flow matching

The papers

Most central to this idea first, not most recent.

• Industrycs.AI

  Robots can act at 50 Hz while updating visual understanding once a second

  Leap-VLA decouples slow semantic from fast reactive control, reusing cached VLM context to match same-frequency performance.

  Leapmotor · Nov 2026

• Academiccs.AI

  Flow matching speeds up speculative decoding by 39% over existing drafters

  Drafting speculative tokens as a short flow over the target model's hidden states speeds generation faster than autoregressive or diffusion drafters.

  Yildiz Technical University · Nov 2026

• Independentq-bio.BM

  AI models sample more protein folding stages, but miss folded endpoints

  FoldScape predicts protein structures at any folding stage from sequence, with constant accuracy across the folding coordinate.

  Nov 2026

• Big Techcs.SD

  Image generation models can be repurposed for speech synthesis while preserving voice

  ImageTTS repurposes a pretrained image diffusion transformer for speech synthesis using spectrograms as images.

  Wuhan University, Microsoft Corporation · Oct 2026

• Big Techcs.CV

  Image generators can gain quality while largely preserving visual variety

  Adding competition among candidate velocity predictions lets flow-matching generators improve both image quality and coverage instead of trading one for the other.

  Cornell University, Adobe Research · Nov 2026

• Chinese Techcs.CV

  Adaptive noise timing boosts video compression realism with 40% bitrate savings

  Choosing the right moment to start denoising adaptively balances fidelity and realism in generative video compression.

  Beihang University, Alibaba Group · Nov 2026

• Top Universitycs.CV

  Joint video and 3D motion prediction reduces trajectory errors by over half

  WorldFlow jointly predicts future video and queried three-dimensional object motion, achieving the best displacement errors among evaluated learning-based methods on three test sets.

  RoboScience, National University of Singapore · Oct 2026

• Industrycs.CV

  Image generators can improve their own samples without external judges

  A training recipe, Scout-Refine-Correct, lets flow-matching image generators refine their own samples during inference, matching an oracle-based best-of-N baseline on ImageNet without a verifier.

  FusionBrain, AXXX, HSE Unifersity · Nov 2026

• Chinese Techcs.LG

  A frozen AI model can imitate an unseen dataset from one example

  A pretrained generator can create realistic new time series for an unseen domain from a single reference sequence, without retraining, by reusing learned shape and shape-transition vocabularies.

  State Key Laboratory of AI Safety, Institute of Computing Technology, Chinese Academy of Sciences, University of Chinese Academy of Sciences · Nov 2026

• Top Universitycs.CV

  A single flow matching Transformer generates images and text jointly in pretrained encoder spaces

  A single flow matching transformer over frozen encoder spaces generates both images and text jointly without fixed modality order.

  New York University, Harvard University · Oct 2026

• Top Universitycs.LG

  Predicting a spread of returns helps most when training data is scarce

  Supervising a flow-matching return model at every noise level, not just the endpoint, yields more accurate value estimates in offline reinforcement learning, especially with limited data.

  Princeton University, University of California, Berkeley · Nov 2026

• Academiccs.CV

  Generating video directly in pixels can beat models that compress first

  Generating video directly in pixel space, without a pretrained autoencoder, beats latent-space models when the transformer's patch size is varied with network depth.

  FAIR, Meta, Sorbonne University, Paris · Nov 2026

• Top Universitycs.CV

  Text-to-motion model animates unseen skeletons from their skeleton file alone

  TopX generates motion for any skeleton from its file alone, without requiring clips of that skeleton.

  University of Southampton, Nanjing University · Oct 2026

• Chinese Techcs.LG

  AI image generators and robot controllers can follow guidance more reliably

  Coupled Guidance preserves the latent-data pairing during flow-matching generation, improving stability and sample quality in robotics and image tasks without extra overhead.

  Huawei · Oct 2026

• Chinese Techcs.LG

  Time-resolved diagnostic separates irreducible velocity ambiguity from reducible training error

  A time-resolved Bayes-risk decomposition separates irreducible velocity ambiguity from reducible error, guiding flow-matching time sampling that improves velocity loss and feature Fréchet distance.

  Beijing Institute of Technology, ByteDance · Oct 2026

• Big Techcs.CV

  One model generates matched video and motion for arbitrary rigs

  JCA co-generates scene-aligned video, camera, and replayable 3D motion for arbitrary rigs more faithfully than evaluated video-first pipelines.

  NVIDIA, Technion – Israel Institute of Technology · Oct 2026

• Academiccs.CV

  Noise-rejuvenated SMC enables test-time scaling for flow matching image generation

  Noise-rejuvenated sequential Monte Carlo enables reward-guided text-to-image generation with flow matching models, without training or gradient optimization.

  Uppsala University, Linköping University · Oct 2026

• Top Universitycs.LG

  Direct distribution rewards preserve diversity during diffusion model fine-tuning

  Reward fine-tuning of diffusion and flow models directly regularizes the generated distribution via distribution-dependent per-sample rewards, improving diversity and reducing drift at matched reward.

  University of Oxford, AITHYRA · Oct 2026

• Top Universitycs.CV

  Self-correcting flow matching via lightweight refiner that reuses hidden states.

  CorrFlow improves flow-matching generation by reusing each base-model pass to predict and apply velocity-residual corrections during sampling.

  Nanyang Technological University, Télécom Paris, Institut Polytechnique de Paris · Oct 2026

• Big Techcs.LG

  MEND selectively moves flow-model samples when reward gains justify distance

  MEND caps rewards per group and accepts a flow-model velocity update only when the capped reward gain exceeds a quadratic displacement price.

  The University of Texas at Austin, Google · Oct 2026

• Top Universitycs.CV

  Large-scale dataset and token-fusion architecture improve 3D asset editing

  Alchemy3D achieves better 3D asset editing fidelity, source preservation, and visual quality than prior methods.

