6arXiv cs.LG (Machine Learning)·
Three-Body Scattering Modeling achieves state-of-the-art one-step image generation on ImageNet-256
A new arXiv preprint introduces Three-Body Scattering Modeling (TBSM), a generative modeling framework that uses energy distance to induce sample-level motion and provide regression supervision for one-step generators, without adversarial critics, prescribed noise paths, or autoregressive factorization. Each 'projectile' sample is attracted toward a real source and repelled from an independently generated source, with the expectation corresponding to the 2-Wasserstein gradient-flow velocity. On ImageNet-256, TBSM achieves FID=2.23 with PixelDiT-XL and FID=1.63 with latent DiT-XL at a single function evaluation (NFE=1), competitive with or surpassing multi-step diffusion methods. The work establishes a new design space connecting diffusion, drift-like, and GAN-like objectives for high-dimensional one-step generation.
Related guides (2)
Related events (8)
FLUX3D: Diffusion-aligned sparse representation for high-fidelity image-to-3D Gaussian Splatting
Researchers introduce FLUX3D, an image-to-3D Gaussian Splatting framework that addresses two structural bottlenecks in sparse voxel-based 3D generation: a representation bottleneck from discriminative 2D features and a cross-modal correspondence bottleneck in diffusion transformers. The system introduces Diffusion-Aligned Structured Latents (DA-SLAT) and a Sparse-structure Multimodal Diffusion Transformer (SMDiT) with Modal-Aware Rotary Positional Embedding (MARoPE) to improve 2D-3D alignment. Benchmark results claim substantial improvements in appearance fidelity over all current state-of-the-art methods for 3DGS asset generation.
World from Motion: generative dynamic 3D Gaussian reconstruction from monocular video
Researchers present World from Motion, a method that generates freely renderable dynamic 3D Gaussian Splatting (3DGS) representations from monocular video by conditioning a video model on pixel-aligned renderings encoding appearance, geometry, and 3D scene motion. The approach trains on aligned multiview video pairs with simulated monocular reconstruction artifacts, then distills model generations back into a consistent dynamic 3DGS at test time. The method claims state-of-the-art results in 4D reconstruction and generalizes to in-the-wild videos with large viewpoint changes and dynamic content.
PTL-Diffusion: Diffusion framework with periodic terminal laws for manifold-aware generation
PTL-Diffusion is a new diffusion modeling framework that replaces the standard single Gaussian terminal distribution with a periodic family of Gaussian terminal laws, embedding phase structure directly into the forward noising dynamics rather than only in the denoising network. The authors derive closed-form forward marginals and reverse posteriors for a periodically forced Ornstein-Uhlenbeck process, enabling standard noise-prediction training. Experiments on torus, cylinder, and face datasets show improvements in manifold-level distributional matching over DDPM baselines. The work is a proof-of-concept motivating structured terminal reference laws as a direction for geometry-aware generative modeling.
The Annotated Diffusion Model
A Hugging Face blog post providing a detailed, annotated walkthrough of diffusion models for image generation, likely covering the mathematical foundations and implementation details of denoising diffusion probabilistic models (DDPMs). The post serves as an educational deep-dive into the architecture and training process of diffusion-based generative models. Published in mid-2022, it coincides with the period of rapid growth in diffusion model adoption.
Meta Research Improves Image Generation via Staged Planning and Self-Revision Fine-Tuning
Researchers from Meta and collaborating universities propose a fine-tuning method that teaches image generators to compose images through discrete plan-sketch-inspect-refine cycles rather than generating all at once. Starting from BAGEL-7B, they construct ~62,000 training examples using GPT-4o and FLUX.1 Kontext to supervise each stage, achieving 83% on GenEval versus 77% for the base model and a competing method (PARM) that required 11x more training data and ~8x more inference steps. The approach improves spatial relationship accuracy, object attribute fidelity, and real-world knowledge grounding in generated images.
DiT-Reward converts text-to-image diffusion transformers into reward models, outperforming HPSv3
DiT-Reward is a new reward modeling approach that repurposes pretrained text-to-image Diffusion Transformers (DiTs) by processing near-clean image latents and aggregating text-conditioned representations across transformer layers. Under matched training data, it outperforms HPSv3 on all four evaluated preference benchmarks, reaching 85.6% on HPDv2 and 77.6% on HPDv3. When used to optimize Stable Diffusion 3.5 Large via Flow-GRPO, it shows clear gains in realism and achieves a 1.65x inference speedup over HPSv3. The work demonstrates that generative DiT representations transfer meaningfully to reward modeling and policy optimization.
Representation-Conditioned Diffusion Models for Controllable Image Generation
This paper explores conditioning diffusion models on representations from pre-trained self-supervised models as an alternative to text prompts or semantic maps, which require large annotated datasets. The self-conditioning mechanism improves unconditional image generation quality and provides a controllable representation space. The authors identify directions of variation in this space and demonstrate smoothness and disentanglement properties, suggesting potential for fine-grained generative control without heavy annotation overhead.
GTCL: Latent trajectory discrimination framework for AI-generated text detection
A new arXiv preprint proposes Geometric Trajectory and Contrastive Learning (GTCL), a framework that reframes AI-generated text detection as a problem of distinguishing latent generation trajectories rather than treating documents as static objects. The method segments documents into ordered local units, encodes them in embedding space, and applies contrastive learning to capture geometric regularities of autoregressive generation. GTCL is evaluated on three benchmarks and consistently outperforms detection baselines, suggesting that modeling sequential dynamics provides more robust discriminative signals than aggregate statistics or global embeddings.