# Flow matching

Published articles for Flow matching.

This is one page of public article previews, not the complete archive. Follow Next page to continue. Summaries are not the original full articles.

## Physical Laws without Physics Loss? Probing Learned Dynamics in Video World Models through Forced Response

DevFeed: [Physical Laws without Physics Loss? Probing Learned Dynamics in Video World Models through Forced Response](<https://devfeed.tech/articles/physical-laws-without-physics-loss-probing-learned-dynamics-in-video-world-models-through-forced-response-145081.md>)

Original publisher: [Read original article](<https://arxiv.org/abs/2611.25626>)

Author: Yutao Li, Chen Wang, Wangmeng Zuo, Hui Li

Published: 2026-11-20T05:00:00Z

Content type: article

Language: en

Sources: [arXiv: Machine Learning](<https://devfeed.tech/sources/arxiv-machine-learning.md>)

Topics: [World models](<https://devfeed.tech/topics/world-models.md>), [Computer vision](<https://devfeed.tech/topics/computer-vision.md>), [Generative AI](<https://devfeed.tech/topics/generative-ai.md>)

Tags: [cs-cv](<https://devfeed.tech/tags/cs-cv.md>), [cs-lg](<https://devfeed.tech/tags/cs-lg.md>), [flow-matching](<https://devfeed.tech/tags/flow-matching.md>), [image-to-video](<https://devfeed.tech/tags/image-to-video.md>), [laws](<https://devfeed.tech/tags/laws.md>), [motion](<https://devfeed.tech/tags/motion.md>), [object](<https://devfeed.tech/tags/object.md>), [physics](<https://devfeed.tech/tags/physics.md>), [probing](<https://devfeed.tech/tags/probing.md>), [response](<https://devfeed.tech/tags/response.md>), [training](<https://devfeed.tech/tags/training.md>), [video](<https://devfeed.tech/tags/video.md>), [world-models](<https://devfeed.tech/tags/world-models.md>)

### AI overview

The study tests whether video world models learn dynamics beyond replay by shaking the support of suspended objects and evaluating forced responses. Six image-to-video backbones are adapted with visual and numeric conditioning using only flow-matching loss. All six reproduce linear response, including resonance, and five capture higher-order effects such as nonlinear softening, despite softening accounting for only 0.74% of the swing's squared displacement in training. Wan generalizes to unseen excitations, objects, poses, and scenes; counterfactual tests suggest it does more than replay training videos. Extensions to a double pendulum and rocking block find sustained dynamics are learned more faithfully than brief impacts. The excerpt also cautions that learned dynamics may not emerge when state estimation or propagation fails.

### Source excerpt

arXiv:2611.25626v1 Announce Type: cross Abstract: Do video world models learn physical laws without a physics loss, or merely imitate motion? Overall motion and appearance scores cannot settle it. We investigate this through nonlinear forced response: an excitation shakes the support of a suspended object. We introduce dual visual-numeric conditioning and adapt six image-to-video backbones to predict this motion using only the flow-matching loss. Wan predicts motion accurately and generalizes to unseen excitations, objects, poses, and scenes. Counterfactual tests suggest capabilities beyond replaying training videos. But what laws have they learned? We design three tiers of tests using Taylor expansions of the dynamics and geometry. All six reproduce the linear response, including resonance; five capture higher-order dynamics, including nonlinear softening. Remarkably, softening is barely present in training: only 0.74% of the swing's squared displacement. Extensions to a double pendulum and a rocking block show sustained dynamics are learned more faithfully than brief impacts. Yet learned dynamics may fail to appear when state estimation or propagation goes wrong. We turn these tests and findings into strategies to improve video training and generation.

## ActMem: From Latent Memory Reasoning to Grounded Robotic Action

DevFeed: [ActMem: From Latent Memory Reasoning to Grounded Robotic Action](<https://devfeed.tech/articles/actmem-from-latent-memory-reasoning-to-grounded-robotic-action-145059.md>)

Original publisher: [Read original article](<https://arxiv.org/abs/2611.25461>)

Author: Linfeng Dong, Jianwei Niu, Jiuyi Zhao, Ruibing Feng, Xinghao Wu, Wu Yang, Shuai Tao, Wulong Liu, Mu Li

Published: 2026-11-20T05:00:00Z

Content type: article

Language: en

Sources: [arXiv: Machine Learning](<https://devfeed.tech/sources/arxiv-machine-learning.md>)

Topics: [Physical AI](<https://devfeed.tech/topics/physical-ai.md>), [Artificial Intelligence](<https://devfeed.tech/topics/ai.md>)

