Yaoru Sun

Siyu Zhang, Yeming Chen, Yaoru Sun et al.

Visual question answering (VQA) has been intensively studied as a multimodal task that requires effort in bridging vision and language to infer answers correctly. Recent attempts have developed various attention-based modules for solving VQA tasks. However, the performance of model inference is largely bottlenecked by visual processing for semantics understanding. Most existing detection methods rely on bounding boxes, remaining a serious challenge for VQA models to understand the causal nexus of object semantics in images and correctly infer contextual information. To this end, we propose a finer model framework without bounding boxes in this work, termed Looking Out of Instance Semantics (LOIS) to tackle this important issue. LOIS enables more fine-grained feature descriptions to produce visual facts. Furthermore, to overcome the label ambiguity caused by instance masks, two types of relation attention modules: 1) intra-modality and 2) inter-modality, are devised to infer the correct answers from the different multi-view features. Specifically, we implement a mutual relation attention module to model sophisticated and deeper visual semantic relations between instance objects and background information. In addition, our proposed attention model can further analyze salient image regions by focusing on important word-related questions. Experimental results on four benchmark VQA datasets prove that our proposed method has favorable performance in improving visual reasoning capability.

Haoran Wang, Zeshen Tang, Leya Yang et al.

Goal-conditioned hierarchical reinforcement learning (HRL) presents a promising approach for enabling effective exploration in complex, long-horizon reinforcement learning (RL) tasks through temporal abstraction. Empirically, heightened inter-level communication and coordination can induce more stable and robust policy improvement in hierarchical systems. Yet, most existing goal-conditioned HRL algorithms have primarily focused on the subgoal discovery, neglecting inter-level cooperation. Here, we propose a goal-conditioned HRL framework named Guided Cooperation via Model-based Rollout (GCMR), aiming to bridge inter-layer information synchronization and cooperation by exploiting forward dynamics. Firstly, the GCMR mitigates the state-transition error within off-policy correction via model-based rollout, thereby enhancing sample efficiency. Secondly, to prevent disruption by the unseen subgoals and states, lower-level Q-function gradients are constrained using a gradient penalty with a model-inferred upper bound, leading to a more stable behavioral policy conducive to effective exploration. Thirdly, we propose a one-step rollout-based planning, using higher-level critics to guide the lower-level policy. Specifically, we estimate the value of future states of the lower-level policy using the higher-level critic function, thereby transmitting global task information downwards to avoid local pitfalls. These three critical components in GCMR are expected to facilitate inter-level cooperation significantly. Experimental results demonstrate that incorporating the proposed GCMR framework with a disentangled variant of HIGL, namely ACLG, yields more stable and robust policy improvement compared to various baselines and significantly outperforms previous state-of-the-art algorithms.

Jun Yang, Lizhi Bai, Yaoru Sun et al.

RGB-D semantic segmentation can be advanced with convolutional neural networks due to the availability of Depth data. Although objects cannot be easily discriminated by just the 2D appearance, with the local pixel difference and geometric patterns in Depth, they can be well separated in some cases. Considering the fixed grid kernel structure, CNNs are limited to lack the ability to capture detailed, fine-grained information and thus cannot achieve accurate pixel-level semantic segmentation. To solve this problem, we propose a Pixel Difference Convolutional Network (PDCNet) to capture detailed intrinsic patterns by aggregating both intensity and gradient information in the local range for Depth data and global range for RGB data, respectively. Precisely, PDCNet consists of a Depth branch and an RGB branch. For the Depth branch, we propose a Pixel Difference Convolution (PDC) to consider local and detailed geometric information in Depth data via aggregating both intensity and gradient information. For the RGB branch, we contribute a lightweight Cascade Large Kernel (CLK) to extend PDC, namely CPDC, to enjoy global contexts for RGB data and further boost performance. Consequently, both modal data's local and global pixel differences are seamlessly incorporated into PDCNet during the information propagation process. Experiments on two challenging benchmark datasets, i.e., NYUDv2 and SUN RGB-D reveal that our PDCNet achieves state-of-the-art performance for the semantic segmentation task.

