diffusion models.

Recent multi-frame lifting methods have dominated the 3D human pose estimation. However, previous methods ignore the intricate dependence within the 2D pose sequence and learn single temporal correlation. To alleviate this limitation, we propose TCPFormer, which leverages an implicit pose proxy as an intermediate representation. Each proxy within the implicit pose proxy can build one temporal correlation therefore helping us learn a more comprehensive temporal correlation of human motion. Specifically, our method consists of three key components: Proxy Update Module (PUM), Proxy Invocation Module (PIM), and Proxy Attention Module (PAM). PUM first uses pose features to update the implicit pose proxy, enabling it to store representative information from the pose sequence. PIM then invocates and integrates the pose proxy with the pose sequence to enhance the motion semantics of each pose. Finally, PAM leverages the above mapping between the pose sequence and pose proxy to enhance the temporal correlation of the whole pose sequence. Experiments on the Human3.6M and MPI-INF-3DHP datasets demonstrate that our proposed TCPFormer outperforms the previous state-of-the-art methods.

Diffusion models have achieved remarkable success in the image and video generation tasks. Nevertheless, they often require a large amount of memory and time overhead during inference, due to the complex network architecture and considerable number of timesteps for iterative diffusion. Recently, the post-training quantization (PTQ) technique has proved a promising way to reduce the inference cost by quantizing the float-point operations to low-bit ones. However, most of them fail to tackle with the large variations in the distribution of activations across distinct channels and timesteps, as well as the inconsistent of input between quantization and inference on diffusion models, thus leaving much room for improvement. To address the above issues, we propose a novel method dubbed Timestep-Channel Adaptive Quantization for Diffusion Models (TCAQ-DM). Specifically, we develop a timestep-channel joint reparameterization (TCR) module to balance the activation range along both the timesteps and channels, facilitating the successive reconstruction procedure. Subsequently, we employ a dynamically adaptive quantization (DAQ) module that mitigate the quantization error by selecting an optimal quantizer for each post-Softmax layers according to their specific types of distributions. Moreover, we present a progressively aligned reconstruction (PAR) strategy to mitigate the bias caused by the input mismatch. Extensive experiments on various benchmarks and distinct diffusion models demonstrate that the proposed method substantially outperforms the state-of-the-art approaches in most cases, especially yielding comparable FID metrics to the full precision model on CIFAR-10 in the W6A6 setting, while enabling generating available images in the W4A4 settings.

We introduce TCAM-Diff, a novel 3D medical image generation model that reduces the memory requirements to encode and generate high-resolution 3D data. This model utilizes a decoder-only autoencoder method to learn triplane representation from dense volume and leverages generalization operations to prevent overfitting. Subsequently, it uses a triplane-aware cross-attention diffusion model to learn and integrate these features effectively. Furthermore, the features generated by the diffusion model can be rapidly transformed into 3D volumes using a pre-trained decoder module. Our experiments on three different scales of medical datasets, BrainTumour 128x128x128, Pancreas 256x256x256, and Colon 512x512x512, demonstrated outstanding results. We utilized MSE and SSIM to evaluate reconstruction quality and leveraged the Wasserstein Generative Adversarial Network (W-GAN) critic to assess generative quality. Comparisons to existing approaches show that our method gives better reconstruction and generation results than other encoder-decoder methods with similar-sized latent spaces.

Multimodal Large Language Models (MLLMs) have significantly improved performance across various image-language applications. Recently, there has been a growing interest in adapting image pre-trained MLLMs for video-related tasks. However, most efforts concentrate on enhancing the vision encoder and projector components, while the core part, Large Language Models (LLMs), remains comparatively under-explored. In this paper, we propose two strategies to enhance the model's capability in video understanding tasks by improving inter-layer attention computation in LLMs. Specifically, the first approach focuses on the enhancement of Rotary Position Embedding (RoPE) with Temporal-Aware Dual RoPE, which introduces temporal position information to strengthen the MLLM's temporal modeling capabilities while preserving the relative position relationships of both visual and text tokens. The second approach involves enhancing the Attention Mask with the Frame-wise Block Causal Attention Mask, a simple yet effective method that broadens visual token interactions within and across video frames while maintaining the causal inference mechanism. Based on these proposed methods, we adapt LLaVA for video understanding tasks, naming it Temporal-Considered LLaVA (TC-LLaVA). Our TC-LLaVA achieves new state-of-the-art performance across various video understanding benchmarks with only supervised fine-tuning (SFT) on video-related datasets.