Tags: [attention](<https://devfeed.tech/tags/attention.md>), [chain-of-thought](<https://devfeed.tech/tags/chain-of-thought.md>), [cs-cv](<https://devfeed.tech/tags/cs-cv.md>), [cs-lg](<https://devfeed.tech/tags/cs-lg.md>), [flow-matching](<https://devfeed.tech/tags/flow-matching.md>), [grounded](<https://devfeed.tech/tags/grounded.md>), [grounding](<https://devfeed.tech/tags/grounding.md>), [hidden-states](<https://devfeed.tech/tags/hidden-states.md>), [history](<https://devfeed.tech/tags/history.md>), [horizon](<https://devfeed.tech/tags/horizon.md>), [inference](<https://devfeed.tech/tags/inference.md>), [memory](<https://devfeed.tech/tags/memory.md>), [reasoning](<https://devfeed.tech/tags/reasoning.md>), [robotic](<https://devfeed.tech/tags/robotic.md>)

### AI overview

ActMem separates memory reasoning from grounding actions in current perception within one vision-language backbone. It infers a subtask from timestamped interaction history, retains the resulting hidden states as a Latent Memory Chain-of-Thought to condition a flow-matching action expert, and lets that expert attend directly to the current view. Efficient Memory CoT Inference reuses computation across consecutive decisions without changing the output. The abstract reports 71.1% task success and 80.8% completion success on RoboMemArena, respectively 7.5 and 8.7 percentage points above the strongest baseline; 73.8% average success on MIKASA-Robo; and 68.8% average success on real-world memory tasks, 10 points above the strongest memory-augmented VLA. These are reported results from the abstract; further experimental details are not provided.

### Source excerpt

arXiv:2611.25461v1 Announce Type: cross Abstract: Vision-Language-Action (VLA) models have shown promising results in real-world long-horizon manipulation, where memory plays a crucial role. However, existing memory-augmented VLAs either retrieve past observations without reasoning about what they imply, or reason over the full history in a shared context where it crowds out current perception. Cognitive science treats them as distinct systems: memory reconstructs past experience to decide what to do, while perception guides how to act. We therefore propose ActMem, an in-context memory framework that decouples memory reasoning from perceptual grounding within a single vision-language backbone. Built on a vision-language model with native video understanding, ActMem infers the current subtask from a timestamped interaction history and keeps the hidden states formed during this inference as a Latent Memory Chain-of-Thought (Memory CoT), which conditions a flow-matching action expert. Perception Attention further lets the expert attend directly to the current view. To keep inference fast under long histories, Efficient Memory CoT Inference reuses computation across consecutive decisions without changing the output. ActMem achieves 71.1% task success and 80.8% completion success on RoboMemArena, surpassing the strongest baseline by 7.5 and 8.7 percentage points, and 73.8% average success on MIKASA-Robo, the best among all compared methods. In the real world, it reaches 68.8% average success on exogenous, endogenous, and traceless memory tasks, 10.0 percentage points above the strongest memory-augmented VLA. Code and checkpoints will be released at https://github.com/LigphiDonk/ActMem and https://huggingface.co/donk111/ActMem.

## LIFT: Latent Flow Matching for Probabilistic Cell Fate Dynamics with Single Cell Snapshots

DevFeed: [LIFT: Latent Flow Matching for Probabilistic Cell Fate Dynamics with Single Cell Snapshots](<https://devfeed.tech/articles/lift-latent-flow-matching-for-probabilistic-cell-fate-dynamics-with-single-cell-snapshots-145117.md>)

Original publisher: [Read original article](<https://arxiv.org/abs/2611.25917>)

Author: Bowen Yang, Junda Ying, Yuxuan Wang, Peijie Zhou, Lei Zhang

Published: 2026-11-20T05:00:00Z

Content type: article

Language: en

Sources: [arXiv: Machine Learning](<https://devfeed.tech/sources/arxiv-machine-learning.md>)

Topics: [VAE](<https://devfeed.tech/topics/vae.md>), [Predictive AI](<https://devfeed.tech/topics/predictive-ai.md>)

