Related papers: PISA Experiments: Exploring Physics Post-Training …
Vision-Language-Action (VLA) models trained via imitation learning suffer from significant performance degradation in data-scarce scenarios due to their reliance on large-scale demonstration datasets. Although reinforcement learning…
AI-generated content has attracted lots of attention recently, but photo-realistic video synthesis is still challenging. Although many attempts using GANs and autoregressive models have been made in this area, the visual quality and length…
When do diffusion models reproduce their training data, and when are they able to generate samples beyond it? A practically relevant theoretical understanding of this interplay between memorization and generalization may significantly…
Text-to-video diffusion models have enabled high-quality video synthesis, yet often fail to generate temporally coherent and physically plausible motion. A key reason is the models' insufficient understanding of complex motions that natural…
Motion models (i.e., transition probability densities) are often deduced from fluorescence widefield tracking experiments by analyzing single-particle trajectories post-processed from data. This analysis immediately raises the question: To…
The ability to learn manipulation skills by watching videos of humans has the potential to unlock a new source of highly scalable data for robot learning. Here, we tackle prehensile manipulation, in which tasks involve grasping an object…
Pre-training is a widely used approach to develop models that are robust to distribution shifts. However, in practice, its effectiveness varies: fine-tuning a pre-trained model improves robustness significantly in some cases but not at all…
Diffusion models have been widely studied for removing unsafe content learned during pre-training. Existing methods require expensive supervised data, either unsafe-text paired with safe-image groundtruth or negative/positive image pairs,…
Contemporary Video Instance Segmentation (VIS) methods typically adhere to a pre-train then fine-tune regime, where a segmentation model trained on images is fine-tuned on videos. However, the lack of temporal knowledge in the pre-trained…
Recent diffusion models have exhibited great potential in generative modeling tasks. Part of their success can be attributed to the ability of training stable on huge sets of paired synthetic data. However, adapting these models to…
Training generalist policies for robotic manipulation has shown great promise, as they enable language-conditioned, multi-task behaviors across diverse scenarios. However, evaluating these policies remains difficult because real-world…
Recent advances in internet-scale video data pretraining have led to the development of text-to-video generative models that can create high-quality videos across a broad range of visual concepts, synthesize realistic motions and render…
Recent advancements in diffusion models have revolutionized video generation, enabling the creation of high-quality, temporally consistent videos. However, generating high frame-rate (FPS) videos remains a significant challenge due to…
World models simulate future states of the world in response to different actions. They facilitate interactive content creation and provides a foundation for grounded, long-horizon reasoning. Current foundation models do not fully meet the…
Video generation models are increasingly used as world simulators for tasks like driving and robotic manipulation. What matters in these settings is not whether a single video looks right, but whether the model's output changes when its…
We investigate the emergence of intuitive physics understanding in general-purpose deep neural network models trained to predict masked regions in natural videos. Leveraging the violation-of-expectation framework, we find that video…
End-to-end self-supervised models have been proposed for estimating the success of future candidate grasps and video predictive models for generating future observations. However, none have yet studied these two strategies side-by-side for…
We propose a framework to continuously learn object-centric representations for visual learning and understanding. Existing object-centric representations either rely on supervisions that individualize objects in the scene, or perform…
Large-scale video generation models have demonstrated emergent physical coherence, positioning them as potential world models. However, a gap remains between contemporary "stateless" video architectures and classic state-centric world model…
While large text-to-image models are able to synthesize "novel" images, these images are necessarily a reflection of the training data. The problem of data attribution in such models -- which of the images in the training set are most…