Robotic motor control necessitates the ability to predict the dynamics of environments and interaction objects. However, advanced self-supervised pre-trained visual representations (PVRs) in robotic motor control, leveraging large-scale egocentric videos, often focus solely on learning the static content features of sampled image frames. This neglects the crucial temporal motion clues in human video data, which implicitly contain key knowledge about sequential interacting and manipulating with the environments and objects. In this paper, we present a simple yet effective robotic motor control visual pre-training framework that jointly performs spatiotemporal predictive learning utilizing large-scale video data, termed as STP. Our STP samples paired frames from video clips. It adheres to two key designs in a multi-task learning manner. First, we perform spatial prediction on the masked current frame for learning content features. Second, we utilize the future frame with an extremely high masking ratio as a condition, based on the masked current frame, to conduct temporal prediction of future frame for capturing motion features. These efficient designs ensure that our representation focusing on motion information while capturing spatial details. We carry out the largest-scale evaluation of PVRs for robotic motor control to date, which encompasses 21 tasks within a real-world Franka robot arm and 5 simulated environments. Extensive experiments demonstrate the effectiveness of STP as well as unleash its generality and data efficiency by further post-pre-training and hybrid pre-training.
Modeling a generalized visuomotor policy has been a longstanding challenge for both computer vision and robotics communities. Existing approaches often fail to efficiently leverage cross-dataset resources or rely on heavy Vision-Language models, which require substantial computational resources, thereby limiting their multi-task performance and application potential. In this paper, we introduce a novel paradigm that effectively utilizes latent modeling of manipulation skills and an efficient visuomotor latent diffusion policy, which enhances the utilizing of existing cross-embodiment and cross-environment datasets, thereby improving multi-task capabilities. Our methodology consists of two decoupled phases: action modeling and policy modeling. Firstly, we introduce a task-agnostic, embodiment-aware trajectory latent autoencoder for unified action skills modeling. This step condenses action data and observation into a condensed latent space, effectively benefiting from large-scale cross-datasets. Secondly, we propose to use a visuomotor latent diffusion policy that recovers target skill latent from noises for effective task execution. We conducted extensive experiments on two widely used benchmarks, and the results demonstrate the effectiveness of our proposed paradigms on multi-tasking and pre-training. Code is available at https://github.com/AlbertTan404/RoLD.
Various Large Language Models(LLMs) from the Generative Pretrained Transformer~(GPT) family have achieved outstanding performances in a wide range of text generation tasks. However, the enormous model sizes have hindered their practical use in real-world applications due to high inference latency. Therefore, improving the efficiencies of LLMs through quantization, pruning, and other means has been a key issue in LLM studies. In this work, we propose a method based on Hessian sensitivity-aware mixed sparsity pruning to prune LLMs to at least 50\% sparsity without the need of any retraining. It allocates sparsity adaptively based on sensitivity, allowing us to reduce pruning-induced error while maintaining the overall sparsity level. The advantages of the proposed method exhibit even more when the sparsity is extremely high. Furthermore, our method is compatible with quantization, enabling further compression of LLMs.
Engaging video comments play an important role in video social media, as they are the carrier of feelings, thoughts, or humor of the audience. Preliminary works have made initial exploration for video comment generation by adopting caption-style encoder-decoder models. However, comment generation presents some unique challenges distinct from caption generation, which makes these methods somewhat less effective at generating engaging comments. In contrast to the objective and descriptive nature of captions, comments tend to be inherently subjective, making it hard to quantify and evaluate the engagement of comments. Furthermore, the scarcity of truly engaging comments brings difficulty to collecting enough high-quality training examples. In this paper, we propose ViCo with three novel designs to tackle the above challenges for generating engaging Video Comments. Firstly, to quantify the engagement of comments, we utilize the number of "likes" each comment receives as a proxy of human preference after an appropriate debiasing procedure. Secondly, to automatically evaluate the engagement of comments, we train a reward model to align its judgment to the above proxy. Our user studies indicate that this reward model effectively aligns with human judgments. Lastly, to alleviate the scarcity of high-quality comments, an initial generator is trained on readily available but noisy data to generate comments. Then the reward model is employed to offer feedback on the generated comments, thus optimizing the initial generator. To facilitate the research of video commenting, we collect a large video comment-dataset (ViCo-20k) with rich metadata from a popular video website. Experiments on ViCo-20k show that the comments generated by our ViCo model exhibit the best performance in terms of both quantitative and qualitative results, particularly when engagement is considered.