Tags: [analysis](<https://devfeed.tech/tags/analysis.md>), [cs-lg](<https://devfeed.tech/tags/cs-lg.md>), [datasets](<https://devfeed.tech/tags/datasets.md>), [decoder](<https://devfeed.tech/tags/decoder.md>), [distribution](<https://devfeed.tech/tags/distribution.md>), [error](<https://devfeed.tech/tags/error.md>), [experiments](<https://devfeed.tech/tags/experiments.md>), [fate](<https://devfeed.tech/tags/fate.md>), [flow](<https://devfeed.tech/tags/flow.md>), [flow-matching](<https://devfeed.tech/tags/flow-matching.md>), [framework](<https://devfeed.tech/tags/framework.md>), [future](<https://devfeed.tech/tags/future.md>), [generation](<https://devfeed.tech/tags/generation.md>), [inference](<https://devfeed.tech/tags/inference.md>), [learning](<https://devfeed.tech/tags/learning.md>), [monte-carlo](<https://devfeed.tech/tags/monte-carlo.md>), [one-to-many](<https://devfeed.tech/tags/one-to-many.md>), [prediction](<https://devfeed.tech/tags/prediction.md>), [probabilistic](<https://devfeed.tech/tags/probabilistic.md>), [sampling](<https://devfeed.tech/tags/sampling.md>), [single-cell](<https://devfeed.tech/tags/single-cell.md>)

### AI overview

LIFT models probabilistic cell fates from lineage-informed couplings, which connect cells through clonal ancestry without identifying unique source-to-descendant pairs. Variational inference represents alternative outcomes in a latent space, and a conditional prior allows sampling using only an initial cell state. Its Mean Flow decoder predicts the future state in a single evaluation per sample, supporting large-scale Monte Carlo prediction. The paper also analyzes conditional prediction error and reports experiments on synthetic and real datasets that recover population distributions and cell-fate probabilities.

### Source excerpt

arXiv:2611.25917v1 Announce Type: new Abstract: Lineage tracing links cell states through clonal ancestry, but does not generally identify unique source-descendant pairings. Learning from such data therefore requires a model of conditional outcomes, rather than only the terminal population distribution. We introduce LIFT, a latent flow framework for probabilistic cell-state generation from lineage-informed couplings. Variational inference encodes alternative outcomes in a latent space, while a conditional prior enables sampling from the initial state alone. A Mean Flow decoder predicts future cell state directly, requiring one decoder evaluation per sampling, which makes it convenient to predict cell fate through large scale Monte-Carlo sampling. This construction combines a representation of one-to-many relationships with efficient sampling and an analysis of conditional prediction error. Experiments on synthetic and real datasets demonstrate LIFT's ability to recover both population distribution and cell fate probabilities.

## One Latent, Many Tokens: Jointly Learning Compressed Embeddings for Efficient Language Diffusion

DevFeed: [One Latent, Many Tokens: Jointly Learning Compressed Embeddings for Efficient Language Diffusion](<https://devfeed.tech/articles/one-latent-many-tokens-jointly-learning-compressed-embeddings-for-efficient-language-diffusion-145381.md>)

Original publisher: [Read original article](<https://arxiv.org/abs/2609.33698>)

Author: Yulin Yuan, Ying Zhang, Xiangming Meng

Published: 2026-11-20T05:00:00Z

Content type: article

Language: en

Sources: [arXiv: Artificial Intelligence](<https://devfeed.tech/sources/arxiv-artificial-intelligence.md>)

Topics: [Embeddings](<https://devfeed.tech/topics/embeddings.md>), [Representation learning](<https://devfeed.tech/topics/representation-learning.md>), [Large Language Model](<https://devfeed.tech/topics/llm.md>), [LLM optimization](<https://devfeed.tech/topics/llm-optimization.md>)

Tags: [compression](<https://devfeed.tech/tags/compression.md>), [cs-ai](<https://devfeed.tech/tags/cs-ai.md>), [diffusion](<https://devfeed.tech/tags/diffusion.md>), [embeddings](<https://devfeed.tech/tags/embeddings.md>), [flow](<https://devfeed.tech/tags/flow.md>), [flow-matching](<https://devfeed.tech/tags/flow-matching.md>), [generation](<https://devfeed.tech/tags/generation.md>), [language](<https://devfeed.tech/tags/language.md>), [language-model](<https://devfeed.tech/tags/language-model.md>), [language-models](<https://devfeed.tech/tags/language-models.md>), [learning](<https://devfeed.tech/tags/learning.md>), [modeling](<https://devfeed.tech/tags/modeling.md>), [structured](<https://devfeed.tech/tags/structured.md>), [throughput](<https://devfeed.tech/tags/throughput.md>), [tokens](<https://devfeed.tech/tags/tokens.md>)

### AI overview

JPEG-DLM reduces diffusion language generation cost by jointly training a compressor, flow matching model, and decoder, so compressed embeddings are designed to be easier to model and decode. This contrasts with two-stage methods that fix the compressed embedding space before training the diffusion model, which the abstract says can hinder modeling and token recovery. The authors report the highest throughput among recent diffusion and flow models on LM1B and OWT, with Gen-PPL lower than most. On OWT at a 0.5 compression rate, JPEG-DLM reports Gen-PPL 34.52 and about 2.3x ELF's throughput.

