{"total":43,"items":[{"citing_arxiv_id":"2606.10023","ref_index":29,"ref_count":1,"confidence":0.98,"is_internal_anchor":true,"paper_title":"Learning the Universe: Posterior Reliability of Neural Generative Models in High-Dimensional Field-Level Inference of Cosmic Initial Conditions","primary_cat":"astro-ph.CO","submitted_at":"2026-06-08T18:08:00+00:00","verdict":"UNVERDICTED","verdict_confidence":"LOW","novelty_score":6.0,"formal_verification":"none","one_line_summary":"Generative models for cosmological field-level inference can reproduce posterior means and cross-correlations yet fail to capture correct uncertainty geometry when validated against HMC reference samples.","context_count":0,"top_context_role":null,"top_context_polarity":null,"context_text":null},{"citing_arxiv_id":"2606.09576","ref_index":3,"ref_count":2,"confidence":0.98,"is_internal_anchor":true,"paper_title":"Characterizing Stellar Streams with Error-Aware Machine Learning","primary_cat":"astro-ph.GA","submitted_at":"2026-06-08T14:51:52+00:00","verdict":"UNVERDICTED","verdict_confidence":"LOW","novelty_score":7.0,"formal_verification":"none","one_line_summary":"SCREAM adapts the CATHODE method to treat stellar streams as feature-space over-densities, incorporates measurement uncertainties into neural network training, and achieves F1=0.745 on GD-1 while recovering faint members and a diffuse cocoon missed by prior methods.","context_count":0,"top_context_role":null,"top_context_polarity":null,"context_text":null},{"citing_arxiv_id":"2605.29373","ref_index":9,"ref_count":1,"confidence":0.98,"is_internal_anchor":true,"paper_title":"Deep Adaptive Dimension Reduction for Bayesian Inference in Inverse Problems","primary_cat":"cs.LG","submitted_at":"2026-05-28T05:21:19+00:00","verdict":"UNVERDICTED","verdict_confidence":"LOW","novelty_score":5.0,"formal_verification":"none","one_line_summary":"A new variational flow model with iterative prior updating and adaptive FNO surrogate for dimension-reduced Bayesian inference in high-dimensional PDE-governed inverse problems, reporting competitive accuracy versus MCMC, UKI, and SVGD on test cases.","context_count":0,"top_context_role":null,"top_context_polarity":null,"context_text":null},{"citing_arxiv_id":"2605.18731","ref_index":213,"ref_count":1,"confidence":0.98,"is_internal_anchor":true,"paper_title":"Pulse profile modelling of the 2024 outburst of the accreting millisecond pulsar SRGA J144459.2-604207","primary_cat":"astro-ph.HE","submitted_at":"2026-05-18T17:53:13+00:00","verdict":"CONDITIONAL","verdict_confidence":"LOW","novelty_score":5.0,"formal_verification":"none","one_line_summary":"Joint NICER+IXPE pulse-profile modeling of SRGA J144459.2-604207 favors large neutron-star mass and radius with two independent hotspots but shows strong sensitivity to joint-analysis methodology.","context_count":0,"top_context_role":null,"top_context_polarity":null,"context_text":null},{"citing_arxiv_id":"2605.17605","ref_index":6,"ref_count":1,"confidence":0.98,"is_internal_anchor":true,"paper_title":"Venom: A PyTorch Generative Modeling Toolkit","primary_cat":"cs.LG","submitted_at":"2026-05-17T19:06:46+00:00","verdict":"UNVERDICTED","verdict_confidence":"LOW","novelty_score":3.0,"formal_verification":"none","one_line_summary":"Venom is an educational PyTorch toolkit that packages multiple generative modeling families under a single MNIST-first interface with reproducible scripts and tutorials.","context_count":0,"top_context_role":null,"top_context_polarity":null,"context_text":null},{"citing_arxiv_id":"2605.17511","ref_index":78,"ref_count":1,"confidence":0.98,"is_internal_anchor":true,"paper_title":"A flow-matching generative model for event-by-event jet-induced hydro response in high-energy heavy-ion collisions","primary_cat":"nucl-th","submitted_at":"2026-05-17T15:46:39+00:00","verdict":"UNVERDICTED","verdict_confidence":"LOW","novelty_score":6.0,"formal_verification":"none","one_line_summary":"A flow-matching generative model trained on CoLBT-hydro data conditionally generates marginal final-state hadron spectra from jet-induced hydro responses in 0-10% Pb+Pb collisions at 5.02 TeV, matching training data statistics with approximately six orders of magnitude computational speedup.","context_count":0,"top_context_role":null,"top_context_polarity":null,"context_text":null},{"citing_arxiv_id":"2605.17137","ref_index":3,"ref_count":1,"confidence":0.98,"is_internal_anchor":true,"paper_title":"Latent Heuristic Search: Continuous Optimization for Automated Algorithm Design","primary_cat":"cs.AI","submitted_at":"2026-05-16T20:03:01+00:00","verdict":"UNVERDICTED","verdict_confidence":"LOW","novelty_score":7.0,"formal_verification":"none","one_line_summary":"Latent