  The University of Hong Kong, Shenzhen Loop Area Institute · Oct 2026

• Big Techcs.CV

  Variable-size tokens cut diffusion costs where less detail is needed

  LoT Diffusion enables image and video generation with fewer tokens in low-detail regions, achieving up to 2.04x and 3.53x speedups with quality preserved.

  Stanford University, Google · Oct 2026

• Independentcs.CV

  Robot policies that read early vision layers handle camera shifts better

  Vision-language-action policies perform better when action models access intermediate visual representations instead of only the final encoder output.

  Nov 2026

• Industrycs.RO

  Human corrections on rollback segments help desktop robots reach 70% success

  A vision-language-action foundation model with three-stage training and Rollback reaches 93.6% success on RoboTwin 2.0.

  Meituan Robotics · Nov 2026

• Big Techcs.SD

  One audio model can now describe, generate and edit 3D sound scenes

  Spatial Transfusion unifies spatial audio understanding, generation, and editing, preserving understanding while outperforming full fine-tuning.

  Microsoft · Nov 2026

• Big Techcs.RO

  Robots can learn door lifting with ten times fewer demonstrations

  Reusing a frozen reinforcement-learned action decoder and learning a visual latent interface makes dexterous manipulation from visual demonstrations far more data-efficient.

  University of California, Berkeley, Amazon · Nov 2026

• Big Techcs.RO

  Robots complete more two-handed tasks when AI repeatedly refines actions

  Vision-language models that actively participate in action denoising outperform those that only provide static conditioning.

  Peking University, The Hong Kong University of Science and Technology (Guangzhou) · Nov 2026

• Big Techcs.RO

  Driving models can recognize hazards yet fail to brake

  Driving VLAs can encode and describe hazards yet ignore them when planning; inference-time hazard guidance restores braking without retraining.

  Politecnico di Milano, Stanford University · Oct 2026

• Big Techcs.CV

  Image generators can add detail while largely preserving image structure

  Flattening a token's energy distribution by scaling its channel magnitudes gives continuous control over generated image detail.

  The Australian National University, Microsoft · Nov 2026

• Top Universitycs.CV

  Persistent memory and reinforcement learning reduce drift in long driving simulations

  WAM-Sim uses persistent hierarchical memory and online flow-matching RL to generate consistent and realistic driving simulations over long horizons.

  Fudan University, Shenzhen Yinwang Intelligent Technology Co., Ltd · Nov 2026

• Research Labcs.CV

  Adaptive steering fields control image generation while preserving structure

  Steering Fields adaptively adjust steering direction per generation step, achieving state-of-the-art safety and editing performance.

  Max Planck Institute for Informatics, Sapienza University of Rome · Oct 2026

• Chinese Techcs.AI

  Fast generative models improve with RL when noise split is optimized geometrically

  Geo-FSRL selects stochastic policies minimizing conditional KL divergence, achieving higher reward scores than prior methods in few-step flow models.

  Beihang University, Sensetime Research · Nov 2026

• Academiccs.CV

  Diagnose-to-Generate aligns flow models with localized structural corrections

  Diagnose-to-Generate (D2G) improves structural fidelity in flow-based image generators using patch-level feedback and trajectory-preserving stochastic sampling.

  University of Science and Technology of China, Meituan Inc. · Oct 2026

• Big Techcs.CV

  Velocity scaling improves flow models by correcting sampling time lag

  Velocity scaling improves flow matching by reducing population time lag, not by correcting an MSE velocity deficit.

  University of Geneva, Meta FAIR · Oct 2026

• Big Techcs.LG

  PathFM generates complete forecast trajectories without sacrificing marginal accuracy

  PathFM uses flow matching to generate complete multivariate forecast paths, matching marginal forecast accuracy while preserving trajectory dependence at sampling speeds comparable to quantile models.

  California Institute of Technology, Amazon · Oct 2026

• Big Techcs.LG

  Learning powered conditional distributions directly with importance-weighted flow matching

  Importance-Weighted Conditional Flow Matching learns an ω-powered conditional distribution more faithfully than classifier-free guidance.

  University of Geneva, HES-SO/HEG Genève · Oct 2026

• Top Universitycs.CV

  Symmetry-regularized flow matching enables video generation from any camera pose without ground truth novel views

  SymRegFlow generates multi-view consistent videos across continuous viewpoints without novel-view RGB supervision, using masked dual-anchor supervision and cross-anchor denoising consistency.

  Tsinghua University, Robert Bosch GmbH · Oct 2026

• Top Universitycs.LG

  Larger optimal transport batches make flow matching more stable

  Larger online entropic OT minibatches reduce gradient variance and integration error in flow matching.

  University of California, Berkeley · Oct 2026

• Research Labcs.LG

  Spectral hybrid parameterizations robustly accelerate flow matching optimization

  Hybrid spectral parameterizations for flow matching that adapt across time and data covariance directions accelerate optimization with no extra training cost.

  Inria, ENS de Lyon · Oct 2026

• Big Techstat.ML

  Flow matching avoids the curse of dimensionality by adapting to intrinsic data dimension

  Flow matching achieves finite-sample Wasserstein generalization rates governed by data’s intrinsic Wasserstein dimension rather than its ambient dimension.

  University of Michigan, Google DeepMind · Oct 2026

• Top Universitycs.LG

  Steering input noise aligns few-step generators without retraining them

  Noise Policy Optimization aligns few-step generative models by learning a flow policy over the noise space of a frozen generator.

  KAIST, Stanford University · Oct 2026

• Top Universitycs.LG

  Conditioning actions on returns improves value learning in reinforcement learning

  A Q-learning critic can be trained by learning return-conditioned action likelihoods, using actions themselves as supervision, improving RL performance.