### Source excerpt

arXiv:2609.33698v2 Announce Type: replace Abstract: Most continuous diffusion language models process one latent position per token at each sampling step, making generation expensive. Two-stage methods lower the cost by reducing the latent length, but they fix the compressed embedding space before training the diffusion model. Embeddings from the fixed space can be difficult to model with diffusion and decode reliably into tokens, which limits generation quality after compression. To address this problem, we introduce JPEG-DLM (Joint-embedding Prediction for Efficient Generation with Diffusion Language Model), which jointly trains a compressor, a flow matching model and a decoding module. With joint-embedding prediction, JPEG-DLM learns compressed embeddings that are more structured, easier to model with diffusion and reliably decodable into tokens.JPEG-DLM achieves the highest throughput among recent diffusion and flow models on LM1B and OWT, with a Gen-PPL lower than most of them. At a compression rate of 0.5 on OWT, it reaches a Gen-PPL of 34.52 and approximately 2.3 times ELF's throughput. These results suggest that jointly learning compressed embeddings offers a promising path toward efficient diffusion language modeling.Code is available at https://github.com/infiniteYuanyl/JPEG-DLM.

## Friction-Augmented Drifting Models for Resource-Efficient Domain Translation

DevFeed: [Friction-Augmented Drifting Models for Resource-Efficient Domain Translation](<https://devfeed.tech/articles/friction-augmented-drifting-models-for-resource-efficient-domain-translation-144887.md>)

Original publisher: [Read original article](<https://arxiv.org/abs/2604.18194>)

Author: Arkadii Kazanskii, Tatiana Petrova, Andrey Ustyuzhanin, Konstantin Bagrianskii, Aleksandr Puzikov, Radu State

Published: 2026-11-20T05:00:00Z

Content type: article

Language: en

Sources: [arXiv: Machine Learning](<https://devfeed.tech/sources/arxiv-machine-learning.md>)

Topics: [Transformers](<https://devfeed.tech/topics/transformers.md>)

Tags: [cost](<https://devfeed.tech/tags/cost.md>), [cs-cv](<https://devfeed.tech/tags/cs-cv.md>), [cs-lg](<https://devfeed.tech/tags/cs-lg.md>), [domain](<https://devfeed.tech/tags/domain.md>), [drift](<https://devfeed.tech/tags/drift.md>), [flow-matching](<https://devfeed.tech/tags/flow-matching.md>), [generator](<https://devfeed.tech/tags/generator.md>), [gradient-descent](<https://devfeed.tech/tags/gradient-descent.md>), [learning](<https://devfeed.tech/tags/learning.md>), [machine-learning](<https://devfeed.tech/tags/machine-learning.md>), [models](<https://devfeed.tech/tags/models.md>), [translation](<https://devfeed.tech/tags/translation.md>)

### AI overview

DMF scales a Drifting Model's kernel-drift displacement linearly from full drift to zero. Under plain gradient descent, this matches linear learning-rate decay; with Adam, it also changes gradient moments, and a constant multiplier c has the exact-arithmetic effect of replacing ε with ε/c. In a two-particle model with unnormalised Laplace interactions, the authors derive where attraction balances repulsion. For FFHQ adult-to-child translation, DMF had lower FID and CMMD than fixed-rate DM in all five seed pairs. A separate comparison favored linear learning-rate decay on mean FID and CMMD, though both 95% intervals for paired differences included zero. At batch size 32 on one RTX A6000, mean training times were 6.85 minutes for DMF, 6.93 for DM, and 133.48 for OFM; DMF also had lower FID and CMMD than OFM, with paired p-values 0.002 and 0.021.