Heuristic Search performs continuous optimization over learned embeddings of heuristics, using normalizing flows and LLM prompting to discover competitive solvers for TSP, CVRP, KSP, and OBP.","context_count":0,"top_context_role":null,"top_context_polarity":null,"context_text":null},{"citing_arxiv_id":"2605.16486","ref_index":5,"ref_count":1,"confidence":0.98,"is_internal_anchor":true,"paper_title":"StAD: Stein Amortized Divergence for Fast Likelihoods with Diffusion and Flow","primary_cat":"stat.ML","submitted_at":"2026-05-15T18:00:00+00:00","verdict":"UNVERDICTED","verdict_confidence":"LOW","novelty_score":7.0,"formal_verification":"none","one_line_summary":"StAD distills divergence of PF-ODEs via the Langevin-Stein operator for faster, lower-variance likelihood estimation in generative models without Jacobian costs.","context_count":0,"top_context_role":null,"top_context_polarity":null,"context_text":null},{"citing_arxiv_id":"2605.15579","ref_index":18,"ref_count":1,"confidence":0.98,"is_internal_anchor":true,"paper_title":"TVRN: Invertible Neural Networks for Compression-Aware Temporal Video Rescaling","primary_cat":"eess.IV","submitted_at":"2026-05-15T03:44:14+00:00","verdict":"UNVERDICTED","verdict_confidence":"LOW","novelty_score":6.0,"formal_verification":"none","one_line_summary":"TVRN combines invertible wavelet-based networks with a surrogate gradient approximator and compression-aware asymmetric design to improve frame-rate rescaling quality under real codecs.","context_count":0,"top_context_role":null,"top_context_polarity":null,"context_text":null},{"citing_arxiv_id":"2605.08561","ref_index":15,"ref_count":1,"confidence":0.9,"is_internal_anchor":true,"paper_title":"CONTRA: Conformal Prediction Region via Normalizing Flow Transformation","primary_cat":"stat.ML","submitted_at":"2026-05-08T23:43:14+00:00","verdict":"UNVERDICTED","verdict_confidence":"LOW","novelty_score":6.0,"formal_verification":"none","one_line_summary":"CONTRA generates sharp multi-dimensional conformal prediction regions by defining nonconformity scores as distances from the center in the latent space of a normalizing flow.","context_count":1,"top_context_role":"background","top_context_polarity":"background","context_text":"Given a generative model for estimating the conditional density, PCP defines the non-conformity score of a data point (xi,y i)as si = min 1≤k≤K ∥yi − ˆyk i ∥, where{ ˆyi},K k=1 is a sample of the output generated from the estimated conditional density. Then the prediction region atx n+1 is obtained by generating a sample{ ˆyk n+1}K k=1 and form bCPCP(xn+1) = K[ k=1 {y:∥y− ˆyk n+1∥2 ≤s 1−α}, wheres 1−α is the⌈(1−α)(n 2 + 1)⌉-th smallest member of{s i, i∈I 2}. Note that each PCP prediction region is the union ofKballs. They have flexible, but often irregular, disconnected shapes, and are sensitive to the choice ofKandα. This brings challenge to interpreting the regions. Next, the ST-DQR method was proposed by Feldman et al. (2023). It learns an r-dimensional latent representation of the output, e.g., using the conditional variational auto-encoder (CV AE). The latent variable is encouraged to follow a unimodal distribution, so that methods like the directional quantile regression (DQR) are applicable to form convex probability regions for it. Samples are generated in the output space corresponding to points in the latent region. Calibration and the final prediction region are created similarly to the PCP. ST-DQR and our CONTRA are similar in that they both depend on latent representations, but differ substantially in the latent representation. CONTRA uses bijection and are often able to learn the latent variable to follow the Gaussian reference distribution rather closely. In contrast, the latent variable in ST-DQR is typically of lower dimension than the output, and its distribution is only coarsely similar to a reference. And this is why additional steps like DQR are needed to form latent probability regions in ST-DQR, but our CONTRA can directly use HDR of the reference Gaussian. When calibrating the regions, ST-DQR (and PCP) had to introduce yet another step: union balls around generated samples, and calibrate the common radius of the balls. Whereas CONTR"},{"citing_arxiv_id":"2605.08485","ref_index":18,"ref_count":1,"confidence":0.9,"is_internal_anchor":true,"paper_title":"Sinkhorn Treatment Effects: A Causal Optimal Transport Measure","primary_cat":"stat.ML","submitted_at":"2026-05-08T21:03:11+00:00","verdict":"UNVERDICTED","verdict_confidence":"LOW","novelty_score":7.0,"formal_verification":"none","one_line_summary":"The Sinkhorn treatment effect is a new entropic optimal transport