2026-03-31 AI 리서치 브리핑

Trace2Skill: Distill Trajectory-Local Lessons into Transferable Agent Skills에 관한 최근 업데이트입니다. 자세한 내용은 원문 링크에서 확인할 수 있습니다.

Most recent advances in 3D generative modeling rely on diffusion or flow-matching formulations. We instead explore a fully autoregressive alternative and introduce GaussianGPT, a transformer-based model that directly generates 3D Gaussians via next-token prediction, thus facilitating full 3D scene generation. We first compress Gaussian primitives into a discrete latent grid using a sparse 3D convolutional autoencoder with vector quantization. The resulting tokens are serialized and modeled using a causal transformer with 3D rotary positional embedding, enabling sequential generation of spatial structure and appearance. Unlike diffusion-based methods that refine scenes holistically, our formulation constructs scenes step-by-step, naturally supporting completion, outpainting, controllable sampling via temperature, and flexible generation horizons. This formulation leverages the compositional inductive biases and scalability of autoregressive modeling while operating on explicit representations compatible with modern neural rendering pipelines, positioning autoregressive transformers as a complementary paradigm for controllable and context-aware 3D generation.

Out of Sight but Not Out of Mind: Hybrid Memory for Dynamic Video World Models에 관한 최근 업데이트입니다. 자세한 내용은 원문 링크에서 확인할 수 있습니다.

ShotStream: Streaming Multi-Shot Video Generation for Interactive Storytelling에 관한 최근 업데이트입니다. 자세한 내용은 원문 링크에서 확인할 수 있습니다.

PackForcing: Short Video Training Suffices for Long Video Sampling and Long Context Inference에 관한 최근 업데이트입니다. 자세한 내용은 원문 링크에서 확인할 수 있습니다.

MedOpenClaw: Auditable Medical Imaging Agents Reasoning over Uncurated Full Studies에 관한 최근 업데이트입니다. 자세한 내용은 원문 링크에서 확인할 수 있습니다.

Reconstructing 3D geometry and appearance from a sparse set of fixed cameras is a foundational task with broad applications, yet it remains fundamentally constrained by the limited viewpoints. We show that this bound can be broken by exploiting opportunistic object motion: as a person manipulates an object~(e.g., moving a chair or lifting a mug), the static cameras effectively ``orbit'' the object in its local coordinate frame, providing additional virtual viewpoints. Harnessing this object motion, however, poses two challenges: the tight coupling of object pose and geometry estimation and the complex appearance variations of a moving object under static illumination. We address these by formulating a joint pose and shape optimization using 2D Gaussian splatting with alternating minimization of 6DoF trajectories and primitive parameters, and by introducing a novel appearance model that factorizes diffuse and specular components with reflected directional probing within the spherical harmonics space. Extensive experiments on synthetic and real-world datasets with extremely sparse viewpoints demonstrate that our method recovers significantly more accurate geometry and appearance than state-of-the-art baselines.

Depth from Defocus (DfD) is the task of estimating a dense metric depth map from a focus stack. Unlike previous works overfitting to a certain dataset, this paper focuses on the challenging and practical setting of zero-shot generalization. We first propose a new real-world DfD benchmark ZEDD, which contains 8.3x more scenes and significantly higher quality images and ground-truth depth maps compared to previous benchmarks. We also design a novel network architecture named FOSSA. FOSSA is a Transformer-based architecture with novel designs tailored to the DfD task. The key contribution is a stack attention layer with a focus distance embedding, allowing efficient information exchange across the focus stack. Finally, we develop a new training data pipeline allowing us to utilize existing large-scale RGBD datasets to generate synthetic focus stacks. Experiment results on ZEDD and other benchmarks show a significant improvement over the baselines, reducing errors by up to 55.7%. The ZEDD benchmark is released at https://zedd.cs.princeton.edu. The code and checkpoints are released at https://github.com/princeton-vl/FOSSA.

Equivariance is a fundamental property in computer vision models, yet strict equivariance is rarely satisfied in real-world data, which can limit a model's performance. Controlling the degree of equivariance is therefore desirable. We propose a general framework for constructing soft equivariant models by projecting the model weights into a designed subspace. The method applies to any pre-trained architecture and provides theoretical bounds on the induced equivariance error. Empirically, we demonstrate the effectiveness of our method on multiple pre-trained backbones, including ViT and ResNet, across image classification, semantic segmentation, and human-trajectory prediction tasks. Notably, our approach improves the performance while simultaneously reducing equivariance error on the competitive ImageNet benchmark.

We introduce PerceptionComp, a manually annotated benchmark for complex, long-horizon, perception-centric video reasoning. PerceptionComp is designed so that no single moment is sufficient: answering each question requires multiple temporally separated pieces of visual evidence and compositional constraints under conjunctive and sequential logic, spanning perceptual subtasks such as objects, attributes, relations, locations, actions, and events, and requiring skills including semantic recognition, visual correspondence, temporal reasoning, and spatial reasoning. The benchmark contains 1,114 highly complex questions on 279 videos from diverse domains including city walk tours, indoor villa tours, video games, and extreme outdoor sports, with 100% manual annotation. Human studies show that PerceptionComp requires substantial test-time thinking and repeated perception steps: participants take much longer than on prior benchmarks, and accuracy drops to near chance (18.97%) when rewatching is disallowed. State-of-the-art MLLMs also perform substantially worse on PerceptionComp than on existing benchmarks: the best model in our evaluation, Gemini-3-Flash, reaches only 45.96% accuracy in the five-choice setting, while open-source models remain below 40%. These results suggest that perception-centric long-horizon video reasoning remains a major bottleneck, and we hope PerceptionComp will help drive progress in perceptual reasoning.