### Source excerpt

arXiv:2604.18194v3 Announce Type: replace Abstract: Drifting Models (DMs) train a one-step generator by regressing its output towards a target displaced by a kernel drift. DMF (Drifting Model with Friction) scales this displacement linearly from full drift to zero. With plain gradient descent, scaling the drift gives the same updates as linear learning-rate decay. With Adam, it also changes the gradient moments. For a constant multiplier $c>0$, its effect in exact arithmetic is to replace Adam's $\epsilon$ by $\epsilon/c$. In a two-particle model with unnormalised Laplace interactions, we derive the distance at which attraction balances repulsion. On FFHQ adult-to-child translation, DMF lowered FID and CMMD relative to fixed-rate DM in all five seed pairs. In a separate paired comparison, mean FID and CMMD were lower for DM with linear learning-rate decay than for DMF. Both 95% intervals for the paired differences included zero. DMF inherits DM's low training cost. At batch size 32 on one RTX A6000, mean training times were 6.85 minutes for DMF, 6.93 for DM and 133.48 for Optimal Flow Matching (OFM). At this batch size, DMF also achieved lower FID and CMMD than OFM (paired $p=0.002$ and $0.021$, respectively).

## VG-Steer: Visual-Geometric Policy Steering from In-Context Aligned Abstractions

DevFeed: [VG-Steer: Visual-Geometric Policy Steering from In-Context Aligned Abstractions](<https://devfeed.tech/articles/vg-steer-visual-geometric-policy-steering-from-in-context-aligned-abstractions-145481.md>)

Original publisher: [Read original article](<https://arxiv.org/abs/2611.25199>)

Author: Arthur Bucker, Pramod Anantharam, Jonathan Francis, Jean Oh

Published: 2026-11-20T05:00:00Z

Content type: article

Language: en

Sources: [arXiv: Artificial Intelligence](<https://devfeed.tech/sources/arxiv-artificial-intelligence.md>)

Topics: [Robot Manipulation](<https://devfeed.tech/topics/robot-manipulation.md>), [llm-reasoning](<https://devfeed.tech/topics/llm-reasoning.md>)

Tags: [abstractions](<https://devfeed.tech/tags/abstractions.md>), [adapting](<https://devfeed.tech/tags/adapting.md>), [benchmark](<https://devfeed.tech/tags/benchmark.md>), [bottleneck](<https://devfeed.tech/tags/bottleneck.md>), [cs-ai](<https://devfeed.tech/tags/cs-ai.md>), [cs-ro](<https://devfeed.tech/tags/cs-ro.md>), [data](<https://devfeed.tech/tags/data.md>), [datasets](<https://devfeed.tech/tags/datasets.md>), [distribution](<https://devfeed.tech/tags/distribution.md>), [efficiently](<https://devfeed.tech/tags/efficiently.md>), [flow-matching](<https://devfeed.tech/tags/flow-matching.md>), [inference](<https://devfeed.tech/tags/inference.md>), [manipulation](<https://devfeed.tech/tags/manipulation.md>), [ood](<https://devfeed.tech/tags/ood.md>), [pairs](<https://devfeed.tech/tags/pairs.md>), [policies](<https://devfeed.tech/tags/policies.md>), [policy](<https://devfeed.tech/tags/policy.md>), [real-world](<https://devfeed.tech/tags/real-world.md>), [reasoning](<https://devfeed.tech/tags/reasoning.md>), [region](<https://devfeed.tech/tags/region.md>), [robot](<https://devfeed.tech/tags/robot.md>), [steering](<https://devfeed.tech/tags/steering.md>)

### AI overview

VG-Steer adapts frozen robot policies at inference by transferring motion patterns from successful in-distribution rollouts. It represents those patterns as object-relational distance pairs, then combines task-relevant entities and geometric projections using Large Vision Model semantics and Vision-Language Model reasoning to produce visual-geometric guidance. The abstract reports efficient correction of flow-matching policies under pose and goal shifts in simulation. In a real-robot experiment, it more than doubled success on cable placements one grid step beyond training positions, expanded workspace coverage, and steered a single-task policy toward unseen objects while nearly eliminating object confusion. These are abstract-level claims; details of the experiments and limitations are not provided.

### Source excerpt

arXiv:2611.25199v1 Announce Type: cross Abstract: Adapting robot policies to out-of-distribution (OOD) settings with limited real-world data remains a key bottleneck to generalist robots. The multimodality and scarcity of robotics data lead end-to-end approaches to overfit and break under small distribution shifts. Test-time adaptation can alleviate this by correcting frozen policies at inference, but most methods ground their correction in a signal learned implicitly from robotic datasets, making it susceptible to the same shifts as the policy it corrects. Recent methods that instead exploit explicit spatial information are typically limited to coarse corrections, missing the nuanced, task-relevant motion most manipulation tasks require. We propose VG-Steer, which corrects existing policies by transferring the nuanced motion of successful in-distribution rollouts, captured along meaningful geometric axes, known here as object-relational distance pairs. VG-Steer leverages the semantic understanding of Large Vision Models (LVMs) and Vision-Language Models (VLMs) reasoning, which combines the task-relevant entity pairs and their object-relational distance projections into a single tailored visual-geometric guidance signal. In a simulation benchmark, it efficiently corrects flow-matching policies under pose and task-goal shifts; on a real robot experiment, it widens the region of the workspace the frozen policy covers, more than doubling its success rate on cable placements one grid step outside the training positions, and steers a single-task policy toward objects it was never trained on while nearly eliminating object confusion. Project website: https://arthurfenderbucker.github.io/VG-Steer/