  Stanford University · Oct 2026

• Big Techcs.CV

  One LiDAR model generates, forecasts and resimulates driving scenes in real time

  A single conditional LiDAR world model unifies generation, forecasting, and resimulation, real-time at 11 FPS on one H100 GPU.

  NVIDIA, University of Illinois, Urbana-Champaign · Oct 2026

• Research Labcs.RO

  Compact robot policy uses lightweight memory to stabilize action generation

  A 2.4B vision-language-action model trained only on public data is outperformed only by models about twice its size, and a register-token memory cuts pose oscillations fourfold.

  Kyutai, Inria · Oct 2026

• Big Techcs.CV

  Correcting classifier-free guidance in early steps improves image editing fidelity

  Standard classifier-free guidance harms source preservation in early, noisy flow steps for text-and-image-to-image editing; replacing the unconditional velocity with a no-edit velocity fixes it.

  Imperial College London, Microsoft · Oct 2026

• Chinese Techcs.CV

  World model forecasts future 3D geometry directly without recursive rollouts

  GeoWM directly predicts future 3D geometry at requested horizons, improving accuracy and cutting long-horizon inference time by avoiding recursive rollouts.

  Huawei Noah's Ark Lab · Oct 2026

• Big Techcs.RO

  Flexible block-causal models sustain efficient long-horizon action-conditioned rollouts

  A block-causal world-action model enables real-time, stable long-horizon imagination and planning, solving benchmark tasks entirely in imagination.

  Center for Robotics and Embodied Intelligence (CREO), New York University, Courant Institute of Mathematical Sciences · Oct 2026

• Chinese Techcs.RO

  Robot-relative 3D features improve pretrained manipulation policies across diverse tasks

  Spatial Grafting binds frozen 3D reconstruction features to metric robot-relative geometry to improve pretrained robot manipulation policies.

  University of Toronto, Huawei Noah’s Ark Lab · Sept 2026

• Top Universitycs.LG

  Loop-specific targets let diffusion transformer improve with extra inference loops

  LiFT improves ImageNet generation by repeatedly applying a shared DiT core with depth-specific trajectory targets, cutting parameters and compute versus a dense DiT baseline.

  University of Amsterdam, TNO · Oct 2026

• Big Techcs.RO

  Recurrent memory policy improves robot manipulation on tasks requiring spatial recall, counting, and timing

  ReCAT uses structured recurrent memory with Mamba-2 and causal attention to achieve high success on memory-dependent robot manipulation tasks.

  University of Edinburgh, Karlsruhe Institute of Technology · Oct 2026

• Big Techcs.RO

  Shared 3D geometry aligns predicted driving scenes and vehicle motion

  PhysWAM uses a geometric constraint between depth and ego motion to jointly predict future scenes and actions for autonomous driving.

  University of Southern California, Woven by Toyota · Oct 2026

• Independentcs.RO

  Filtered critic gradients speed reinforcement learning for flow-based robot policies

  QF3 is the first off-policy flow RL method to train humanoid locomotion from scratch, achieve 10x speedup, and transfer zero-shot to hardware.

  Oct 2026

• Big Techcs.LG

  Permutation-equivariant flow matching generates neural network weights without alignment

  A permutation-equivariant flow matching model generates neural network weights directly from independently trained networks without alignment.

  Technion – Israel Institute of Technology, Technical University of Munich · Oct 2026

• Top Universitycs.LG

  Data-free distillation of diffusion models with a single consistent objective

  A single objective unifying sample generation and score estimation lets a student distill a frozen diffusion teacher data-free, with Wasserstein convergence and ImageNet FID 2.04 at one step.

  University of British Columbia, Vector Institute for AI · Oct 2026

• Top Universitycs.CV

  One Flow Model Can Reconstruct Scenes or Generate Missing Detail

  A shared clean-target predictor performs direct reconstruction at one step and conditional generation through multi-step flow.

  Peking University · Oct 2026

• Academiccs.CV

  Repelling crossing trajectories makes flow matching smoother and easier to sample

  Contrastive repulsion of training trajectories produces smoother velocity fields for few-step flow matching.

  The Hong Kong University of Science and Technology · Oct 2026

• Chinese Techcs.CV

  Flow matching plus physics yields more trackable 3D human motion from video

  FlowHMR uses flow matching with reinforcement learning to recover physically plausible global 3D human motion from monocular video, beating prior methods in fidelity and tracking.

  Peking University, The Hong Kong University of Science and Technology · Oct 2026

• Big Techcs.LG

  Distilled language models generate sequences through one-step continuous flows

  Gumbel Straight Flow distills autoregressive models into one-step flow maps that outperform few-step baselines.

  Google DeepMind Amsterdam · Oct 2026

• Big Techcs.CV

  Foundation-model initialization improves feed-forward 4D hand-object reconstruction from video

  4D-HOF uses conditional flow matching to refine foundation model estimates into accurate, stable 4D hand-object reconstructions.

  NVIDIA, Texas A&M University · Oct 2026

• Chinese Techcs.CV

  Sparse flow matching generates 3D shapes at ultra-high voxel resolution

  Filigree3D generates detailed 3D geometry from one image at 2048^3 voxel resolution by combining sparse computation, local-global attention, and multi-scale visual conditioning.

  Alibaba Group · Oct 2026

• Big Techcs.RO

  Generated language gives robot policies a working memory

  Jointly generating video, action, and language in one model improves robot manipulation and lets generated language serve as working memory.

  Brown University, Amazon · Oct 2026

• Chinese Techcs.CV

  Spatial gradient maps improve RL alignment of flow-based image models

  SGA-Flow-GRPO uses spatial gradient-guided credit assignment to achieve state-of-the-art alignment quality for flow-based generative models on GenEval.

  Sun Yat-Sen University, Baidu · Sept 2026

• Chinese Tech

  Dynamic optimization loop aligns 3D generators with 2D diffusion priors to improve realism

  OREO enhances 3D generation fidelity by editing rendered views and distilling the gap back into the generator via latent contrastive learning.