measure of divergence between counterfactual distributions that admits first- and second-order pathwise differentiability, debiased estimators, and asymptotically valid tests for distributional treatment effects.","context_count":1,"top_context_role":"background","top_context_polarity":"background","context_text":"We refer the reader to Luedtke & Chung (2024) for a detailed exposition of pathwise differentiability. We briefly recall the key definitions needed for our analysis. As introduced in Sec. 2, let P ⊂ P(Z) denote our statistical model, where Z:=X × {0,1} × Y is a Polish space equipped with its Borel σ-algebra BZ. We assume that all distributions in P are dominated by aσ-finite measureλ. A parametric submodel (Pt :t∈[0, δ))⊂ P with P0 =P is said to bequadratic mean differentiable(QMD) at P if there exists a (Fisher) score functions∈L 2 0(P)such that √pt − √p−ts √p L2(λ) =o(t), where pt =dP t/dλ and p=dP/dλ . We denote by P(P,P, s) the collection of all QMD submodels at P with score function s. The set {s∈L 2 0(P) :P(P,P, s)̸=∅} is called thetangent setof P at P . Its closed linear span in L2 0(P) is the tangent space, denoted by ˙PP . Let Φ :P → H be a Hilbert-valued parameter. We say that Φ ispathwise differentiableat P , relative to the model P, if there exists a continuous linear operatorDΦ P : ˙PP → Hsuch that, for all(P t :t∈[0, δ))∈P(P,P, s), Φ(Pt)−Φ(P)−tDΦ P (s) H =o(t).(16) The operator DΦP is called thelocal parameterof Φ at P . Its image is a closed subspace of H, denoted by ˙HP , and referred to as the local parameter space. The Hermitian adjoint of DΦP , denoted by D∗ΦP :H → ˙PP , is called theefficient influence operator. The operators DΦP and D∗ΦP satisfy the adjoint relationship ⟨h, DΦP (s)⟩H =⟨D ∗ΦP (h), s⟩L2(P) for all h∈ H, s∈ ˙PP . We say that Φ admits an efficient influence function (EIF) atP if, for P -almost every z∈ Z , the map h7→D ∗ΦP (h)(z) defines a bounded linear functional on H. In this case, by the Riesz representation theorem, there exists a functionϕ P :Z → Hsuch that D∗ΦP (h)(z) =⟨h, ϕ P (z)⟩H ,for allh∈ HandP-a.e.z.(17) Sufficient conditions for the existence of the EIF when H and ˙HP are RKHS are given in (Luedtke & Chung, 2024, Theorem 1). In the special case of real-valued parameters, H=R , a function ϕP :Z →R is an influe"},{"citing_arxiv_id":"2605.08078","ref_index":9,"ref_count":2,"confidence":0.98,"is_internal_anchor":true,"paper_title":"Normalizing Trajectory Models","primary_cat":"cs.CV","submitted_at":"2026-05-08T17:57:14+00:00","verdict":"UNVERDICTED","verdict_confidence":"LOW","novelty_score":7.0,"formal_verification":"none","one_line_summary":"NTM models each generative reverse step as a conditional normalizing flow with a hybrid shallow-deep architecture, enabling exact-likelihood training and strong four-step sampling performance on text-to-image tasks.","context_count":0,"top_context_role":null,"top_context_polarity":null,"context_text":null},{"citing_arxiv_id":"2605.08029","ref_index":7,"ref_count":1,"confidence":0.9,"is_internal_anchor":true,"paper_title":"STARFlow2: Bridging Language Models and Normalizing Flows for Unified Multimodal Generation","primary_cat":"cs.CV","submitted_at":"2026-05-08T17:14:43+00:00","verdict":"UNVERDICTED","verdict_confidence":"LOW","novelty_score":6.0,"formal_verification":"none","one_line_summary":"STARFlow2 presents an autoregressive flow-based architecture for unified multimodal text-image generation by interleaving a VLM stream with a TarFlow stream via residual skips and a unified latent space.","context_count":0,"top_context_role":null,"top_context_polarity":null,"context_text":null},{"citing_arxiv_id":"2605.07100","ref_index":45,"ref_count":1,"confidence":0.9,"is_internal_anchor":true,"paper_title":"TRACE: Transport Alignment Conformal Prediction via Diffusion and Flow Matching Models","primary_cat":"stat.ML","submitted_at":"2026-05-08T01:28:57+00:00","verdict":"UNVERDICTED","verdict_confidence":"LOW","novelty_score":7.0,"formal_verification":"none","one_line_summary":"TRACE creates valid conformal prediction sets for complex generative models by scoring outputs via averaged denoising or velocity errors along stochastic transport paths instead of likelihoods.","context_count":0,"top_context_role":null,"top_context_polarity":null,"context_text":null},{"citing_arxiv_id":"2605.01217","ref_index":4,"ref_count":1,"confidence":0.9,"is_internal_anchor":true,"paper_title":"Asymmetric Invertible Threat: Learning Reversible Privacy Defense for Face Recognition","primary_cat":"cs.CV","submitted_at":"2026-05-02T03:18:36+00:00","verdict":"UNVERDICTED","verdict_confidence":"LOW","novelty_score":6.0,"formal_verification":"none","one_line_summary":"ARFP