Tropical cyclones (TCs) are complex weather systems with strong winds and heavy rainfall, causing substantial loss of life and property. Therefore, accurate TC forecasting is crucial for the effective prevention of disasters caused by TCs. TC forecasting can be regarded as a spatio-temporal prediction problem. It has been proven that using multi-modal data can effectively introduce atmospheric information to achieve better prediction results and higher interpretability. But it also introduces inevitably introduces noise into the prediction process. The diffusion model's unique noise modeling capability can reduce prediction noise when using multi-modal datasets. However, adapting it to TC forecasting has two main challenges: how to extract valuable information from multi-modal data, and how to utilize them to guide the generation process. For the first challenge, while recent methods can predict multiple TC attributes using multi-modal data, they often overlook the interdependence of multiple attributes and the semantic gap between modalities. Considering the interdependence of attributes, we propose two condition generators that capture the commonalities and characteristics of TC attributes, extracting spatio-temporal and environmental features and incorporating expert knowledge. To reduce the semantic gap between multi-modal data, we introduce the PGSA-LSTM module to map primary and auxiliary modalities. For the second challenge, we propose a novel Bi-condition diffusion model that sequentially processes conditions from the characteristics to commonalities of attributes, thereby expanding the guidance information that the diffusion model can accept. Our results surpass state-of-the-art deep learning models and outperform the numerical weather prediction model used by the China Central Meteorological Observatory. TC-Diffuser shows high generalizability across global ocean areas, strong robustness in handling missing data, and higher computational efficiency.

The concept of function and affordance is a critical aspect of 3D scene understanding and supports task-oriented objectives. In this work, we develop a model that learns to structure and vary functional affordance across a 3D hierarchical scene graph representing the spatial organization of a scene. The varying functional affordance is designed to integrate with the varying spatial context of the graph. More specifically, we develop an algorithm that learns to construct a 3D hierarchical scene graph (3DHSG) that captures the spatial organization of the scene. Starting from segmented object point clouds and object semantic labels, we develop a 3DHSG with a top node that identifies the room label, child nodes that define local spatial regions inside the room with region-specific affordances, and grand-child nodes indicating object locations and object-specific affordances. To support this work, we create a custom 3DHSG dataset that provides ground truth data for local spatial regions with region-specific affordances and also object-specific affordances for each object. We employ a Transformer Based Hierarchical Scene Understanding (TB-HSU) model to learn the 3DHSG. We use a multi-task learning framework that learns both room classification and learns to define spatial regions within the room with region-specific affordances. Our work improves on the performance of state-of-the-art baseline models and shows one approach for applying transformer models to 3D scene understanding and the generation of 3DHSGs that capture the spatial organization of a room. The code and dataset are publicly available.

Handwritten Mathematical Expression Recognition (HMER) has extensive applications in automated grading and office automation. However, existing sequence-based decoding methods, which directly predict LaTeX sequences, struggle to understand and model the inherent tree structure of LaTeX and often fail to ensure syntactic correctness in the decoded results. To address these challenges, we propose a novel model named TAMER (Tree-Aware Transformer) for handwritten mathematical expression recognition. TAMER introduces an innovative Tree-aware Module while maintaining the flexibility and efficient training of Transformer. TAMER combines the advantages of both sequence decoding and tree decoding models by jointly optimizing sequence prediction and tree structure prediction tasks, which enhances the model's understanding and generalization of complex mathematical expression structures. During inference, TAMER employs a Tree Structure Prediction Scoring Mechanism to improve the structural validity of the generated LaTeX sequences. Experimental results on CROHME datasets demonstrate that TAMER outperforms traditional sequence decoding and tree decoding models, especially in handling complex mathematical structures, achieving state-of-the-art (SOTA) performance.