  The Hong Kong Polytechnic University, Tencent ARC Lab · Oct 2026

• Top Universitycs.RO

  Robots maintain control speed as their AI models grow tenfold

  Decoupling the vision-language backbone from the action head lets humanoid control rate stay at 20 Hz regardless of model size.

  XPENG Robotics, University of Washington · Nov 2026

• Top Universityeess.AS

  229K-hour dataset lets speech models generate minutes-long audio with control

  Emilia2 is a 229K-hour dataset for controllable long-form speech generation across 41 languages.

  The Chinese University of Hong Kong, Shenzhen, University of Illinois at Urbana-Champaign · Nov 2026

• Academiccs.CV

  Predicting motion, not video, improves robot learning from human footage

  Predicting 3D scene flow instead of video from pretrained models improves robot policy learning.

  Georgia Institute of Technology · Nov 2026

• Big Techcs.CV

  AI models forecast satellite changes better without predicting individual pixels

  Forecasting future satellite observations as foundation-model embeddings instead of pixel spectra halves prediction error and gives far better-calibrated uncertainty.

  University of Cambridge, Microsoft AI for Good Research Lab · Nov 2026

• Top Universitycs.RO

  Robot AI speed depends more on action generation than model size

  Action-generation strategy determines VLA inference latency on edge hardware more than model size.

  National University of Singapore, 0G Labs · Oct 2026

• Big Techcs.RO

  AI driving models go farther when action feedback updates perception

  The benefit of VLM gradient insulation depends on action-token supervision and evaluation, questioning its default use in VLA training.

  NVIDIA, Princeton University · Nov 2026

• Top Universitycs.CV

  High-resolution image editing gets a 4x speedup by keeping a low-resolution anchor

  FLASH enables faithful 2K/4K image editing by anchoring a high-resolution detail branch to a frozen 1K editing backbone.

  Shanghai Innovation Institute, Fudan University · Nov 2026

• Independentcs.LG

  AI diffusion models can follow deterministic paths through discrete data

  Exponential residual flows provide an ODE-based sampler for discrete diffusion models that preserves marginals and improves sampling.

  Oct 2026

• Independentcs.RO

  Bimanual robot learns precise tasks without any robot-collected data

  A two-armed robot skill system trained only on 16,151 hours of handheld-gripper human demonstrations performs long, precise manipulation tasks at a new site, averaging 78% success.

  Nov 2026

• Top Universitycs.RO

  Metric depth helps robotic manipulation policies generalize across different robots and settings

  Using metric depth from stereo cameras boosts robot manipulation success and generalization from UMI demonstrations.

  USC Physical Superintelligence Lab, CMU · Nov 2026

• Chinese Techcs.RO

  Future 3D motion prediction boosts robot learning with fewer demonstrations

  Augmenting robot action learning with future 3D motion prediction improves manipulation performance and generalization.

  The Hong Kong Polytechnic University, Alibaba Group · Oct 2026

• Academiccs.CV

  Tuning only the first denoising step restores lost diversity in distilled models

  Diversity loss in distilled diffusion models is concentrated in the first step and can be recovered by a lightweight first-step bias adapter.

  Graz University of Technology, Complexity Science Hub Austria · Oct 2026

• Independentcs.CL

  Scalable flow language model rivals autoregressive quality at 2.5x speed

  Continuous flow language models can scale to 8B parameters, match autoregressive quality, and decode in parallel faster.

  Oct 2026

• Big Techcs.SD

  Audio training teaches one model to edit scenes across modalities

  Training on synthetic audio edits and masked audiovisual reconstruction enables zero-shot, instruction-driven editing of movie scenes across image, audio, and video modalities.

  Adobe Research, Carnegie Mellon University · Oct 2026

• Top Universitycs.AIcode

  EvoSteer Repairs Multi-Agent Workflows While They Are Still Running

  EvoSteer outperforms baselines across twelve datasets with online self-evolving graph orchestration using AnchorTB and validated skill admission.

  The Chinese University of Hong Kong, Shenzhen, Dalian University of Technology · Oct 2026

• Big Techstat.ML

  Average generator yields 16x faster discrete diffusion language model sampling

  The Discrete Average Generator achieves state of the art generative perplexity on OpenWebText with 16x fewer generation steps.

  University of California, Los Angeles, Google · Oct 2026

• Big Techcs.CV

  Pixel diffusion benefits from adapting its prediction target by data direction

  A direction-adaptive interpolation between clean-image and velocity prediction, plus a post-hoc convolutional refiner, improves pixel-space diffusion transformers.

  University of Maryland, College Park, Adobe Research · Oct 2026

• Independentcs.CV

  One model combines text, image and video references for editing

  OmniXRef outperforms task-specific models and commercial systems on multimodal multi-reference video editing.

  Oct 2026

• Top Universitycs.RO

  Classifier gradients let frozen robot policies adapt to new bodies without demonstrations

  A lightweight embodiment classifier guides a frozen flow-based robot policy to new dynamics, raising average success from 72% to 80% in LIBERO simulation.

  UC Berkeley · Oct 2026

• Chinese Techcs.CV

  Generative framework reconstructs complete large-scale 3D meshes from unposed photos

  WorldRecon reconstructs complete, large-scale 3D scene meshes from unposed multi-view images.

  Zhejiang University, Tencent Hunyuan · Oct 2026

• Academiccs.LG

  Backpropagating verifier rewards improves reasoning in continuous flow language models

  A critic predicting per-token values lets continuous flow language models improve reasoning through backpropagated verifier rewards.

  Tufts University · Oct 2026

• Top Universitycs.CV

  Single generator reconstructs 3D scenes at any scale without per-scene optimization

  T3lescope reconstructs arbitrary-resolution high-fidelity 3D scene meshes from posed multi-view images without per-scene optimization.