Training deep generative models usually requires a large amount of data. To alleviate the data collection cost, the task of zero-shot GAN adaptation aims to reuse well-trained generators to synthesize images of an unseen target domain without any further training samples. Due to the data absence, the textual description of the target domain and the vision-language models, e.g., CLIP, are utilized to effectively guide the generator. However, with only a single representative text feature instead of real images, the synthesized images gradually lose diversity as the model is optimized, which is also known as mode collapse. To tackle the problem, we propose a novel method to find semantic variations of the target text in the CLIP space. Specifically, we explore diverse semantic variations based on the informative text feature of the target domain while regularizing the uncontrolled deviation of the semantic information. With the obtained variations, we design a novel directional moment loss that matches the first and second moments of image and text direction distributions. Moreover, we introduce elastic weight consolidation and a relation consistency loss to effectively preserve valuable content information from the source domain, e.g., appearances. Through extensive experiments, we demonstrate the efficacy of the proposed methods in ensuring sample diversity in various scenarios of zero-shot GAN adaptation. We also conduct ablation studies to validate the effect of each proposed component. Notably, our model achieves a new state-of-the-art on zero-shot GAN adaptation in terms of both diversity and quality.
The exploration of the latent space in StyleGANs and GAN inversion exemplify impressive real-world image editing, yet the trade-off between reconstruction quality and editing quality remains an open problem. In this study, we revisit StyleGANs' hyperspherical prior $\mathcal{Z}$ and combine it with highly capable latent spaces to build combined spaces that faithfully invert real images while maintaining the quality of edited images. More specifically, we propose $\mathcal{F}/\mathcal{Z}^{+}$ space consisting of two subspaces: $\mathcal{F}$ space of an intermediate feature map of StyleGANs enabling faithful reconstruction and $\mathcal{Z}^{+}$ space of an extended StyleGAN prior supporting high editing quality. We project the real images into the proposed space to obtain the inverted codes, by which we then move along $\mathcal{Z}^{+}$, enabling semantic editing without sacrificing image quality. Comprehensive experiments show that $\mathcal{Z}^{+}$ can replace the most commonly-used $\mathcal{W}$, $\mathcal{W}^{+}$, and $\mathcal{S}$ spaces while preserving reconstruction quality, resulting in reduced distortion of edited images.
Large Pre-trained Transformers exhibit an intriguing capacity for in-context learning. Without gradient updates, these models can rapidly construct new predictors from demonstrations presented in the inputs. Recent works promote this ability in the vision-language domain by incorporating visual information into large language models that can already make in-context predictions. However, these methods could inherit issues in the language domain, such as template sensitivity and hallucination. Also, the scale of these language models raises a significant demand for computations, making learning and operating these models resource-intensive. To this end, we raise a question: ``How can we enable in-context learning for general models without being constrained on large language models?". To answer it, we propose a succinct and general framework, Self-supervised IN-Context learning (SINC), that introduces a meta-model to learn on self-supervised prompts consisting of tailored demonstrations. The learned models can be transferred to downstream tasks for making in-context predictions on-the-fly. Extensive experiments show that SINC outperforms gradient-based methods in various vision-language tasks under few-shot settings. Furthermore, the designs of SINC help us investigate the benefits of in-context learning across different tasks, and the analysis further reveals the essential components for the emergence of in-context learning in the vision-language domain.
Improving the generalization capabilities of general-purpose robotic agents has long been a significant challenge actively pursued by research communities. Existing approaches often rely on collecting large-scale real-world robotic data, such as the RT-1 dataset. However, these approaches typically suffer from low efficiency, limiting their capability in open-domain scenarios with new objects, and diverse backgrounds. In this paper, we propose a novel paradigm that effectively leverages language-grounded segmentation masks generated by state-of-the-art foundation models, to address a wide range of pick-and-place robot manipulation tasks in everyday scenarios. By integrating precise semantics and geometries conveyed from masks into our multi-view policy model, our approach can perceive accurate object poses and enable sample-efficient learning. Besides, such design facilitates effective generalization for grasping new objects with similar shapes observed during training. Our approach consists of two distinct steps. First, we introduce a series of foundation models to accurately ground natural language demands across multiple tasks. Second, we develop a Multi-modal Multi-view Policy Model that incorporates inputs such as RGB images, semantic masks, and robot proprioception states to jointly predict precise and executable robot actions. Extensive real-world experiments conducted on a Franka Emika robot arm validate the effectiveness of our proposed paradigm. Real-world demos are shown in YouTube (https://www.youtube.com/watch?v=1m9wNzfp_4E ) and Bilibili (https://www.bilibili.com/video/BV178411Z7H2/ ).