## Mamba Flow Matching Neural Processes: Linear-Time Inference for Irregularly Observed Spatial Fields

DevFeed: [Mamba Flow Matching Neural Processes: Linear-Time Inference for Irregularly Observed Spatial Fields](<https://devfeed.tech/articles/mamba-flow-matching-neural-processes-linear-time-inference-for-irregularly-observed-spatial-fields-145168.md>)

Original publisher: [Read original article](<https://arxiv.org/abs/2611.26355>)

Author: Cosmo Santoni, Giovanni Charles, Timothy James Hitge, Oliver Watson, Elizaveta Semenova

Published: 2026-11-20T05:00:00Z

Content type: article

Language: en

Sources: [arXiv: Machine Learning](<https://devfeed.tech/sources/arxiv-machine-learning.md>)

Topics: [Mamba](<https://devfeed.tech/topics/mamba.md>), [Neural Network](<https://devfeed.tech/topics/neural-network.md>), [Deep neural networks](<https://devfeed.tech/topics/deep-neural-networks.md>)

Tags: [2](<https://devfeed.tech/tags/2.md>), [3](<https://devfeed.tech/tags/3.md>), [abstract](<https://devfeed.tech/tags/abstract.md>), [arxiv](<https://devfeed.tech/tags/arxiv.md>), [attention](<https://devfeed.tech/tags/attention.md>), [based](<https://devfeed.tech/tags/based.md>), [benchmarks](<https://devfeed.tech/tags/benchmarks.md>), [but](<https://devfeed.tech/tags/but.md>), [combined](<https://devfeed.tech/tags/combined.md>), [cs-lg](<https://devfeed.tech/tags/cs-lg.md>), [fields](<https://devfeed.tech/tags/fields.md>), [flow](<https://devfeed.tech/tags/flow.md>), [flow-matching](<https://devfeed.tech/tags/flow-matching.md>), [inference](<https://devfeed.tech/tags/inference.md>), [linear](<https://devfeed.tech/tags/linear.md>), [mamba](<https://devfeed.tech/tags/mamba.md>), [matching](<https://devfeed.tech/tags/matching.md>), [neural](<https://devfeed.tech/tags/neural.md>), [processes](<https://devfeed.tech/tags/processes.md>), [spatial](<https://devfeed.tech/tags/spatial.md>), [time](<https://devfeed.tech/tags/time.md>)

### AI overview

MambaFlowNP uses conditional flow matching and a multi-directional Mamba-2 scan to jointly predict irregularly observed spatial fields with linear compute. Context and query tokens share a sequence, and each query reads the scan state at its own position; this avoids reducing context to one hidden summary and makes predictions consistent across query sets. On synthetic Matérn Gaussian-process benchmarks it was competitive with leading attention-based joint-predictive baselines. For NOAA temperature interpolation, it achieved the lowest reported RMSE and CRPS, near-nominal coverage across tiers, and ran 114-784 times faster than the strongest joint-predictive baseline. A forward pass handled one million points in under a second on an NVIDIA L40S, though predictive quality was evaluated only up to 5,580 points.

### Source excerpt

arXiv:2611.26355v1 Announce Type: new Abstract: Interpolating a spatial field, such as daily maximum temperature, from sparse, irregularly located observations recurs across the environmental and health sciences. Standard Gaussian-process kriging refits hyperparameters and solves an $O(N^3)$ system for every new field. When many related fields are observed, such as daily maps from one sensor network, neural processes (NPs) instead learn the interpolation map from past fields and predict a new field in one forward pass. Joint-predictive variants, which sample whole fields rather than independent per-location values, typically use self-attention over context and query points, costing $O(K^2)$ in the combined point count $K$ and limiting practical use to a few thousand points. State-space models offer a linear-time alternative, but a direct state-space adaptation compresses the context into one hidden state before reading queries, reducing what the decoder sees. We propose MambaFlowNP, a linear-compute joint-predictive NP trained with conditional flow matching. A multi-directional Mamba-2 scan reads context and query tokens in one shared sequence, so each query's velocity is read from its own scan position rather than a compressed context summary, and because queries only read the scan state, predictions are consistent across query sets. MambaFlowNP is competitive with the strongest attention-based joint-predictive baselines on synthetic Mat\'ern Gaussian-process benchmarks, and on regional to continental-scale NOAA temperature interpolation it attains the lowest RMSE and CRPS and near-nominal coverage on every tier, while running 114--784$\times$ faster than the strongest joint-predictive baseline. A single forward pass scales to $10^6$ points on one NVIDIA L40S GPU in under a second. This makes joint prediction feasible for continental-scale, irregularly sampled sensor networks; predictive quality is evaluated up to $K=5{,}580$.