  Preferred Networks, Inc., The University of Tokyo · Oct 2026

• Big Techcs.CV

  Generative chroma keying from unaligned background references preserves foreground performance

  FreeCHROMA generates composite videos from unaligned background references while preserving foreground identity and motion.

  University of Maryland, College Park, Amazon · Oct 2026

• Big Techcs.CV

  Polynomial student paths improve four-step video distillation quality

  Parametric Trajectory Distillation fits teacher trajectory segments to polynomials and supervises along the student's own predicted path, preserving motion and detail in four-step video generation.

  EPFL, NVIDIA · Oct 2026

• Academiccs.CV

  Splitting diffusion distillation into two phase-specific half-sized experts beats a full-size student.

  Phase-wise Velocity Distillation uses two half-sized phase-specific experts to generate high-quality images in one full-backbone forward pass.

  The Hong Kong Polytechnic University, OPPO Research Institute · Oct 2026

• Top Universitycs.CV

  Better video predictions improve robots’ performance on unfamiliar manipulation tasks

  Video Prediction Policy 2 achieves strong zero-shot generalization in video prediction and action generation for robot manipulation.

  Robotera, Tsinghua University · Oct 2026

• Chinese Techcs.LG

  Direct data-space refinement improves few-step generation without flow schedules

  Data-space iteration directly refines predictions in data space and outperforms standard flow discretization for few-step generation.

  ByteDance Seed · Oct 2026

• Big Techcs.CV

  One-step video refiner achieves 8.91x speedup for 4K upscaling

  SoL-Refiner transforms low-resolution video outputs into 4K videos with a single denoising step, outperforming multi-step refiners.

  NVIDIA · Oct 2026

• Big Techcs.AI

  Reasoning supervision lets action models skip decoding without losing performance

  Action generation does not need to wait for reasoning text; TacitVLA matches explicit-reasoning VLA performance at up to 3.4x lower latency.

  NVIDIA, New York University · Oct 2026

• Chinese Techcs.CV

  Stochastic attention routing enables efficient RL alignment for diffusion models

  Routing Policy Optimization uses temporary stochastic attention routes to improve reward-aligned visual generation with less training compute while preserving dense deterministic inference.

  National University of Singapore, ByteDance · Oct 2026

• Big Techcs.CV

  High-fidelity latent priors improve efficient high-resolution image generation

  UltraDiff distills a joint f8-f32 diffusion teacher into an f32 student, approaching f8 image quality at f32 speed.

  The Hong Kong University of Science and Technology (Guangzhou), Adobe Research · Oct 2026

• Big Techcs.CV

  Calibrated depth reward improves 4D consistency without harming visual quality

  A single calibrated geometric reward improves both 4D consistency and video quality without auxiliary rewards.

  KAIST, ETH Zürich · Oct 2026

• Big Techcs.CV

  Observation-only videos can directly pretrain robot action policies

  NAVA-WAM directly pretrains an action policy from observation-only videos using future-video flow matching and outperforms prior world action models.

  Meta AI, University of Copenhagen · Oct 2026

• Big Techcs.CV

  Generative and agentic pipeline reconstructs scene-aligned 3D objects from casual photos

  GATOR reconstructs complete, scene-aligned 3D objects with pose from one or more casual images using generative and agentic methods.

  NVIDIA, Simon Fraser University · Oct 2026

• Big Techcs.CV

  Simulator-aligned training makes world models more responsive to actions

  CtrlWAM aligns perturbed actions with their visual consequences in a simulator to improve intent-foresight consistency in world action models.

  NVIDIA, UC Berkeley · Oct 2026

• Independentcs.CV

  Encoder-free model unifies image and video understanding and generation in pixel space

  PixelUMM unifies image and video understanding and generation in pixel space, delivering comparable performance without pretrained visual encoders or VAE tokenizers.

  Oct 2026

• Chinese Techcs.LG

  Distribution matching distillation and drifting models are equivalent under finite-sample analysis

  Distribution Matching Distillation is equivalent to Drifting Models, and multi-bandwidth training improves DMD convergence and quality.

  Baidu Inc., Central South University · Oct 2026

• Top Universitycs.LGcode

  Dual-expert diffusion improves long-horizon predictions of collisions and smooth dynamics

  HamiFormer reduces phase-space prediction error by up to 26.3% using a dual-expert diffusion field with affine symplectic maps.

  University of Science and Technology of China, Tsinghua University · Sept 2026

• Academiccs.RO

  Adaptive action chunking improves robot control by varying horizon based on prediction reliability.

  GeoAAC adapts action chunking horizon from denoising trajectory geometry, improving VLA policy success rates over fixed horizons.

  Tongji University, The Hong Kong Polytechnic University · Sept 2026

• Top Universitycs.CV

  Revising skipped diffusion steps with later feedback improves caching efficiency and quality.

  PoSTCache revises earlier diffusion predictions using corrective feedback from later computations, improving caching fidelity without training.

  Tsinghua University · Oct 2026

• Independentcs.CV

  Robot policies generalize better when future tokens remain at inference

  Latent world action models match explicit ones in distribution but lose generalization; conditioning on fully noised future tokens recovers most of the benefit.

  Oct 2026

• Academiccs.LG

  Weighting schedules in score-based models determine which data features are learned and at what rate

  Weighting schedules govern the rate at which each feature of multimodal data is acquired during score-based model training.

  Laboratoire de Physique de l’École normale supérieure, ENS, Université PSL, CNRS, Sorbonne Université, Université Paris Cité · Oct 2026

• Top Universitycs.CV

  Touch signals improve predictions of how hands manipulate objects

  A world model conditioned on tactile signals reduces hand motion underestimation from 23% to 9%.

  Yale University, Texas A&M University · Oct 2026

• Big Techcs.CL

  Coupling continuous latent with token feedback improves diffusion language models

  Coupling continuous latent trajectories with token readout and feedback improves diffusion language models on reasoning and language modeling.

