GEAR jointly trains VQ tokenizer and AR generator end-to-end via dual hard/soft read-out and representation alignment, achieving up to 10x faster ImageNet gFID convergence than LlamaGen-REPA while generalizing across quantizers and to text-to-image.
hub
The unreasonable effectiveness of deep features as a perceptual metric
29 Pith papers cite this work. Polarity classification is still indexing.
29
Pith papers citing it
hub tools
citation-role summary
method 3
background 1
citation-polarity summary
representative citing papers
ReMatch corrects train-test residual distribution mismatch in probabilistic downscaling via optimal transport in low-dimensional PCA space, reducing under-dispersion and improving SSR and CRPS on HRRR-ERA5 wind data.
GazePrior learns a 3D prior over eyes to synthesize realistic ground-truth data for training eye trackers on new devices without new real data collection.
3DEditSafe adds generation-stage guidance, 3D safety regularization, semantic projection, residue suppression, and mask-aware preservation to reduce unsafe semantic alignment in 3D editing while noting a safety-quality tradeoff.
AID amortizes guidance for diffusion inpainting by training a reusable module via an auxiliary Gaussian formulation and continuous-time actor-critic algorithm, improving quality-speed trade-off with under 1% overhead.
GeoSR-Bench is the first SR benchmark that directly measures how super-resolved remote sensing imagery improves performance on land cover segmentation, infrastructure mapping, and biophysical variable estimation rather than relying on fidelity metrics.
LPNSR derives optimal intermediate noise for diffusion SR via MLE and implements it with an LR-guided noise predictor, reaching SOTA perceptual quality in 4 steps without text priors.
Dreamer 4 is the first agent to obtain diamonds in Minecraft from only offline data by reinforcement learning inside a scalable world model that accurately predicts game mechanics.
SCOPE adds per-pixel action conditioning to pretrained video diffusion models and releases the CrossFPS multi-game dataset to support cross-game FPS world model simulation with zero-shot transfer.
Spatial Gram Alignment aligns internal self-similarities of LDM features with foundation priors to reconcile global structure and fine details in ultra-high-resolution text-to-image synthesis.
AVIS applies autoregressive diffusion models to video inverse problems by streaming restoration with measurement-consistent initialization, reducing latency from 114s to 4s and raising throughput to 1.18 FPS (or 5.91 FPS in the Flash variant).
RAE v2 reaches gFID 1.06 on ImageNet-256 in 80 epochs by combining multi-layer encoder sums, complementary REPA targets, and free guidance via output reparameterization.
InsightTok improves text and face fidelity in discrete image tokenization via content-aware perceptual losses, with gains transferring to autoregressive generation.
MotionCache accelerates autoregressive video generation up to 6.28x by motion-weighted cache reuse based on inter-frame differences, with negligible quality loss on SkyReels-V2 and MAGI-1.
Sphere clouds neutralize density attacks on private 3D maps for visual localization while depth guidance from ToF sensors restores translation scale for accurate pose estimation.
An end-to-end autoregressive model with a jointly trained 1D semantic tokenizer achieves state-of-the-art FID 1.48 on ImageNet 256x256 generation without guidance.
PEPS decomposes positional encodings into projected points with unique frequency-dependent motions to support more efficient learned grid-based encodings in INRs, outperforming prior methods on image, texture, and SDF tasks with often 25% fewer parameters.
UniT creates a unified physical language via visual anchoring and tri-branch reconstruction to enable scalable human-to-humanoid transfer for policy learning and world modeling.
A self-supervised framework using SURE and equivariant constraints produces super-resolved Sentinel-5P images comparable to supervised baselines without HR references and with physically plausible structures validated against EMIT data.
SciTikZer-8B uses a new dataset, benchmark, and dual self-consistency RL to generate TikZ code for scientific graphics, outperforming much larger models like Gemini-2.5-Pro.
RefTon is a flux-based virtual try-on method that uses unpaired reference images of the target garment on different people to guide texture and detail preservation in a streamlined person-to-person pipeline without body parsing or masks.