is a key-conditioned reversible face cloaking method that resists unauthorized restoration attacks while enabling authorized recovery with tamper indication.","context_count":0,"top_context_role":null,"top_context_polarity":null,"context_text":null},{"citing_arxiv_id":"2604.25129","ref_index":1,"ref_count":1,"confidence":0.9,"is_internal_anchor":true,"paper_title":"8DNA: 8D Neural Asset Light Transport by Distribution Learning","primary_cat":"cs.GR","submitted_at":"2026-04-28T02:07:42+00:00","verdict":"UNVERDICTED","verdict_confidence":"LOW","novelty_score":6.0,"formal_verification":"none","one_line_summary":"8DNA learns the complete 8D light transport function from path-traced samples via distribution learning to support accurate near-field global illumination rendering of complex 3D assets.","context_count":0,"top_context_role":null,"top_context_polarity":null,"context_text":null},{"citing_arxiv_id":"2604.24322","ref_index":38,"ref_count":1,"confidence":0.9,"is_internal_anchor":true,"paper_title":"Generative Design of a Gas Turbine Combustor Using Invertible Neural Networks","primary_cat":"cs.AI","submitted_at":"2026-04-27T11:14:06+00:00","verdict":"UNVERDICTED","verdict_confidence":"LOW","novelty_score":5.0,"formal_verification":"none","one_line_summary":"Invertible Neural Networks are used to generate gas turbine combustor designs that meet specified performance criteria from a training database of parameterized designs and simulations.","context_count":0,"top_context_role":null,"top_context_polarity":null,"context_text":null},{"citing_arxiv_id":"2604.21209","ref_index":12,"ref_count":1,"confidence":0.9,"is_internal_anchor":true,"paper_title":"Align Generative Artificial Intelligence with Human Preferences: A Novel Large Language Model Fine-Tuning Method for Online Review Management","primary_cat":"cs.AI","submitted_at":"2026-04-23T02:01:37+00:00","verdict":"UNVERDICTED","verdict_confidence":"LOW","novelty_score":5.0,"formal_verification":"none","one_line_summary":"A preference fine-tuning method for LLMs that combines context augmentation, theory-driven preference pair construction, curriculum learning, and a density estimation support constraint to produce domain-aligned review responses with reduced hallucinations and over-conservatism.","context_count":0,"top_context_role":null,"top_context_polarity":null,"context_text":null},{"citing_arxiv_id":"2604.17257","ref_index":40,"ref_count":1,"confidence":0.9,"is_internal_anchor":true,"paper_title":"REZE: Representation Regularization for Domain-adaptive Text Embedding Pre-finetuning","primary_cat":"cs.CL","submitted_at":"2026-04-19T04:41:55+00:00","verdict":"UNVERDICTED","verdict_confidence":"LOW","novelty_score":6.0,"formal_verification":"none","one_line_summary":"REZE controls representation shifts in contrastive pre-finetuning of text embeddings via eigenspace decomposition of anchor-positive pairs and adaptive soft-shrinkage on task-variant directions.","context_count":0,"top_context_role":null,"top_context_polarity":null,"context_text":null},{"citing_arxiv_id":"2605.04060","ref_index":6,"ref_count":1,"confidence":0.9,"is_internal_anchor":true,"paper_title":"Lookahead Drifting Model","primary_cat":"cs.LG","submitted_at":"2026-04-10T11:12:21+00:00","verdict":"UNVERDICTED","verdict_confidence":"LOW","novelty_score":6.0,"formal_verification":"none","one_line_summary":"The lookahead drifting model improves upon the drifting model by sequentially computing multiple drifting terms that incorporate higher-order gradient information, leading to better performance on toy examples and CIFAR10.","context_count":0,"top_context_role":null,"top_context_polarity":null,"context_text":null},{"citing_arxiv_id":"2604.07664","ref_index":49,"ref_count":1,"confidence":0.9,"is_internal_anchor":true,"paper_title":"Monocular Depth Estimation From the Perspective of Feature Restoration: A Diffusion Enhanced Depth Restoration Approach","primary_cat":"cs.CV","submitted_at":"2026-04-09T00:13:53+00:00","verdict":"CONDITIONAL","verdict_confidence":"LOW","novelty_score":6.0,"formal_verification":"none","one_line_summary":"Monocular depth estimation is recast as indirect feature restoration via an invertible diffusion module plus auxiliary viewpoint enhancement, delivering 4-38% RMSE gains on KITTI over baselines.","context_count":0,"top_context_role":null,"top_context_polarity":null,"context_text":null},{"citing_arxiv_id":"2604.06348","ref_index":32,"ref_count":1,"confidence":0.9,"is_internal_anchor":true,"paper_title":"Dartmouth Stellar Evolution Emulator (DSEE) 1: Generative Stellar Evolution Model