Lizhi Bai, Jun Yang, Chunqi Tian et al.

Combining RGB images and the corresponding depth maps in semantic segmentation proves the effectiveness in the past few years. Existing RGB-D modal fusion methods either lack the non-linear feature fusion ability or treat both modal images equally, regardless of the intrinsic distribution gap or information loss. Here we find that depth maps are suitable to provide intrinsic fine-grained patterns of objects due to their local depth continuity, while RGB images effectively provide a global view. Based on this, we propose a pixel differential convolution attention (DCA) module to consider geometric information and local-range correlations for depth data. Furthermore, we extend DCA to ensemble differential convolution attention (EDCA) which propagates long-range contextual dependencies and seamlessly incorporates spatial distribution for RGB data. DCA and EDCA dynamically adjust convolutional weights by pixel difference to enable self-adaptive in local and long range, respectively. A two-branch network built with DCA and EDCA, called Differential Convolutional Network (DCANet), is proposed to fuse local and global information of two-modal data. Consequently, the individual advantage of RGB and depth data are emphasized. Our DCANet is shown to set a new state-of-the-art performance for RGB-D semantic segmentation on two challenging benchmark datasets, i.e., NYUDv2 and SUN-RGBD.

Guorun Wang, Jun Yang, Yaoru Sun

The Outstanding performance and growing size of Large Language Models has led to increased attention in parameter efficient learning. The two predominant approaches are Adapters and Pruning. Adapters are to freeze the model and give it a new weight matrix on the side, which can significantly reduce the time and memory of training, but the cost is that the evaluation and testing will increase the time and memory consumption. Pruning is to cut off some weight and re-distribute the remaining weight, which sacrifices the complexity of training at the cost of extremely high memory and training time, making the cost of evaluation and testing relatively low. So efficiency of training and inference can't be obtained in the same time. In this work, we propose a task-oriented Pruning-Adapter method that achieve a high memory efficiency of training and memory, and speeds up training time and ensures no significant decrease in accuracy in GLUE tasks, achieving training and inference efficiency at the same time.

Zheyuan Yang, Liqiang Shang, Junjie Chen et al.

We introduce TableVista, a comprehensive benchmark for evaluating foundation models in multimodal table reasoning under visual and structural complexity. TableVista consists of 3,000 high-quality table reasoning problems, where each instance is expanded into 10 distinct visual variants through our multi-style rendering and transformation pipeline. This process encompasses diverse scenario styles, robustness perturbations, and vision-only configurations, culminating in 30,000 multimodal samples for a multi-dimensional evaluation. We conduct an extensive evaluation of 29 state-of-the-art open-source and proprietary foundation models on TableVista. Through comprehensive quantitative and qualitative analysis, we find that while evaluated models remain largely stable across diverse rendering styles, they exhibit pronounced performance degradation on complex structural layouts and vision-only settings, revealing that current models struggle to maintain reasoning consistency when structural complexity combines with visually integrated presentations. These findings highlight critical gaps in current multimodal capabilities, providing insights for advancing more robust and reliable table understanding models.

Yeming Chen, Siyu Zhang, Yaoru Sun et al.

With the success of self-supervised learning, multimodal foundation models have rapidly adapted a wide range of downstream tasks driven by vision and language (VL) pretraining. State-of-the-art methods achieve impressive performance by pre-training on large-scale datasets. However, bridging the semantic gap between the two modalities remains a nonnegligible challenge for VL tasks. In this work, we propose an efficient computation framework for multimodal alignment by introducing a novel visual semantic module to further improve the performance of the VL tasks. Specifically, we propose a flexible model, namely Artificial-Spiking Hierarchical Networks (ASH-Nets), which combines the complementary advantages of Artificial neural networks (ANNs) and Spiking neural networks (SNNs) to enrich visual semantic representations. In particular, a visual concrete encoder and a semantic abstract encoder are constructed to learn continuous and discrete latent variables to enhance the flexibility of semantic encoding. Considering the spatio-temporal properties of SNNs modeling, we introduce a contrastive learning method to optimize the inputs of similar samples. This can improve the computational efficiency of the hierarchical network, while the augmentation of hard samples is beneficial to the learning of visual representations. Furthermore, the Spiking to Text Uni-Alignment Learning (STUA) pre-training method is proposed, which only relies on text features to enhance the encoding ability of abstract semantics. We validate the performance on multiple well-established downstream VL tasks. Experiments show that the proposed ASH-Nets achieve competitive results.