A post-adaptation pipeline on 3D foundation models yields higher-fidelity, faster, animation-ready single-image 3D characters than prior open methods on reported benchmarks.
Edit-GRPO decouples editing and preservation objectives via region-specific signals in a policy optimization framework to improve locality in image editing tasks.
SANA-WM is a 2.6B-parameter efficient world model that synthesizes minute-scale 720p videos with 6-DoF camera control, trained on 213K public clips in 15 days on 64 H100s and runnable on single GPUs at 36x higher throughput than prior open baselines.
citing papers explorer
-
GEAR: Guided End-to-End AutoRegression for Image Synthesis
GEAR jointly trains VQ tokenizer and AR generator end-to-end via dual hard/soft read-out and representation alignment, achieving up to 10x faster ImageNet gFID convergence than LlamaGen-REPA while generalizing across quantizers and to text-to-image.
-
Mind the Residual Gap: Probabilistic Downscaling under Real-World Bias
ReMatch corrects train-test residual distribution mismatch in probabilistic downscaling via optimal transport in low-dimensional PCA space, reducing under-dispersion and improving SSR and CRPS on HRRR-ERA5 wind data.
-
GazePrior: Zero-Shot AR/VR Eye Tracking via Learned 3D Gaze Reconstruction
GazePrior learns a 3D prior over eyes to synthesize realistic ground-truth data for training eye trackers on new devices without new real data collection.
-
3DEditSafe: Defending 3D Editing Pipelines from Unsafe Generation
3DEditSafe adds generation-stage guidance, 3D safety regularization, semantic projection, residue suppression, and mask-aware preservation to reduce unsafe semantic alignment in 3D editing while noting a safety-quality tradeoff.
-
Amortized Guidance for Image Inpainting with Pretrained Diffusion Models
AID amortizes guidance for diffusion inpainting by training a reusable module via an auxiliary Gaussian formulation and continuous-time actor-critic algorithm, improving quality-speed trade-off with under 1% overhead.
-
Beyond Visual Fidelity: Benchmarking Super-Resolution Models for Large-Scale Remote Sensing Imagery via Downstream Task Integration
GeoSR-Bench is the first SR benchmark that directly measures how super-resolved remote sensing imagery improves performance on land cover segmentation, infrastructure mapping, and biophysical variable estimation rather than relying on fidelity metrics.
-
LPNSR: Optimal Noise-Guided Diffusion Image Super-Resolution Via Learnable Noise Prediction
LPNSR derives optimal intermediate noise for diffusion SR via MLE and implements it with an LR-guided noise predictor, reaching SOTA perceptual quality in 4 steps without text priors.
-
Training Agents Inside of Scalable World Models
Dreamer 4 is the first agent to obtain diamonds in Minecraft from only offline data by reinforcement learning inside a scalable world model that accurately predicts game mechanics.
-
SCOPE: Simulating Cross-game Operations in Playable Environments for FPS World Models
SCOPE adds per-pixel action conditioning to pretrained video diffusion models and releases the CrossFPS multi-game dataset to support cross-game FPS world model simulation with zero-shot transfer.
-
Spatial Gram Alignment for Ultra-High-Resolution Image Synthesis
Spatial Gram Alignment aligns internal self-similarities of LDM features with foundation priors to reconcile global structure and fine details in ultra-high-resolution text-to-image synthesis.
-
Accelerating Video Inverse Problem Solvers with Autoregressive Diffusion Models
AVIS applies autoregressive diffusion models to video inverse problems by streaming restoration with measurement-consistent initialization, reducing latency from 114s to 4s and raising throughput to 1.18 FPS (or 5.91 FPS in the Flash variant).
-
Improved Baselines with Representation Autoencoders
RAE v2 reaches gFID 1.06 on ImageNet-256 in 80 epochs by combining multi-layer encoder sums, complementary REPA targets, and free guidance via output reparameterization.
-
InsightTok: Improving Text and Face Fidelity in Discrete Tokenization for Autoregressive Image Generation
InsightTok improves text and face fidelity in discrete image tokenization via content-aware perceptual losses, with gains transferring to autoregressive generation.
