09 Dec 2025
Wan-Move presents a framework for motion-controllable video generation that utilizes latent trajectory guidance to directly edit image condition features within a pre-trained image-to-video model. This method yields superior visual quality and precise motion adherence compared to state-of-the-art academic approaches and rivals commercial solutions, while also establishing MoveBench, a new comprehensive evaluation benchmark.
06 Dec 2025
This paper presents the Universal Weight Subspace Hypothesis, demonstrating empirically that deep neural networks trained across diverse tasks and modalities converge to shared low-dimensional parametric subspaces. This convergence enables significant memory savings, such as up to 100x for Vision Transformers and LLaMA models, and 19x for LoRA adapters, while preserving model performance and enhancing efficiency in model merging and adaptation.
08 Dec 2025
Researchers from Fudan University and Shanghai Innovation Institute introduced RoPE++, an extension of Rotary Position Embeddings that re-incorporates the previously discarded imaginary component of attention scores to improve long-context modeling in Large Language Models. This method consistently outperforms standard RoPE on various benchmarks and offers significant KV-cache and parameter efficiency.
08 Dec 2025
Apple researchers introduced FAE (Feature Auto-Encoder), a minimalist framework using a single attention layer and a double-decoder architecture to adapt high-dimensional self-supervised visual features into compact, generation-friendly latent spaces. FAE achieves competitive FID scores on ImageNet (1.29) and MS-COCO (6.90) for image generation while preserving semantic understanding capabilities of the original pre-trained encoders.
08 Dec 2025
The DEMOCRITUS system establishes a new framework for building large causal models (LCMs) by extracting and structuring textual knowledge from Large Language Models (LLMs) across diverse domains. It leverages a Geometric Transformer to embed and organize vast causal claims into coherent, navigable manifolds, which, unlike raw LLM outputs, exhibit global causal coherence and interpretable local structures.
09 Dec 2025
Terrain Diffusion introduces a diffusion-based framework for generating infinite, real-time procedural terrain, delivering highly realistic, boundless virtual worlds with seed-consistency and constant-time random access. The system achieves competitive FID scores and real-time generation latency on consumer hardware, demonstrating its practical applicability.
09 Dec 2025
Researchers from Google DeepMind, University College London, and the University of Oxford developed D4RT, a unified feedforward model for reconstructing dynamic 4D scenes, encompassing depth, spatio-temporal correspondence, and camera parameters, from video using a single, flexible querying interface. The model achieved state-of-the-art accuracy across various 4D reconstruction and tracking benchmarks, with 3D tracking throughput 18-300 times faster and pose estimation over 100 times faster than prior methods.
08 Dec 2025
Researchers from University of Wisconsin-Madison, UCLA, and Adobe Research introduce a computational framework for "relational visual similarity," which identifies image commonalities based on abstract logic rather than surface features. Their `relsim` model, trained on a novel dataset of images paired with anonymous group-derived captions, aligns significantly with human perception of relational similarity and outperforms existing attribute-based metrics in retrieval tasks.
10 Dec 2025
The growing adoption of XR devices has fueled strong demand for high-quality stereo video, yet its production remains costly and artifact-prone. To address this challenge, we present StereoWorld, an end-to-end framework that repurposes a pretrained video generator for high-fidelity monocular-to-stereo video generation. Our framework jointly conditions the model on the monocular video input while explicitly supervising the generation with a geometry-aware regularization to ensure 3D structural fidelity. A spatio-temporal tiling scheme is further integrated to enable efficient, high-resolution synthesis. To enable large-scale training and evaluation, we curate a high-definition stereo video dataset containing over 11M frames aligned to natural human interpupillary distance (IPD). Extensive experiments demonstrate that StereoWorld substantially outperforms prior methods, generating stereo videos with superior visual fidelity and geometric consistency. The project webpage is available at this https URL.
07 Dec 2025
MeshSplatting generates connected, opaque, and colored triangle meshes from images using differentiable rendering, enabling direct integration of neurally reconstructed scenes into traditional 3D graphics pipelines. The method achieves a +0.69 dB PSNR improvement over MiLo on the Mip-NeRF360 dataset and trains 2x faster while requiring 2.5x less memory.
