In end-to-end speech translation, acoustic representations learned by the encoder are usually fixed and static, from the perspective of the decoder, which is not desirable for dealing with the cross-modal and cross-lingual challenge in speech translation. In this paper, we show the benefits of varying acoustic states according to decoder hidden states and propose an adaptive speech-to-text translation model that is able to dynamically adapt acoustic states in the decoder. We concatenate the acoustic state and target word embedding sequence and feed the concatenated sequence into subsequent blocks in the decoder. In order to model the deep interaction between acoustic states and target hidden states, a speech-text mixed attention sublayer is introduced to replace the conventional cross-attention network. Experiment results on two widely-used datasets show that the proposed method significantly outperforms state-of-the-art neural speech translation models.

In robotic manipulation tasks, achieving a designated target state for the manipulated object is often essential to facilitate motion planning for robotic arms. Specifically, in tasks such as hanging a mug, the mug must be positioned within a feasible region around the hook. Previous approaches have enabled the generation of multiple feasible target states for mugs; however, these target states are typically generated randomly, lacking control over the specific generation locations. This limitation makes such methods less effective in scenarios where constraints exist, such as hooks already occupied by other mugs or when specific operational objectives must be met. Moreover, due to the frequent physical interactions between the mug and the rack in real-world hanging scenarios, imprecisely generated target states from end-to-end models often result in overlapping point clouds. This overlap adversely impacts subsequent motion planning for the robotic arm. To address these challenges, we propose a Linguistically Guided Hybrid Gaussian Diffusion (LHGD) network for generating manipulation target states, combined with a gravity coverage coefficient-based method for target state refinement. To evaluate our approach under a language-specified distribution setting, we collected multiple feasible target states for 10 types of mugs across 5 different racks with 10 distinct hooks. Additionally, we prepared five unseen mug designs for validation purposes. Experimental results demonstrate that our method achieves the highest success rates across single-mode, multi-mode, and language-specified distribution manipulation tasks. Furthermore, it significantly reduces point cloud overlap, directly producing collision-free target states and eliminating the need for additional obstacle avoidance operations by the robotic arm.

Distinguishing spatial relations is a basic part of human cognition which requires fine-grained perception on cross-instance. Although benchmarks like MME, MMBench and SEED comprehensively have evaluated various capabilities which already include visual spatial reasoning(VSR). There is still a lack of sufficient quantity and quality evaluation and optimization datasets for Vision Large Language Models(VLLMs) specifically targeting visual positional reasoning. To handle this, we first diagnosed current VLLMs with the VSR dataset and proposed a unified test set. We found current VLLMs to exhibit a contradiction of over-sensitivity to language instructions and under-sensitivity to visual positional information. By expanding the original benchmark from two aspects of tunning data and model structure, we mitigated this phenomenon. To our knowledge, we expanded spatially positioned image data controllably using diffusion models for the first time and integrated original visual encoding(CLIP) with other 3 powerful visual encoders(SigLIP, SAM and DINO). After conducting combination experiments on scaling data and models, we obtained a VLLM VSR Expert(VSRE) that not only generalizes better to different instructions but also accurately distinguishes differences in visual positional information. VSRE achieved over a 27\% increase in accuracy on the VSR test set. It becomes a performant VLLM on the position reasoning of both the VSR dataset and relevant subsets of other evaluation benchmarks. We open-sourced the expanded model with data and Appendix at \url{https://github.com/peijin360/vsre} and hope it will accelerate advancements in VLLM on VSR learning.

Decision-making in robotics using denoising diffusion processes has increasingly become a hot research topic, but end-to-end policies perform poorly in tasks with rich contact and have limited controllability. This paper proposes Hierarchical Diffusion Policy (HDP), a new imitation learning method of using objective contacts to guide the generation of robot trajectories. The policy is divided into two layers: the high-level policy predicts the contact for the robot's next object manipulation based on 3D information, while the low-level policy predicts the action sequence toward the high-level contact based on the latent variables of observation and contact. We represent both level policies as conditional denoising diffusion processes, and combine behavioral cloning and Q-learning to optimize the low level policy for accurately guiding actions towards contact. We benchmark Hierarchical Diffusion Policy across 6 different tasks and find that it significantly outperforms the existing state of-the-art imitation learning method Diffusion Policy with an average improvement of 20.8%. We find that contact guidance yields significant improvements, including superior performance, greater interpretability, and stronger controllability, especially on contact-rich tasks. To further unlock the potential of HDP, this paper proposes a set of key technical contributions including snapshot gradient optimization, 3D conditioning, and prompt guidance, which improve the policy's optimization efficiency, spatial awareness, and controllability respectively. Finally, real world experiments verify that HDP can handle both rigid and deformable objects.