Database","primary_cat":"astro-ph.SR","submitted_at":"2026-04-07T18:27:53+00:00","verdict":"UNVERDICTED","verdict_confidence":"LOW","novelty_score":6.0,"formal_verification":"none","one_line_summary":"DSEE is a flow-based emulator that generates stellar evolution tracks and isochrones as probabilistic outputs from a single model trained on millions of simulations, enabling fast interpolation and uncertainty-aware analyses.","context_count":0,"top_context_role":null,"top_context_polarity":null,"context_text":null},{"citing_arxiv_id":"2604.03511","ref_index":24,"ref_count":1,"confidence":0.9,"is_internal_anchor":true,"paper_title":"Monte Carlo Event Generation with Continuous Normalizing Flows","primary_cat":"hep-ph","submitted_at":"2026-04-03T23:18:12+00:00","verdict":"CONDITIONAL","verdict_confidence":"LOW","novelty_score":6.0,"formal_verification":"none","one_line_summary":"Continuous normalizing flows improve unweighting efficiency in Monte Carlo event generation for high-jet-multiplicity collider processes by factors up to 184, with wall-time gains of about ten when combined with coupling-layer flows.","context_count":1,"top_context_role":"baseline","top_context_polarity":"baseline","context_text":"the Flow Matching method [34-37] to solve this problem. We evaluate our approach on the two most computation- ally intensive processes simulated for the LHC: lepton- pair production and top-quark pair production with mul- tiple jets. We compare performance using the unweight- ing efficiencyϵas our key metric, benchmarking against Vegas-based methods and Normalizing Flows based on Coupling Layers [24, 38, 39]. Our results demonstrate a significant increase inϵ: for the highest-multiplicity processes, CNFs improveϵby factors of up to 184, com- pared to the other methods. This is enabled in part by conditioning on helicity configurations, which allows the model to learn correlations between discrete and contin- uous features. Phase-space sampling- In collider simulations, the"},{"citing_arxiv_id":"2604.08582","ref_index":33,"ref_count":1,"confidence":0.98,"is_internal_anchor":true,"paper_title":"Multivariate Time Series Anomaly Detection via Dual-Branch Reconstruction and Autoregressive Flow-based Residual Density Estimation","primary_cat":"cs.LG","submitted_at":"2026-03-29T15:18:14+00:00","verdict":"UNVERDICTED","verdict_confidence":"LOW","novelty_score":6.0,"formal_verification":"none","one_line_summary":"DBR-AF decouples cross-variable correlations in reconstruction and applies autoregressive flows to model residual densities for improved anomaly detection in multivariate time series.","context_count":0,"top_context_role":null,"top_context_polarity":null,"context_text":null},{"citing_arxiv_id":"2603.14135","ref_index":29,"ref_count":1,"confidence":0.98,"is_internal_anchor":true,"paper_title":"Conditional flow matching for physics-constrained inverse problems with finite training data","primary_cat":"stat.ML","submitted_at":"2026-03-14T21:43:48+00:00","verdict":"UNVERDICTED","verdict_confidence":"LOW","novelty_score":6.0,"formal_verification":"none","one_line_summary":"Conditional flow matching learns a velocity field to sample from measurement-conditioned posteriors in physics inverse problems, with early stopping to prevent variance collapse and selective memorization under finite training data.","context_count":0,"top_context_role":null,"top_context_polarity":null,"context_text":null},{"citing_arxiv_id":"2602.03875","ref_index":6,"ref_count":1,"confidence":0.98,"is_internal_anchor":true,"paper_title":"Reversible Deep Learning for 13C NMR in Chemoinformatics: On Structures and Spectra","primary_cat":"cs.LG","submitted_at":"2026-02-01T18:48:31+00:00","verdict":"UNVERDICTED","verdict_confidence":"LOW","novelty_score":6.0,"formal_verification":"none","one_line_summary":"A conditional invertible neural network unifies forward prediction of 13C NMR spectra from structures and inverse generation of structure candidates from spectra.","context_count":0,"top_context_role":null,"top_context_polarity":null,"context_text":null},{"citing_arxiv_id":"2601.15739","ref_index":8,"ref_count":1,"confidence":0.98,"is_internal_anchor":true,"paper_title":"Breaking the Resolution Barrier: Arbitrary-resolution Deep Image Steganography Framework","primary_cat":"cs.CV","submitted_at":"2026-01-22T08:07:10+00:00","verdict":"UNVERDICTED","verdict_confidence":"LOW","novelty_score":7.0,"formal_verification":"none","one_line_summary":"ARDIS enables arbitrary-resolution deep image steganography via frequency decoupling in hiding and latent-guided implicit reconstruction for blind recovery.","context_count":0,"top_context_role":null,"top_context_polarity":null,"context_text":null},{"citing_arxiv_id":"2512.04153","ref_index":63,"ref_count":1,"confidence":0.98,"is_internal_anchor":true,"paper_title":"Data-Driven