09 Dec 2025
Visionary introduces a WebGPU-powered platform for 3D Gaussian Splatting (3DGS) that enables real-time, client-side rendering and inference for dynamic and generative 3DGS models. The platform demonstrates up to 135x speedup compared to WebGL-based viewers, while maintaining or improving visual quality and ensuring robust depth-aware composition.
08 Dec 2025
The paper introduces Group Representational Position Encoding (GRAPE), a unified group-theoretic framework that re-conceptualizes and unifies existing positional encoding mechanisms like RoPE and ALiBi. It provides a principled design space for new encodings, demonstrating improved training stability and superior zero-shot performance in large language models.
08 Dec 2025
A new framework, Distribution Matching Variational AutoEncoder (DMVAE), explicitly aligns a VAE's aggregate latent distribution with a pre-defined reference distribution using score-based matching. The approach achieves a state-of-the-art gFID of 1.82 on ImageNet 256x256, demonstrating superior training efficiency for downstream generative models, particularly when utilizing Self-Supervised Learning features as the reference.
09 Dec 2025
Researchers at HKUST developed TrackingWorld, a framework for dense, world-centric 3D tracking of nearly all pixels in monocular videos, effectively disentangling camera and object motion. This method integrates foundation models with a novel optimization pipeline to track objects, including newly emerging ones, demonstrating superior camera pose estimation and 3D depth consistency, achieving, for example, an Abs Rel depth error of 0.218 on Sintel compared to 0.636 from baselines.
09 Dec 2025
Multimodal Large Language Models (MLLMs) exhibit substantial cross-modal inconsistency, producing different answers for semantically identical information presented across image, text, and mixed modalities. This problem persists even with perfect Optical Character Recognition (OCR), revealing an inherent reasoning challenge where text inputs generally achieve higher accuracy than image inputs.
08 Dec 2025
Researchers from Meta AI and the University of Copenhagen developed OneStory, a framework for coherent multi-shot video generation that utilizes adaptive memory modules to model long-range cross-shot context. The method consistently outperforms existing baselines, achieving higher inter-shot coherence scores (e.g., 0.5813 average inter-shot coherence in text-to-multi-shot video tasks) and enhanced shot-level quality.
10 Dec 2025
Researchers from Columbia University and NYU introduced Online World Modeling (OWM) and Adversarial World Modeling (AWM) to mitigate the train-test gap in world models for gradient-based planning (GBP). These methods enabled GBP to achieve performance comparable to or better than search-based planning algorithms like CEM, while simultaneously reducing computation time by an order of magnitude across various robotic tasks.
08 Dec 2025
WorldReel develops a unified, feed-forward 4D generator that integrates geometry, motion, and appearance directly into a latent diffusion model, yielding videos with explicit 4D scene representations. The model achieves state-of-the-art photorealism and significantly improves geometric consistency and dynamic range, particularly for complex scenes with moving cameras.
09 Dec 2025
We present WonderZoom, a novel approach to generating 3D scenes with contents across multiple spatial scales from a single image. Existing 3D world generation models remain limited to single-scale synthesis and cannot produce coherent scene contents at varying granularities. The fundamental challenge is the lack of a scale-aware 3D representation capable of generating and rendering content with largely different spatial sizes. WonderZoom addresses this through two key innovations: (1) scale-adaptive Gaussian surfels for generating and real-time rendering of multi-scale 3D scenes, and (2) a progressive detail synthesizer that iteratively generates finer-scale 3D contents. Our approach enables users to "zoom into" a 3D region and auto-regressively synthesize previously non-existent fine details from landscapes to microscopic features. Experiments demonstrate that WonderZoom significantly outperforms state-of-the-art video and 3D models in both quality and alignment, enabling multi-scale 3D world creation from a single image. We show video results and an interactive viewer of generated multi-scale 3D worlds in this https URL
08 Dec 2025
Researchers at Zhejiang University developed LIVINGSWAP, a high-fidelity video face swapping framework designed for cinematic quality by directly leveraging complete source video attributes and employing keyframe conditioning. The system outperforms existing methods on new cinematic benchmarks and reduces manual editing effort by approximately 40 times.
There are no more papers matching your filters at the moment.