  University of Illinois Urbana-Champaign, Amazon.com, Inc. · Oct 2026

• Chinese Techcs.CV

  High-fidelity 3D texture models can train on images alone

  3D texture generation can be trained using only 2D images, without any 3D assets.

  MMLab, CUHK, Tencent Hunyuan · Oct 2026

• Big Techcs.CV

  Last-frame layouts guide video generation through large camera moves

  LIFT enables specifying layout of future views in video generation via on-policy self-distillation from a dense-layout teacher, improving controllability under large viewpoint changes.

  UC San Diego, University of Virginia · Oct 2026

• Top Universitystat.ML

  Steepest Guidance aligns generative models by maximizing local reward improvement

  Inference-time alignment of flow and diffusion models can be cast as sequential optimization in probability-measure space, where steepest local ascent of the reward functional gives unbiased guidance.

  LY Corporation, The University of Tokyo · Oct 2026

• Chinese Techcs.LG

  Three diffusion RL methods unified as divergence-constrained reward maximizers

  DiffusionRFT uses exact sparsemax projection to converge faster, train more stably, and achieve top performance in regression-based diffusion RL.

  Tencent · Oct 2026

• Big Techcs.AI

  Reasoning improves when latent next thoughts remain a distribution

  Modeling each latent next thought as a multimodal distribution improves compact mathematical reasoning and preserves diverse solution paths better than point predictions.

  Rutgers University, Amazon · Sept 2026

• Big Techcs.RO

  Rolling predictions make robot replanning 4.5 times faster

  Rolling-WAM distributes video-action denoising across replanning cycles to achieve 4.5x faster replanning without sacrificing manipulation performance.

  University of Southern California, Brown University · Sept 2026

• Top Universitycs.CV

  Routing 3D geometry into diffusion improves sparse-view scene synthesis

  VGGT-Diff achieves state-of-the-art novel view synthesis from sparse views by routing visual geometry latents from VGGT-Ω into a pretrained video diffusion model.

  XPeng Motors, The Chinese University of Hong Kong · Oct 2026

• Chinese Tech

  Canonical gripper-frame representation achieves zero-shot cross-embodiment transfer for two-finger manipulation

  Interaction-centric representation with a canonical gripper frame enables cross-embodiment zero-shot transfer for two-finger gripper manipulation.

  Renmin University of China, Zhipu AI · Oct 2026

• Independentcs.LG

  Looped flows enable deeper reasoning by chaining denoising steps

  Looped flows train recurrent models with local denoising objectives, allowing more inference steps and improving reasoning benchmark performance.

  Sept 2026

• Independentcs.CVcode

  Image-only pretraining helps build a strong open image generator

  A 6B diffusion transformer trained from scratch with a mostly real 220M-sample pipeline achieves open-source state-of-the-art Qwen-Image-Bench scores of 53.53 and 53.38.

  Sept 2026

• Top Universitycs.LG

  AI agents can learn visual tasks through longer imagined futures

  MAC-v2 uses action chunking and reconstruction-free latent modeling to enable long-horizon value expansion from pixels.

  UC Berkeley · Nov 2026

• Big Techcs.CV

  Learning visual concepts before rendering improves image generation in controlled tests

  Visual generation is better when autoregressive pretraining learns semantic tokens and diffusion post-training renders them.

  University of California, San Diego, Adobe Research · Oct 2026

• Chinese Techcs.RO

  Robots can use human videos before learning their own controls

  Robot policies can learn reusable, three-dimensional motion representations from videos without robot command labels, then adapt them to different robot controls.

  Hong Kong Embodied AI Lab, CUHK · Oct 2026

• Chinese Techcs.CV

  Jointly generating object shapes and camera poses improves sparse-view reconstruction

  A native 3D diffusion model jointly recovers object shape and camera motion from a few uncalibrated views.

  The University of Hong Kong, Transengram · Oct 2026

• Chinese Techcs.CV

  Image generators can train over 140 times faster with pixel feedback

  SpeeDiff enables end-to-end training of latent diffusion models using a Tweedie loss that prevents latent collapse, achieving dramatically faster convergence.

  ByteDance Seed, California Institute of Technology · Oct 2026

• Research Labcs.CV

  AI video generators can match quality with eight times fewer tokens

  TrackTok encodes video into object-centric persistent tokens, cutting latent tokens 8x while matching generative FVD on UCF-101 and beating 1D baselines on frame prediction.

  Max Planck Institute for Informatics, Saarland University · Oct 2026

• Independentcs.CV

  Adversarial post-training fixes high-frequency deficit in pixel diffusion models

  Adversarial post-training improves pixel diffusion outputs by restoring missing high-frequency image statistics.

  Oct 2026

• Big Techphysics.ao-ph

  Joint deterministic-generative model extends reliable extreme-precipitation nowcasting to six hours

  A radar nowcasting model extends skilful extreme precipitation forecasts to 6 hours by separating predictable storm structure from uncertain local details.

  Microsoft Corporation, City University of Hong Kong · Oct 2026

• Top Universitycs.RO

  Human-like scripted demonstrations reproduce behavioral cloning’s puzzling failures

  Behavioral cloning performance can improve with overfitting, and fully closed-loop policies often fail without action chunking.

  University of California, Berkeley · Oct 2026

• Chinese Techcs.CV

  One-step diffusion framework restores 1080p video in real time with efficient streaming attention

  FastVR restores high-resolution video at 11 FPS using a one-step diffusion model with a lightweight VAE and chunk-wise causal attention.

  Alibaba Group, Xidian University · Oct 2026

• Big Techcs.LG

  Denoiser Jacobian spectrum links to generative model quality and enables targeted improvements

  Better generative denoisers have larger principal Jacobian eigenvalues, and regularizing those responses while suppressing noisy directions improves generation quality.