Predictions for Dark Photon and Millicharged Particle Production","primary_cat":"hep-ph","submitted_at":"2025-12-03T19:00:00+00:00","verdict":"UNVERDICTED","verdict_confidence":"LOW","novelty_score":7.0,"formal_verification":"none","one_line_summary":"A data-driven framework using normalizing flows predicts the rate and kinematic distributions of dark photon and millicharged particle production directly from measured dilepton events.","context_count":0,"top_context_role":null,"top_context_polarity":null,"context_text":null},{"citing_arxiv_id":"2511.00956","ref_index":36,"ref_count":1,"confidence":0.98,"is_internal_anchor":true,"paper_title":"RefTon: Reference person shot assist virtual Try-on","primary_cat":"cs.CV","submitted_at":"2025-11-02T14:32:31+00:00","verdict":"UNVERDICTED","verdict_confidence":"LOW","novelty_score":6.0,"formal_verification":"none","one_line_summary":"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.","context_count":0,"top_context_role":null,"top_context_polarity":null,"context_text":null},{"citing_arxiv_id":"2510.21033","ref_index":25,"ref_count":1,"confidence":0.98,"is_internal_anchor":true,"paper_title":"Iso-Riemannian Optimization on Learned Data Manifolds","primary_cat":"math.OC","submitted_at":"2025-10-23T22:34:55+00:00","verdict":"UNVERDICTED","verdict_confidence":"LOW","novelty_score":7.0,"formal_verification":"none","one_line_summary":"Iso-Riemannian descent algorithm with convergence analysis under iso-convexity, iso-monotonicity and iso-Lipschitz conditions for optimization on learned Riemannian manifolds from data.","context_count":0,"top_context_role":null,"top_context_polarity":null,"context_text":null},{"citing_arxiv_id":"2509.12884","ref_index":8,"ref_count":1,"confidence":0.98,"is_internal_anchor":true,"paper_title":"Modeling nonstationary spatial processes with normalizing flows","primary_cat":"stat.ME","submitted_at":"2025-09-16T09:37:18+00:00","verdict":"UNVERDICTED","verdict_confidence":"LOW","novelty_score":7.0,"formal_verification":"none","one_line_summary":"Neural autoregressive flows enable flexible high-dimensional spatial warpings for nonstationary anisotropic processes, with simulations showing greater representational capacity than standard models and an application to 3D Argo Floats data.","context_count":0,"top_context_role":null,"top_context_polarity":null,"context_text":null},{"citing_arxiv_id":"2402.17177","ref_index":9,"ref_count":1,"confidence":0.9,"is_internal_anchor":true,"paper_title":"Sora: A Review on Background, Technology, Limitations, and Opportunities of Large Vision Models","primary_cat":"cs.CV","submitted_at":"2024-02-27T03:30:58+00:00","verdict":"UNVERDICTED","verdict_confidence":"LOW","novelty_score":2.0,"formal_verification":"none","one_line_summary":"The paper reviews the background, technology, applications, limitations, and future directions of OpenAI's Sora text-to-video generative model based on public information.","context_count":1,"top_context_role":"background","top_context_polarity":"background","context_text":"However, these methods were limited in their capacity to produce complex and vivid images. 4 Figure 3: History of Generative AI in Vision Domain. The introduction of Generative Adversarial Networks (GANs) [7] and Variational Autoencoders (V AEs) [8] marked a significant turning point due to its remarkable capabilities across various applications. Subsequent developments, such as flow models [9] and diffusion models [10], further enhanced image generation with greater detail and quality. The recent progress in Artificial Intelligence Generated Content (AIGC) technolo- gies has democratized content creation, enabling users to generate desired content through simple textual instructions [11]. Over the past decade, the development of generative CV models has taken various routes, as shown"},{"citing_arxiv_id":"2402.14212","ref_index":10,"ref_count":1,"confidence":0.98,"is_internal_anchor":true,"paper_title":"Moonwalk: Inverse-Forward Differentiation","primary_cat":"cs.LG","submitted_at":"2024-02-22T01:33:31+00:00","verdict":"UNVERDICTED","verdict_confidence":"LOW","novelty_score":7.0,"formal_verification":"none","one_line_summary":"Moonwalk enables memory-efficient training of deep networks via mixed-mode gradient computation with vector-inverse-Jacobian products for submersive layers and fragmental checkpointing otherwise, matching backprop runtime at over twice the depth.","context_count":0,"top_context_role":null,"top_context_polarity":null,"context_text":null},{"citing_arxiv_id":"2302.08453","ref_index":7,"ref_count":1,"confidence":0.98,"is_internal_anchor":true,"paper_title":"T2I-Adapter: Learning Adapters to Dig out More Controllable Ability for Text-to-Image Diffusion Models","primary_cat":"cs.CV","submitted_at":"2023-02-16T17:56:08+00:00","verdict":"UNVERDICTED","verdict_confidence":"LOW","novelty_score":6.0,"formal_verification":"none","one_line_summary":"T2I-Adapters are lightweight modules that enable fine-grained control over color and structure in text-to-image diffusion models by aligning external conditions with the frozen model's internal knowledge.","context_count":0,"top_context_role":null,"top_context_polarity":null,"context_text":null},{"citing_arxiv_id":"2209.03003","ref_index":13,"ref_count":1,"confidence":0.9,"is_internal_anchor":true,"paper_title":"Flow Straight and Fast: Learning to Generate and Transfer Data with Rectified Flow","primary_cat":"cs.LG","submitted_at":"2022-09-07T08:59:55+00:00","verdict":"UNVERDICTED","verdict_confidence":"LOW","novelty_score":8.0,"formal_verification":"none","one_line_summary":"Rectified flow learns straight-path neural ODEs for distribution transport, yielding efficient generative models and domain transfers that work well even with a single simulation step.","context_count":1,"top_context_role":"background","top_context_polarity":"background","context_text":"[11] Valentin De Bortoli, James Thornton, Jeremy Heng, and Arnaud Doucet. Diffusion Schr ¨odinger bridge with applications to score-based generative modeling. Advances in Neural Information Pro- cessing Systems, 34, 2021. [12] Prafulla Dhariwal and Alexander Nichol. Diffusion models beat GANs on image synthesis. Advances in Neural Information Processing Systems, 34, 2021. [13] Laurent Dinh, David Krueger, and Yoshua Bengio. Nice: Non-linear independent components esti- mation. arXiv preprint arXiv:1410.8516, 2014. [14] Laurent Dinh, Jascha Sohl-Dickstein, and Samy Bengio. Density estimation using real nvp. arXiv preprint arXiv:1605.08803, 2016. [15] Alessio Figalli and Federico Glaudo. An Invitation to Optimal Transport, Wasserstein Distances, and"},{"citing_arxiv_id":"2104.10157","ref_index":12,"ref_count":1,"confidence":0.9,"is_internal_anchor":true,"paper_title":"VideoGPT: Video Generation using VQ-VAE and Transformers","primary_cat":"cs.CV","submitted_at":"2021-04-20T17:58:03+00:00","verdict":"ACCEPT","verdict_confidence":"MODERATE","novelty_score":6.0,"formal_verification":"none","one_line_summary":"VideoGPT generates competitive natural videos by learning discrete latents with VQ-VAE and modeling them autoregressively with a transformer.","context_count":0,"top_context_role":null,"top_context_polarity":null,"context_text":null},{"citing_arxiv_id":"2011.13456","ref_index":55,"ref_count":1,"confidence":0.98,"is_internal_anchor":true,"paper_title":"Score-Based Generative Modeling through Stochastic Differential Equations","primary_cat":"cs.LG","submitted_at":"2020-11-26T19:39:10+00:00","verdict":"UNVERDICTED","verdict_confidence":"LOW","novelty_score":8.0,"formal_verification":"none","one_line_summary":"Introduces an SDE-based framework for score-based generative modeling that unifies prior methods, enables predictor-corrector sampling and neural ODE likelihoods, and achieves SOTA unconditional image generation on CIFAR-10.","context_count":0,"top_context_role":null,"top_context_polarity":null,"context_text":null},{"citing_arxiv_id":"2006.11239","ref_index":9,"ref_count":1,"confidence":0.9,"is_internal_anchor":true,"paper_title":"Denoising Diffusion Probabilistic Models","primary_cat":"cs.LG","submitted_at":"2020-06-19T17:24:44+00:00","verdict":"ACCEPT","verdict_confidence":"MODERATE","novelty_score":8.0,"formal_verification":"none","one_line_summary":"Denoising diffusion probabilistic models generate high-quality images by learning to reverse a fixed forward diffusion process, achieving FID 3.17 on CIFAR10.","context_count":1,"top_context_role":"background","top_context_polarity":"background","context_text":"reverse process produces high-quality reconstructions, and plausible interpolations that smoothly vary attributes such as pose, skin tone, hairstyle, expression and background, but not eyewear. Larger t results in coarser and more varied interpolations, with novel samples att = 1000 (Appendix Fig. 9). 5 Related Work While diffusion models might resemble flows [ 9, 46, 10, 32, 5, 16, 23] and V AEs [33, 47, 37], diffusion models are designed so thatq has no parameters and the top-level latent xT has nearly zero mutual information with the data x0. Our ϵ-prediction reverse process parameterization establishes a connection between diffusion models and denoising score matching over multiple noise levels with annealed Langevin dynamics for sampling [55, 56]."