-
Motion-Aware Caching for Efficient Autoregressive Video Generation
MotionCache accelerates autoregressive video generation up to 6.28x by motion-weighted cache reuse based on inter-frame differences, with negligible quality loss on SkyReels-V2 and MAGI-1.
-
Depth-Guided Privacy-Preserving Visual Localization Using 3D Sphere Clouds
Sphere clouds neutralize density attacks on private 3D maps for visual localization while depth guidance from ToF sensors restores translation scale for accurate pose estimation.
-
End-to-End Autoregressive Image Generation with 1D Semantic Tokenizer
An end-to-end autoregressive model with a jointly trained 1D semantic tokenizer achieves state-of-the-art FID 1.48 on ImageNet 256x256 generation without guidance.
-
PEPS: Positional Encoding Projected Sampling -- Extended
PEPS decomposes positional encodings into projected points with unique frequency-dependent motions to support more efficient learned grid-based encodings in INRs, outperforming prior methods on image, texture, and SDF tasks with often 25% fewer parameters.
-
UniT: Toward a Unified Physical Language for Human-to-Humanoid Policy Learning and World Modeling
UniT creates a unified physical language via visual anchoring and tri-branch reconstruction to enable scalable human-to-humanoid transfer for policy learning and world modeling.
-
Self-Supervised Super-Resolution for Sentinel-5P Hyperspectral Images
A self-supervised framework using SURE and equivariant constraints produces super-resolved Sentinel-5P images comparable to supervised baselines without HR references and with physically plausible structures validated against EMIT data.
-
Scientific Graphics Program Synthesis via Dual Self-Consistency Reinforcement Learning
SciTikZer-8B uses a new dataset, benchmark, and dual self-consistency RL to generate TikZ code for scientific graphics, outperforming much larger models like Gemini-2.5-Pro.
-
RefTon: Reference person shot assist virtual Try-on
RefTon is a flux-based virtual try-on method that uses unpaired reference images of the target garment on different people to guide texture and detail preservation in a streamlined person-to-person pipeline without body parsing or masks.
-
DreamCharacter-1: From 3D Generative Foundation Models to Product-Ready Character Generation
A post-adaptation pipeline on 3D foundation models yields higher-fidelity, faster, animation-ready single-image 3D characters than prior open methods on reported benchmarks.
-
Edit-GRPO: A Locality-Preserving Policy Optimization Framework for Image Editing
Edit-GRPO decouples editing and preservation objectives via region-specific signals in a policy optimization framework to improve locality in image editing tasks.
-
SANA-WM: Efficient Minute-Scale World Modeling with Hybrid Linear Diffusion Transformer
SANA-WM is a 2.6B-parameter efficient world model that synthesizes minute-scale 720p videos with 6-DoF camera control, trained on 213K public clips in 15 days on 64 H100s and runnable on single GPUs at 36x higher throughput than prior open baselines.
-
ML-CLIPSim: Multi-Layer CLIP Similarity for Machine-Oriented Image Quality
ML-CLIPSim aggregates multi-layer patch and global similarities from frozen CLIP to approximate machine utility for images and outperforms standard IQA metrics on machine-preference tasks while staying competitive on human data.
-
FREPix: Frequency-Heterogeneous Flow Matching for Pixel-Space Image Generation
FREPix generates images by transporting low- and high-frequency wavelet components along separate schedules, reaching 1.91 FID on ImageNet 256×256.
-
Video Generation with Predictive Latents
PV-VAE improves video latent spaces for generation by unifying reconstruction with future-frame prediction, reporting 52% faster convergence and 34.42 FVD gain over Wan2.2 VAE on UCF101.
-
Seedance 1.0: Exploring the Boundaries of Video Generation Models
Seedance 1.0 generates 5-second 1080p videos in about 41 seconds with claimed superior motion quality, prompt adherence, and multi-shot consistency compared to prior models.
- HEART: Exploiting Head Heterogeneity in Sparse Attention for Video Diffusion