## ReliFlow: Reliability-Calibrated Residual Transport for Combinatorial Single-Cell Perturbation Prediction

DevFeed: [ReliFlow: Reliability-Calibrated Residual Transport for Combinatorial Single-Cell Perturbation Prediction](<https://devfeed.tech/articles/reliflow-reliability-calibrated-residual-transport-for-combinatorial-single-cell-perturbation-prediction-142735.md>)

Original publisher: [Read original article](<https://arxiv.org/abs/2611.23999>)

Author: Jinrui David, Yun Peng

Published: 2026-11-19T05:00:00Z

Content type: article

Language: en

Sources: [arXiv: Artificial Intelligence](<https://devfeed.tech/sources/arxiv-artificial-intelligence.md>), [arXiv: Machine Learning](<https://devfeed.tech/sources/arxiv-machine-learning.md>)

Topics: [Advanced AI workflows](<https://devfeed.tech/topics/advanced-ai-workflows.md>), [Tensorflow](<https://devfeed.tech/topics/tensorflow.md>)

Tags: [combined](<https://devfeed.tech/tags/combined.md>), [component](<https://devfeed.tech/tags/component.md>), [conditional](<https://devfeed.tech/tags/conditional.md>), [correlation](<https://devfeed.tech/tags/correlation.md>), [cost](<https://devfeed.tech/tags/cost.md>), [cross](<https://devfeed.tech/tags/cross.md>), [cs-ai](<https://devfeed.tech/tags/cs-ai.md>), [cs-lg](<https://devfeed.tech/tags/cs-lg.md>), [distribution](<https://devfeed.tech/tags/distribution.md>), [effect](<https://devfeed.tech/tags/effect.md>), [error](<https://devfeed.tech/tags/error.md>), [field](<https://devfeed.tech/tags/field.md>), [flow](<https://devfeed.tech/tags/flow.md>), [flow-matching](<https://devfeed.tech/tags/flow-matching.md>), [matching](<https://devfeed.tech/tags/matching.md>), [model](<https://devfeed.tech/tags/model.md>), [prediction](<https://devfeed.tech/tags/prediction.md>), [reference](<https://devfeed.tech/tags/reference.md>), [reliability](<https://devfeed.tech/tags/reliability.md>), [residual](<https://devfeed.tech/tags/residual.md>), [single-cell](<https://devfeed.tech/tags/single-cell.md>), [stat-ml](<https://devfeed.tech/tags/stat-ml.md>), [transport](<https://devfeed.tech/tags/transport.md>)

### AI overview

ReliFlow separates errors in estimated single-perturbation effects from genuine interactions by estimating effect reliability through training-only cross-fitting and using it to calibrate an additive source distribution. Its conditional flow-matching velocity combines a known correction to that reference with a learned residual transport field. In the reported evaluations, MAE fell from 0.142 to 0.131 for unseen combinations of observed Norman genetic perturbations, from 0.198 to 0.178 when constituent singles were withheld, and from 0.191 to 0.176 on held-out ComboSciPlex drug combinations. Non-additive residual correlation rose from 0.574 to 0.638; ablations indicate reliability weighting contributes to performance. The authors report a moderate increase in training cost.

### Source excerpt

arXiv:2611.23999v1 Announce Type: cross Abstract: Predicting the transcriptional response to combined perturbations requires distinguishing interactions between interventions from errors in their individual effect estimates. The latter become more pronounced when single-perturbation evidence is sparse or withheld. We introduce a conditional flow-matching model that estimates the reliability of single-gene effects from training-only cross-fitting and uses those estimates to calibrate the additive source distribution. The velocity is decomposed into a known correction of the additive reference and a learned residual transport field. On Norman genetic perturbations, reduces mean absolute error (MAE) from 0.142 to 0.131 for unseen combinations of observed singles and from 0.198 to 0.178 when constituent singles are withheld. On held-out ComboSciPlex drug combinations, MAE decreases from 0.191 to 0.176. Non-additive residual correlation improves from 0.574 to 0.638, and component ablations identify the contribution of reliability weighting. The results indicate that accounting for uncertain constituent effects improves generalization to unseen genetic and chemical combinations, with a moderate increase in training cost.