Vision-language pre-training (VLP) on large-scale datasets has shown premier performance on various downstream tasks. In contrast to plenty of available benchmarks with English corpus, large-scale pre-training datasets and downstream datasets with Chinese corpus remain largely unexplored. In this work, we build a large-scale high-quality Chinese Cross-Modal Benchmark named CCMB for the research community, which contains the currently largest public pre-training dataset Zero and five human-annotated fine-tuning datasets for downstream tasks. Zero contains 250 million images paired with 750 million text descriptions, plus two of the five fine-tuning datasets are also currently the largest ones for Chinese cross-modal downstream tasks. Along with the CCMB, we also develop a VLP framework named R2D2, applying a pre-Ranking + Ranking strategy to learn powerful vision-language representations and a two-way distillation method (i.e., target-guided Distillation and feature-guided Distillation) to further enhance the learning capability. With the Zero and the R2D2 VLP framework, we achieve state-of-the-art performance on twelve downstream datasets from five broad categories of tasks including image-text retrieval, image-text matching, image caption, text-to-image generation, and zero-shot image classification. The datasets, models, and codes are available at https://github.com/yuxie11/R2D2

Manipulating objects without grasping them enables more complex tasks, known as non-prehensile manipulation. Most previous methods only learn one manipulation skill, such as reach or push, and cannot achieve flexible object manipulation.In this work, we introduce MRLM, a Multi-stage Reinforcement Learning approach for non-prehensile Manipulation of objects.MRLM divides the task into multiple stages according to the switching of object poses and contact points.At each stage, the policy takes the point cloud-based state-goal fusion representation as input, and proposes a spatially-continuous action that including the motion of the parallel gripper pose and opening width.To fully unlock the potential of MRLM, we propose a set of technical contributions including the state-goal fusion representation, spatially-reachable distance metric, and automatic buffer compaction.We evaluate MRLM on an Occluded Grasping task which aims to grasp the object in configurations that are initially occluded.Compared with the baselines, the proposed technical contributions improve the success rate by at least 40\% and maximum 100\%, and avoids falling into local optimum.Our method demonstrates strong generalization to unseen object with shapes outside the training distribution.Moreover, MRLM can be transferred to real world with zero-shot transfer, achieving a 95\% success rate.Code and videos can be found at https://sites.google.com/view/mrlm.

Grasp detection in cluttered scenes is a very challenging task for robots. Generating synthetic grasping data is a popular way to train and test grasp methods, as is Dex-net and GraspNet; yet, these methods generate training grasps on 3D synthetic object models, but evaluate at images or point clouds with different distributions, which reduces performance on real scenes due to sparse grasp labels and covariate shift. To solve existing problems, we propose a novel on-policy grasp detection method, which can train and test on the same distribution with dense pixel-level grasp labels generated on RGB-D images. A Parallel-Depth Grasp Generation (PDG-Generation) method is proposed to generate a parallel depth image through a new imaging model of projecting points in parallel; then this method generates multiple candidate grasps for each pixel and obtains robust grasps through flatness detection, force-closure metric and collision detection. Then, a large comprehensive Pixel-Level Grasp Pose Dataset (PLGP-Dataset) is constructed and released; distinguished with previous datasets with off-policy data and sparse grasp samples, this dataset is the first pixel-level grasp dataset, with the on-policy distribution where grasps are generated based on depth images. Lastly, we build and test a series of pixel-level grasp detection networks with a data augmentation process for imbalance training, which learn grasp poses in a decoupled manner on the input RGB-D images. Extensive experiments show that our on-policy grasp method can largely overcome the gap between simulation and reality, and achieves the state-of-the-art performance. Code and data are provided at https://github.com/liuchunsense/PLGP-Dataset.

Visual context provides grounding information for multimodal machine translation (MMT). However, previous MMT models and probing studies on visual features suggest that visual information is less explored in MMT as it is often redundant to textual information. In this paper, we propose an object-level visual context modeling framework (OVC) to efficiently capture and explore visual information for multimodal machine translation. With detected objects, the proposed OVC encourages MMT to ground translation on desirable visual objects by masking irrelevant objects in the visual modality. We equip the proposed with an additional object-masking loss to achieve this goal. The object-masking loss is estimated according to the similarity between masked objects and the source texts so as to encourage masking source-irrelevant objects. Additionally, in order to generate vision-consistent target words, we further propose a vision-weighted translation loss for OVC. Experiments on MMT datasets demonstrate that the proposed OVC model outperforms state-of-the-art MMT models and analyses show that masking irrelevant objects helps grounding in MMT.