optimal transport
Imagine you have two piles of dirt and want to reshape one into the other with minimal effort. Optimal transport is the mathematical way to find the cheapest plan for moving each grain, where "cost" is usually distance. It tells you exactly how much of each source point should go to each target point, and the minimal total cost is called the Wasserstein distance.
The papers show optimal transport is a versatile tool beyond earth moving. They use it to align different data types with few examples, to couple random samples more effectively during training, to reformulate robust optimization as a transport problem, to estimate how noisy each human preference is, and even to repair hallucinated features in vision-language models. In each case, enforcing or approximating optimal transport's structure, like monotone assignments or entropy-regularized matching, improves performance, robustness, or data efficiency.
- Papers
- 5
- Released code
- 0
- First seen
- Oct 2026
- Latest
- Oct 2026
5 papers in the last two months, against 0 in the two before.
Start here
Three papers to read first, in the order they came out- 1Optimal transport aligns pretrained modalities with only hundreds of paired anchors
Aligning two frozen unimodal encoders requires only 100 to 500 paired anchors when anchor-relative coordinates, entropic optimal transport, and distillation are combined.
Oct 2026 - 2Larger optimal transport batches make flow matching more stable
Larger online entropic OT minibatches reduce gradient variance and integration error in flow matching.
Oct 2026 - 3A few clean labels can fix sample-wise noise in reward model training
Reward models trained on noisy human feedback can be made unbiased by calibrating sample-wise noise with a few clean labels via optimal transport.
Oct 2026
How fast it is moving
Often studied with
adversarial trainingTraining a model on adversarially perturbed inputs to make it robust to attacks, often trading some clean accuracy for resilience.1 togethercontrastive alignmentA method that learns embeddings by minimizing distance between positive pairs and maximizing distance between negative pairs.1 togetherdistribution matching distillationA training method that aligns student and teacher output distributions to compress many-step generative models into efficient few-step versions.1 togetherThe papers
Most central to this idea first, not most recent.
- Research Labcs.AI
Optimal transport aligns pretrained modalities with only hundreds of paired anchors
Aligning two frozen unimodal encoders requires only 100 to 500 paired anchors when anchor-relative coordinates, entropic optimal transport, and distillation are combined.
Sorbonne Université, Institut du Cerveau - Paris Brain Institute - ICM · Oct 2026
- Independentcs.AI
A few clean labels can fix sample-wise noise in reward model training
Reward models trained on noisy human feedback can be made unbiased by calibrating sample-wise noise with a few clean labels via optimal transport.
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.CV
Training-free method repairs object hallucination in LVLMs instead of suppressing it
ResOT repairs hallucinated representations by aligning them to faithful distributions in a residual subspace, reducing object hallucination without suppressing useful information.
Wuhan University, Institute of Automation, Chinese Academy of Sciences · Oct 2026
- Big Tech
Cyclical monotonicity enforces efficient adversarial transport in robust learning
Adversarial training violates optimal transport geometry; enforcing cyclical monotonicity via MPA or ICNN improves robustness.
EPFL, Apple · Oct 2026
Concepts are extracted from each paper and reused across the corpus, so this page grows on its own as the desk reads.