This study addresses the challenge of detecting anomalies in multivariate time series data. Considering a bag (e.g., multi-sensor data) consisting of two-dimensional spaces of time points and multivariate instances (e.g., individual sensors), we aim to detect anomalies at both the bag and instance level with a unified model. To circumvent the practical difficulties of labeling at the instance level in such spaces, we adopt a multiple instance learning (MIL)-based approach, which enables learning at both the bag- and instance- levels using only the bag-level labels. In this study, we introduce time-aware and instance-learnable MIL (simply, TAIL-MIL). We propose two specialized attention mechanisms designed to effectively capture the relationships between different types of instances. We innovatively integrate these attention mechanisms with conjunctive pooling applied to the two-dimensional structure at different levels (i.e., bag- and instance-level), enabling TAIL-MIL to effectively pinpoint both the timing and causative multivariate factors of anomalies. We provide theoretical evidence demonstrating TAIL-MIL's efficacy in detecting instances with two-dimensional structures. Furthermore, we empirically validate the superior performance of TAIL-MIL over the state-of-the-art MIL methods and multivariate time-series anomaly detection methods.

The multi-instance multi-label (MIML) problem is a new supervised learning paradigm that has emerged to efficiently represent complex data. Therefore, various similarity-based algorithms have been proposed, but existing algorithms commonly measure similarity by considering only the structural relationships in the feature space without utilizing information from the label space. As these approaches do not adequately reflect the complex properties of MIML data, it is essential to improve the accuracy of MIML classification by utilizing information from both feature and label spaces. Thus, we propose a new algorithm, T-MDML: triplet-based multiple distance metric learning for MIML. T-MDML defines a distance metric by learning a global property shared by the entire label space and a label-specific property for each label. In addition, we simultaneously consider the structural characteristics of features and label space to extract label correlation and incorporate it into the optimization process. In experiments, we demonstrate the efficiency of our label correlation estimation method and verify its performance by applying it to MIMLkNN. We also demonstrate T-MDML’s relative superiority over existing MIML algorithms, as well as its scalability when applied to similarity-based MIML methods.

Small lesions play a critical role in early disease diagnosis and intervention of severe infections. Popular models often face challenges in segmenting small lesions, as it occupies only a minor portion of an image, while down-sampling operations may inevitably lose focus on local features of small lesions. To tackle the challenges, we propose a Small-Size-Sensitive Mamba (S³-Mamba), which promotes the sensitivity to small lesions across three dimensions: channel, spatial, and training strategy. Specifically, an Enhanced Visual State Space block is designed to focus on small lesions through multiple residual connections to preserve local features, and selectively amplify important details while suppressing irrelevant ones through channel-wise attention. A Tensor-based Cross-feature Multi-scale Attention is designed to integrate input image features and intermediate-layer features with edge features and exploit the attentive support of features across multiple scales, thereby retaining spatial details of small lesions at various granularities. Finally, we introduce a novel regularized curriculum learning to automatically assess lesion size and sample difficulty, and gradually focus from easy samples to hard ones like small lesions. Extensive experiments on three medical image segmentation datasets show the superiority of our S³-Mamba, especially in segmenting small lesions.

The task of music-to-lyric generation aims to create lyrics that can be sung in harmony with the music while capturing the music’s intrinsic meaning. Previous efforts in this area have struggled to effectively handle both the structural and semantic alignments of music and lyrics, often relying on rigid, manually crafted rules or overlooking the semantic essence of music, which deviates from the natural lyric-writing process of humans. In this paper, we bridge the structural and semantic gap between music and lyrics by proposing an end-to-end model for music-driven lyric generation. Our model aims at generating well-formatted lyrics based solely on the music while capturing its inherent semantic essence. In the music processing phase, we introduce a hierarchical music information extractor, which operates at both the song and sentence levels. The song-level extractor focuses on discerning the overall semantic content of the music, such as themes and emotions. Simultaneously, the sentence-level extractor captures the local semantic and structural details from note sequences. Additionally, we propose a lyric length predictor that determines the optimal length for the generated lyrics. During the lyric generation phase, the information gathered by the above modules is integrated, providing essential guidance for the downstream lyric generation module to produce coherent and meaningful lyrics. Experimental results on objective and subjective benchmarks demonstrate the capabilities of our proposed model in capturing semantics and generating well-formatted lyrics.