Siyu Zhang, Yeming Chen, Yaoru Sun et al.

The key to integrating visual language tasks is to establish a good alignment strategy. Recently, visual semantic representation has achieved fine-grained visual understanding by dividing grids or image patches. However, the coarse-grained semantic interactions in image space should not be ignored, which hinders the extraction of complex contextual semantic relations at the scene boundaries. This paper proposes superpixels as comprehensive and robust visual primitives, which mine coarse-grained semantic interactions by clustering perceptually similar pixels, speeding up the subsequent processing of primitives. To capture superpixel-level semantic features, we propose a Multiscale Difference Graph Convolutional Network (MDGCN). It allows parsing the entire image as a fine-to-coarse visual hierarchy. To reason actual semantic relations, we reduce potential noise interference by aggregating difference information between adjacent graph nodes. Finally, we propose a multi-level fusion rule in a bottom-up manner to avoid understanding deviation by mining complementary spatial information at different levels. Experiments show that the proposed method can effectively promote the learning of multiple downstream tasks. Encouragingly, our method outperforms previous methods on all metrics. Our code will be released upon publication.

Xinfeng Zhao, Yaoru Sun

The Vision Transformer (ViT) has made significant strides in the field of computer vision. However, as the depth of the model and the resolution of the input images increase, the computational cost associated with training and running ViT models has surged dramatically. This paper proposes a hybrid model based on CNN and Vision Transformer, named CI2P-ViT. The model incorporates a module called CI2P, which utilizes the CompressAI encoder to compress images and subsequently generates a sequence of patches through a series of convolutions. CI2P can replace the Patch Embedding component in the ViT model, enabling seamless integration into existing ViT models. Compared to ViT-B/16, CI2P-ViT has the number of patches input to the self-attention layer reduced to a quarter of the original. This design not only significantly reduces the computational cost of the ViT model but also effectively enhances the model's accuracy by introducing the inductive bias properties of CNN. The ViT model's precision is markedly enhanced. When trained from the ground up on the Animals-10 dataset, CI2P-ViT achieved an accuracy rate of 92.37%, representing a 3.3% improvement over the ViT-B/16 baseline. Additionally, the model's computational operations, measured in floating-point operations per second (FLOPs), were diminished by 63.35%, and it exhibited a 2-fold increase in training velocity on identical hardware configurations.

Haoran Wang, Yaoru Sun, Zeshen Tang et al.

Goal-conditioned hierarchical reinforcement learning (HRL) decomposes complex reaching tasks into a sequence of simple subgoal-conditioned tasks, showing significant promise for addressing long-horizon planning in large-scale environments. This paper bridges the goal-conditioned HRL based on graph-based planning to brain mechanisms, proposing a hippocampus-striatum-like dual-controller hypothesis. Inspired by the brain mechanisms of organisms (i.e., the high-reward preferences observed in hippocampal replay) and instance-based theory, we propose a high-return sampling strategy for constructing memory graphs, improving sample efficiency. Additionally, we derive a model-free lower-level Q-function gradient penalty to resolve the model dependency issues present in prior work, improving the generalization of Lipschitz constraints in applications. Finally, we integrate these two extensions, High-reward Graph and model-free Gradient Penalty (HG2P), into the state-of-the-art framework ACLG, proposing a novel goal-conditioned HRL framework, HG2P+ACLG. Experimentally, the results demonstrate that our method outperforms state-of-the-art goal-conditioned HRL algorithms on a variety of long-horizon navigation tasks and robotic manipulation tasks.