# Rectified Flow

Vessels reconstructed from low-resolution contrast-enhanced CT can come out broken, spurious, or wrongly connected, and image-quality metrics do not capture those structural errors. VascuSR separates vascular structure from CT intensity reconstruction: a conditional 3D rectified-flow refiner corrects vessels in label space, a lightweight controller injects the refined structure into the diffusion decoder, and each new super-resolved volume re-extracts the structure for the next round. On re-annotated MSD Task08 and HVM, it beats interpolation and learning-based baselines in Dice and clDice, cuts class-averaged connected-component counts by over 20% on both datasets and mean centerline distance by over 10% on MSD, while keeping image quality comparable to its SR backbone.

Vessels reconstructed from low-resolution contrast-enhanced CT can come out broken, spurious, or wrongly connected, and image-quality metrics do not capture those structural errors. VascuSR separates vascular structure from CT intensity reconstruction: a conditional 3D rectified-flow refiner corrects vessels in label space, a lightweight controller injects the refined structure into the diffusion decoder, and each new super-resolved volume re-extracts the structure for the next round. On re-annotated MSD Task08 and HVM, it beats interpolation and learning-based baselines in Dice and clDice, cuts class-averaged connected-component counts by over 20% on both datasets and mean centerline distance by over 10% on MSD, while keeping image quality comparable to its SR backbone.

SILSA is a topology-aware high-resolution 3D generation framework that replaces voxel tokens with a fixed set of overlapping slice latents along the three canonical axes, combining a SliceVAE, a Volumetric Anchor Lattice, and slice-level persistent-homology supervision to raise PSNR from 30.12 to 32.74, coverage from 73.12% to 79.08%, and reduce Betti error by 9.2% on image-to-3D generation while using only 384 tokens, cutting training memory by 40.4% and inference time by 58.5%.

SILSA is a topology-aware high-resolution 3D generation framework that replaces voxel tokens with a fixed set of overlapping slice latents along the three canonical axes, combining a SliceVAE, a Volumetric Anchor Lattice, and slice-level persistent-homology supervision to raise PSNR from 30.12 to 32.74, coverage from 73.12% to 79.08%, and reduce Betti error by 9.2% on image-to-3D generation while using only 384 tokens, cutting training memory by 40.4% and inference time by 58.5%.

The authors present ALICE, a foundation model trained once on a synthetic distribution family that turns mutual information estimation into in-context estimation of rectified-flow velocity fields for unseen distributions, matching per-distribution neural estimators on the Beyond Normal benchmark with at least twofold lower error at a 1k-sample budget and zero-shot reproducing and extending dedicated estimators' findings on biology, genetics, and neuroscience data the model never saw.

The authors present ALICE, a foundation model trained once on a synthetic distribution family that turns mutual information estimation into in-context estimation of rectified-flow velocity fields for unseen distributions, matching per-distribution neural estimators on the Beyond Normal benchmark with at least twofold lower error at a 1k-sample budget and zero-shot reproducing and extending dedicated estimators' findings on biology, genetics, and neuroscience data the model never saw.

Researchers at Adobe Research and KAIST introduce FlowTool, which uses conditional rectified flow to directly model the distribution of high-quality retouching tool parameters conditioned on the input image and user instruction, combining a vision-language model backbone with a Diffusion Transformer parameter generator and a tool-presence head instead of autoregressive MLLM reasoning and token-by-token numeric generation; across MMArt-Bench, FlowTool-Eval, ArtEdit-Bench, and MIT-Adobe5K it achieves stronger reference-based performance than specialized MLLM editing agents and proprietary MLLMs, remains competitive with proprietary models under reference-free evaluation, and reduces inference latency by at least 50x while requiring nearly 2x less memory.

Researchers at Adobe Research and KAIST introduce FlowTool, which uses conditional rectified flow to directly model the distribution of high-quality retouching tool parameters conditioned on the input image and user instruction, combining a vision-language model backbone with a Diffusion Transformer parameter generator and a tool-presence head instead of autoregressive MLLM reasoning and token-by-token numeric generation; across MMArt-Bench, FlowTool-Eval, ArtEdit-Bench, and MIT-Adobe5K it achieves stronger reference-based performance than specialized MLLM editing agents and proprietary MLLMs, remains competitive with proprietary models under reference-free evaluation, and reduces inference latency by at least 50x while requiring nearly 2x less memory.

The work proposes MPFlow, a zero-shot multi-modal MRI reconstruction framework built on rectified flow that uses a self-supervised pretraining strategy, PAMRI, to learn shared cross-modal representations and jointly guides the unconditional prior with data consistency and cross-modal feature alignment at inference; on HCP T2 4x super-resolution and BraTS FLAIR 8x k-space reconstruction it matches diffusion baselines in image quality using only 20% of the sampling steps while improving tumor segmentation Dice by 15% and reducing the SHAFE hallucination score by 26%.

The work proposes MPFlow, a zero-shot multi-modal MRI reconstruction framework built on rectified flow that uses a self-supervised pretraining strategy, PAMRI, to learn shared cross-modal representations and jointly guides the unconditional prior with data consistency and cross-modal feature alignment at inference; on HCP T2 4x super-resolution and BraTS FLAIR 8x k-space reconstruction it matches diffusion baselines in image quality using only 20% of the sampling steps while improving tumor segmentation Dice by 15% and reducing the SHAFE hallucination score by 26%.

The work introduces Latent-to-Latent Flow (L2L-Flow), which first compresses labels into a latent space with a volumetric label autoencoder and then learns a rectified flow between an image-conditional latent prior and frozen latent label representations, yielding stochastic segmentation on a private radiotherapy clinical-target-volume dataset (55 cases, 5-fold cross-validation) and the CURVAS multi-organ dataset (90 cases), with inference about 14x faster than full-resolution Flow-SSN (0.936 s/image versus 15.296 s/image) at a GED of 0.161, alongside a time-shifted noise schedule that improves Flow-SSN stability on high-dimensional volumetric data.

The work introduces Latent-to-Latent Flow (L2L-Flow), which first compresses labels into a latent space with a volumetric label autoencoder and then learns a rectified flow between an image-conditional latent prior and frozen latent label representations, yielding stochastic segmentation on a private radiotherapy clinical-target-volume dataset (55 cases, 5-fold cross-validation) and the CURVAS multi-organ dataset (90 cases), with inference about 14x faster than full-resolution Flow-SSN (0.936 s/image versus 15.296 s/image) at a GED of 0.161, alongside a time-shifted noise schedule that improves Flow-SSN stability on high-dimensional volumetric data.