Inductive Knowledge Graph Completion (KGC) aims to infer missing facts between newly emerged entities within knowledge graphs (KGs), posing a significant challenge. While recent studies have shown promising results in inferring such entities through knowledge subgraph reasoning, they suffer from (i) the semantic inconsistencies of similar relations, and (ii) noisy interactions inherent in KGs due to the presence of unconvincing knowledge for emerging entities. To address these challenges, we propose a Semantic Structure-aware Denoising Network (S2DN) for inductive KGC. Our goal is to learn adaptable general semantics and reliable structures to distill consistent semantic knowledge while preserving reliable interactions within KGs. Specifically, we introduce a semantic smoothing module over the enclosing subgraphs to retain the universal semantic knowledge of relations. We incorporate a structure refining module to filter out unreliable interactions and offer additional knowledge, retaining robust structure surrounding target links. Extensive experiments conducted on three benchmark KGs demonstrate that S2DN surpasses the performance of state-of-the-art models. These results demonstrate the effectiveness of S2DN in preserving semantic consistency and enhancing the robustness of filtering out unreliable interactions in contaminated KGs.

Handling imbalance in class distribution when building a classifier over tabular data has been a problem of long-standing interest. One popular approach is augmenting the training dataset with synthetically generated data. While classical augmentation techniques were limited to linear interpolation of existing minority class examples, recently higher capacity deep generative models are providing greater promise. However, handling of imbalance in class distribution when building a deep generative model is also a challenging problem, that has not been studied as extensively as imbalanced classifier model training. We show that state-of-the-art deep generative models yield significantly lower-quality minority examples than majority examples. We propose a novel technique of converting the binary class labels to ternary class labels by introducing a class for the region where minority and majority distributions overlap. We show that just this pre-processing of the training set, significantly improves the quality of data generated spanning several state-of-the-art diffusion and GAN-based models. While training the classifier using synthetic data, we remove the overlap class from the training data and justify the reasons behind the enhanced accuracy. We perform extensive experiments on four real-life datasets, five different classifiers, and five generative models, demonstrating that our method enhances not only the synthesizer performance of state-of-the-art models but also the classifier performance

Antibodies defend our health by binding to antigens with high specificity and potentiality, primarily relying on the Complementarity-Determining Region (CDR). Yet, current experimental methods of discovering new antibody CDRs are heavily time-consuming. Computational design could alleviate this burden; especially, protein language models have proven quite beneficial in many recent studies. However, most existing models solely focus on antibody potentiality and struggle to encapsulate the diverse range of plausible CDR candidates, limiting their effectiveness in real-world scenarios as binding is only one factor in the multitude of drug-forming criteria. In this paper, we introduce PG-AbD, a framework uniting Generative Flow Networks (GFlowNets) and pretrained Protein Language Models (PLMs) to successfully generate highly potent, diverse and novel antibody candidates. We innovatively construct a Products of Experts (PoE) composed by the global-distribution-modeling PLM and the local-distribution-modeling Potts Model to serve as the reward function of GFlowNet. The joint training paradigm is introduced, where PoE is trained by contrastive divergence with the negative samples generated by GFlowNet, and then guides GFlowNet to sample diverse antibody candidates. We evaluate PG-AbD on extensive antibody design benchmarks. It significantly outperforms existing methods in diversity (13.5% on RabDab, 31.1% on SabDab) while maintaining optimal potential and novelty. Generated antibodies are also found to form stable, regular 3D structures with their corresponding antigens, demonstrating the great potential of PG-AbD to accelerate real-world antibody discovery.

Recent advancements in Large Language Models (LLMs) have led to significant breakthroughs in various natural language processing tasks. However, generating factually consistent responses in knowledge-intensive scenarios remains a challenge due to issues such as hallucination, difficulty in acquiring long-tailed knowledge, and limited memory expansion. This paper introduces SMART, a novel multi-agent framework that leverages external knowledge to enhance the interpretability and factual consistency of LLM-generated responses. SMART comprises four specialized agents, each performing a specific sub-trajectory action to navigate complex knowledge-intensive tasks. We propose a multi-agent co-training paradigm, Long-Short Trajectory Learning, which ensures synergistic collaboration among agents while maintaining fine-grained execution by each agent. Extensive experiments on five knowledge-intensive tasks demonstrate SMART's superior performance compared to widely adopted knowledge internalization and knowledge enhancement methods. Our framework can extend beyond knowledge-intensive tasks to more complex scenarios.