},{"citing_arxiv_id":"1907.05600","ref_index":11,"ref_count":1,"confidence":0.98,"is_internal_anchor":true,"paper_title":"Generative Modeling by Estimating Gradients of the Data Distribution","primary_cat":"cs.LG","submitted_at":"2019-07-12T07:37:26+00:00","verdict":"UNVERDICTED","verdict_confidence":"LOW","novelty_score":6.0,"formal_verification":"none","one_line_summary":"Score-based generative modeling via multi-noise-level score matching and annealed Langevin dynamics produces samples on par with GANs and sets a new inception score record on CIFAR-10.","context_count":0,"top_context_role":null,"top_context_polarity":null,"context_text":null},{"citing_arxiv_id":"1907.02392","ref_index":7,"ref_count":1,"confidence":0.98,"is_internal_anchor":true,"paper_title":"Guided Image Generation with Conditional Invertible Neural Networks","primary_cat":"cs.CV","submitted_at":"2019-07-04T13:20:57+00:00","verdict":"UNVERDICTED","verdict_confidence":"LOW","novelty_score":7.0,"formal_verification":"none","one_line_summary":"Proposes cINN architecture for conditional image generation that by construction yields diverse sharp samples, demonstrated on MNIST digit generation and image colorization with latent space manipulation.","context_count":0,"top_context_role":null,"top_context_polarity":null,"context_text":null},{"citing_arxiv_id":"1806.07366","ref_index":13,"ref_count":1,"confidence":0.98,"is_internal_anchor":true,"paper_title":"Neural Ordinary Differential Equations","primary_cat":"cs.LG","submitted_at":"2018-06-19T17:50:12+00:00","verdict":"ACCEPT","verdict_confidence":"HIGH","novelty_score":8.0,"formal_verification":"none","one_line_summary":"Neural networks are redefined as continuous dynamical systems by learning the derivative of the hidden state with a neural network and integrating it with an ODE solver.","context_count":0,"top_context_role":null,"top_context_polarity":null,"context_text":null},{"citing_arxiv_id":"1605.08803","ref_index":17,"ref_count":1,"confidence":0.9,"is_internal_anchor":true,"paper_title":"Density estimation using Real NVP","primary_cat":"cs.LG","submitted_at":"2016-05-27T21:24:32+00:00","verdict":"ACCEPT","verdict_confidence":"MODERATE","novelty_score":8.0,"formal_verification":"none","one_line_summary":"Real NVP uses affine coupling layers to create invertible transformations that support exact density estimation, sampling, and latent inference without approximations.","context_count":1,"top_context_role":"extension","top_context_polarity":"extend","context_text":"In this paper, we will tackle the problem of learning highly nonlinear models in high-dimensional continuous spaces through maximum likelihood. In order to optimize the log-likelihood, we introduce a more flexible class of architectures that enables the computation of log-likelihood on continuous data using the change of variable formula. Building on our previous work in [ 17], we define a powerful class of bijective functions which enable exact and tractable density evaluation and exact and tractable inference. Moreover, the resulting cost function does not to rely on a fixed form reconstruction cost such as square error [38, 47], and generates sharper samples as a result. Also, this flexibility helps us leverage recent advances in batch normalization [31] and residual networks"},{"citing_arxiv_id":"1503.03585","ref_index":39,"ref_count":1,"confidence":0.9,"is_internal_anchor":true,"paper_title":"Deep Unsupervised Learning using Nonequilibrium Thermodynamics","primary_cat":"cs.LG","submitted_at":"2015-03-12T04:51:37+00:00","verdict":"ACCEPT","verdict_confidence":"MODERATE","novelty_score":8.0,"formal_verification":"none","one_line_summary":"A forward diffusion process adds noise iteratively to data until it is unstructured, and a neural network learns the reverse process to generate new samples from the original distribution.","context_count":1,"top_context_role":"background","top_context_polarity":"background","context_text":"X(t−1)|X(0) ) −Hq ( X(t)|X(0) ) , (36) where both the upper and lower bounds depend only on the conditional forward trajectory q ( x(1···T)|x(0)) , and can be analytically computed. B. Log Likelihood Lower Bound The lower bound on the log likelihood is L≥K (37) K = ∫ dx(0···T)q ( x(0···T) ) log [ p ( x(T) ) T∏ t=1 p ( x(t−1)|x(t)) q ( x(t)|x(t−1)) ] (38) (39) Deep Unsupervised Learning using Nonequilibrium Thermodynamics B.1. Entropy of p ( X(T)) We can peel off the contribution fromp ( X(T)) , and rewrite it as an entropy, K = ∫ dx(0···T)q ( x(0···T) ) T∑ t=1 log [ p ( x(t−1)|x(t)) q ( x(t)|x(t−1)) ] + ∫ dx(T)q ( x(T) ) logp ( x(T) ) (40) = ∫ dx(0···T)q ( x(0···T) ) T∑ t=1 log [ p ( x(t−1)|x(t)) q ("}],"limit":50,"offset":0}