  Stony Brook University, Microsoft Research · Oct 2026

• Top Universitycs.AI

  Temperature-controlled preference updates reduce manifold drift in flow models

  Manifold drift in flow preference optimization moves terminal samples off the pretrained manifold, causing reward hacking.

  Zhejiang University, Kuaishou Technology · Aug 2026

• Chinese Techcs.CV

  Image-to-3D method fixes orientation and faithfulness in coarse generation

  Coarse3D generates orientation-controllable and faithful sparse 3D structures from arbitrary unposed images.

  Alibaba Group · Nov 2026

• Top Universitycs.RO

  Robots can anticipate physical changes without generating videos of the future

  PDRVLA learns compact physical tokens from vision-language features to condition robot actions on future scene dynamics without online video generation.

  Peking University, JD.com · Oct 2026

• Chinese Techcs.CV

  Video generators can change characters or cameras while preserving other details

  Factor-Aware Predictive Learning strengthens MLLM latent use for temporal inference and factor-aware control in cinematic video generation.

  The Hong Kong University of Science and Technology, Tencent Video · Oct 2026

• Chinese Techcs.CV

  Combining compression and adaptive inference for diffusion transformers without quality loss

  Trajectory-aware student training with learned routing and teacher rescue makes compressed Diffusion Transformers compatible with adaptive execution, reaching FID 2.58 at 2.25× speedup.

  The Hong Kong University of Science and Technology (Guangzhou), University of Electronic Science and Technology of China · Oct 2026

• Chinese Techcs.CV

  Image generators create less distorted panoramas with fewer visible seams

  CubeAtlas representation enables distortion-free and seamless 360-degree panorama generation with better quality and lower attention computation than equirectangular projection.

  Tencent · Oct 2026

• Big Techcs.CV

  More Egocentric Video Improves Agents but Not Their Object Effects

  Scaling ego-centric video to 30,000 hours models agents well but leaves object interaction fidelity far behind, requiring training changes, not just more data.

  University of Illinois Urbana-Champaign, Carnegie Mellon University · Oct 2026

• Academiccs.CV

  A controllable video model can serve as a photorealistic 3D avatar

  Video Diffusion Avatars generate photorealistic multi-view consistent avatars from a single image using a controlled video diffusion model.

  Technical University of Munich, Woven by Toyota · Oct 2026

• Chinese Techcs.CV

  Video sampling needs both model likelihood and external physics guidance

  A two-dimensional framework separates low-density sampling risk from high-density drifting risk to explain and reduce physical hallucinations in video generation.

  Zhejiang University, Nanyang Technological University · Oct 2026

• Big Techcs.AI

  Explicit predictive states boost physical plausibility in video world models

  Physical Registry adds explicit predictive states to video DiTs, improving physical plausibility with minimal parameters.

  Yonsei University, Microsoft Research Asia · Oct 2026

• Big Techcs.CV

  Spatially addressed memory preserves scenes across long generated videos

  SMART uses pose-addressed slots for bounded video memory, enabling long-horizon spatial recall without explicit 3D reconstruction.

  ETH Zurich, Google · Oct 2026

• Big Techcs.CV

  Frequency-aware forecasting cuts diffusion transformer evaluations while preserving fidelity

  SpecKCast accelerates Diffusion Transformers by forecasting DCT coefficients with Kalman filters along noise-schedule-aligned time, achieving up to 6x speedup with minimal memory.

  University of Toronto, Vector Institute · Oct 2026

• Top Universitycs.LG

  Advantage guidance makes supervised policy training competitive with offline RL

  Advantage Guidance trains a flow-matching policy conditioned on advantages and uses contrastive sampling guidance to outperform prior RL methods on 26 long-horizon, sparse reward tasks.

  UC Berkeley · Oct 2026

• Big Techcs.LG

  Imagined goals from geometric horizon models boost behavior cloning generalization

  Goal-conditioned behavior cloning generalizes to unseen compositional tasks when trained on goals sampled from a temporal-difference geometric horizon model.

  Mila, Université de Montréal · Oct 2026

• Chinese Techcs.CV

  World coordinates replace learned features for game model control

  ActionRoPE represents action-driven camera displacement as world-coordinate rotary positions, enabling spatially consistent control in planar-camera game world models.

  Tencent, National University of Singapore · Oct 2026

• Chinese Techcs.SD

  AI speech translators can improve quality with little extra initial delay

  A streaming speech translation model learns when to commit from semantically aligned partial examples, with a tunable threshold controlling quality-latency tradeoffs.

  Shanghai Jiao Tong University, Shanghai Innovation Institute · Oct 2026

• Chinese Techcs.RO

  Systematic study reveals what matters for world modeling in robot VLAs

  2D semantic representations and decoupled world model integration improve VLA robustness on long-horizon and dynamic tasks.

  Chinese Academy of Sciences, University of Chinese Academy of Sciences · Oct 2026

• Chinese Techcs.CV

  AI video generators reduce flickering and preserve detail with few steps

  Adding a predictive representation loss to distribution matching distillation improves few-step video generation spatial fidelity and temporal coherence.

  Shanghai Jiao Tong University, Hunyuan (Tencent) · Oct 2026

• Industrycs.CV

  Parallel adapter composition yields real-time joint audio-video generation without chained training

  Parallel adapter composition achieves real-time streaming audio-video generation by adding separately trained causal and few-step adapters.

  LIGHTSPEED, Independent Researcher · Oct 2026

• Research Labcs.CV

  Part-anchored voting generates more faithful human-object interactions from text

  PAMI uses part-anchored voting to generate text-conditioned human-object interactions with 14.5% higher contact recall than prior methods.

  Tübingen AI Center, University of Tübingen, Max Planck Institute for Informatics · Oct 2026

• Chinese Techcs.CV

  Adaptive reward routing improves joint audio-video diffusion training

  Adaptive Reward Routing dynamically adapts update locations and reward weights, improving joint audio-video diffusion across multiple metrics.

  MMLab@HKUST, The Hong Kong University of Science and Technology · Oct 2026

• Academiccs.RO

  One action-conditioned JEPA improves vision-language-action policies across all lifecycle stages

  Juno uses a single action-conditioned JEPA to improve VLA success rates across pretraining, policy learning, and deployment.

  University of Science and Technology of China, Eastern Institute of Technology, Ningbo · Oct 2026

• Big Techcs.CV

  Diffusion in texture space generates high-quality 3D materials from text or photos

  Generating high quality 3D object materials entirely in texture space avoids view consistency problems.

  NVIDIA · Oct 2026

• Chinese Techcs.CV

  Causal video diffusion matches bidirectional quality without a teacher model

  Conditional Residual Prediction reduces causal video models’ dependence on generated history, nearly matching bidirectional quality without a bidirectional teacher.

  The Chinese University of Hong Kong, Shenzhen, Tencent Hunyuan · Oct 2026

• AI Startupcs.RO

  Dense subtask labels help robots complete long two-handed tasks

  FineART, a bimanual manipulation dataset with dense subtask annotations, enables a vision-language-action model to achieve 100% success on spatial instructions.

  Scale AI, Hugging Face · Oct 2026

• Top Universityeess.AS

  YuE2 unifies symbolic and audio music generation via explicit score planning

  YuE2 unifies symbolic and audio music generation by first writing a readable score then producing full song audio.

  HKUST, M-A-P · Oct 2026

• Chinese Techcs.CV

  Diffusion models benefit most from teacher features they still lack

  RARE selects the teacher layer with the largest recoverability gap, improving FID and training speed.

  Central South University, The Hong Kong University of Science and Technology · Oct 2026

• Big Techcs.CV

  ALIVE inserts objects that respond coherently to video actions

  ALIVE inserts objects into videos that coherently interact with the source actions using an edited first frame and object-only instruction.

  University of Rochester, Adobe Research · Oct 2026

• Big Techcs.CV

  Adaptive tokenization cuts compute for pixel-space diffusion models by up to 75% with no fidelity loss.

  BudgetPix matches full-compute image fidelity using 25% of the compute by adaptively tokenizing based on visual complexity.

  University of Illinois Urbana-Champaign, Google · Oct 2026

• Academiccs.CVcode

  Interactive world model adds synchronized spatial audio to visual simulation

  HelixWorld generates controllable visual scenes and camera-aligned stereo audio at 24 FPS on one GPU while maintaining long-horizon synchronization.

  The Hong Kong University of Science and Technology, Noiz AI · Oct 2026

• Chinese Techcs.CV

  Velocity corrections improve instruction following in few-step text-to-image generation

  A sub-20M-parameter velocity residual corrector improves few-step text-to-image instruction following by 17.60 percentage points without altering the frozen generator or adding forward passes.

  The University of Osaka, Alibaba Group · Oct 2026

• Chinese Techcs.CV

  RL-trained cache scheduler adapts video diffusion to user-specified speedups

  MORCA uses latent-aware offline-to-online reinforcement learning to schedule cache reuse, improving video diffusion fidelity at user-specified acceleration targets over existing caching methods.

  Shanghai Jiao Tong University, Alibaba Group · Oct 2026

• Research Labcs.CV

  Backdoor attacks on interactive video generation make outputs freeze when action triggers appear.

  Backdoor attacks on interactive video generation models can use action-guided triggers to cause frozen frames that ignore user actions.

  Chinese Academy of Sciences, University of Chinese Academy of Sciences · Oct 2026

• Chinese Techcs.RO

  Counterfactual post-training improves robot world model's action following and physical plausibility

  A robot world model using offline geometric calibration and counterfactual post-training improves action following and cuts human-assessed interaction defects from 48.12% to 6.25%.

  Institute of Automation, Chinese Academy of Sciences, Alibaba Group · Oct 2026

• AI Startupcs.CV

  Complete 3D meshes improve control of generated video motion

  Using explicit rigid 3D geometry to control video world models achieves better camera and object control than prior methods.

  Stability AI, University of Illinois Urbana-Champaign · Oct 2026

• Chinese Techcs.CV

  Fine-grained routing makes sparse video attention faster without quality loss

  VSA2 makes video DiT attention 8.9x faster with comparable quality by using a fine-grained router and adaptive sparsity.

  University of California, San Diego, ByteDance Seed · Oct 2026

• Top Universitycs.CV

  Shared 3D controls coordinate camera and object motion in generated video

  Generative Cinematographer (GenCine) lifts a single image to 3D to jointly author camera and foreground motion using local 3D motion handles.

  Johns Hopkins University · Oct 2026

• Academiccs.CV

  A zero-shot pipeline that turns monocular 360° video generators into stereoscopic ones

  A zero-shot pipeline generates 4K stereoscopic 360° video from monocular panoramic video by enforcing epipolar consistency.

  UNC Chapel Hill, Dolby Laboratories · Oct 2026

• Chinese Techcs.CV

  Three-stage latent generator reaches 1.11 FID on ImageNet

  A distill-then-refine latent rollout approach with a staged Transformer achieves FID 1.11 on ImageNet 256x256 without Fréchet losses.

  ByteDance · Oct 2026

• Top Universitycs.LG

  Discrete Wasserstein transport enables one-step generation on finite state spaces

  Discrete Drifting uses discrete Wasserstein geometry to enable one-step generative modeling on finite state spaces via Markov jump processes.

  Imperial College London, Aalborg University · Oct 2026

• Top University

  Dynamics-aware distillation preserves motion in four-step video world models

  DyMD improves few-step video generation by preserving interaction dynamics via distribution matching distillation.

  Beihang University, JD Future Academy · Sept 2026