## Horizontal Flow Matching for Single-Cell Spatiotemporal Dynamics

DevFeed: [Horizontal Flow Matching for Single-Cell Spatiotemporal Dynamics](<https://devfeed.tech/articles/horizontal-flow-matching-for-single-cell-spatiotemporal-dynamics-142828.md>)

Original publisher: [Read original article](<https://arxiv.org/abs/2611.24764>)

Author: Zihan Wang, Jianzhe Li, Zhenyi Zhang, Peijie Zhou

Published: 2026-11-19T05:00:00Z

Content type: article

Language: en

Sources: [arXiv: Artificial Intelligence](<https://devfeed.tech/sources/arxiv-artificial-intelligence.md>), [arXiv: Machine Learning](<https://devfeed.tech/sources/arxiv-machine-learning.md>)

Topics: [Neural Network](<https://devfeed.tech/topics/neural-network.md>)

Tags: [alignment](<https://devfeed.tech/tags/alignment.md>), [cs-ai](<https://devfeed.tech/tags/cs-ai.md>), [cs-lg](<https://devfeed.tech/tags/cs-lg.md>), [drift](<https://devfeed.tech/tags/drift.md>), [expression](<https://devfeed.tech/tags/expression.md>), [flow](<https://devfeed.tech/tags/flow.md>), [flow-matching](<https://devfeed.tech/tags/flow-matching.md>), [machine-learning](<https://devfeed.tech/tags/machine-learning.md>), [q-bio-qm](<https://devfeed.tech/tags/q-bio-qm.md>), [single-cell](<https://devfeed.tech/tags/single-cell.md>), [snapshot](<https://devfeed.tech/tags/snapshot.md>), [spatial](<https://devfeed.tech/tags/spatial.md>), [time](<https://devfeed.tech/tags/time.md>), [transformer](<https://devfeed.tech/tags/transformer.md>)

### AI overview

HORIZON addresses a key issue in learning cell dynamics from spatial transcriptomics: section rotations or translations can be mistaken for cell movement. Its horizontal flow matching models tissue organization, molecular state, and cell abundance while assigning no cost to shared rigid motion; a horizontal projector removes that motion from learning errors and generated velocities. The method also jointly estimates cell correspondence and relative pose in Wasserstein-Fisher-Rao geometry, then trains an equivariant transformer to predict tissue velocity and growth from one initial snapshot. In reported evaluations, alignment improved reconstruction and projection reduced rigid drift on simulated perturbed tissues. Across four time courses, HORIZON reportedly reconstructed geometry, gene expression, and cell abundance more accurately than existing spatial and unbalanced trajectory methods, and interpolated held-out developmental stages.

### Source excerpt

arXiv:2611.24764v1 Announce Type: cross Abstract: Learning the spatiotemporal dynamics of cells from spatial transcriptomics requires connecting snapshots of different cells acquired in different coordinate frames. A rotation or translation of a section can appear as cell movement and distort the inferred correspondence. We introduce horizontal flow matching, which learns changes in tissue organization, molecular state, and cell abundance up to shared rigid motion. The dynamics follow a continuity--reaction equation whose action assigns no cost to shared rigid motion, and the resulting horizontal projector removes this component from spatial prediction errors during learning and from predicted velocities during generation. An equivalent endpoint problem jointly estimates correspondence and relative pose in Wasserstein--Fisher--Rao geometry and yields analytic paths whose velocity and growth serve as regression targets. Regressing these targets recovers the horizontal dynamics under ideal population conditions. Our model, HORIZON, uses an equivariant transformer to predict velocity and growth from the current tissue, so that a tissue can be generated continuously from a single initial snapshot. On a simulated tissue observed in perturbed coordinate frames, alignment improves reconstruction and projection reduces rigid drift. On four spatial transcriptomics time courses, HORIZON reconstructs tissue geometry, gene expression, and cell abundance more accurately than existing spatial and unbalanced trajectory methods, and interpolates held-out developmental stages.

## Estimating aerosol particle number size distributions from PM2.5 with physics-regularized flow matching

DevFeed: [Estimating aerosol particle number size distributions from PM2.5 with physics-regularized flow matching](<https://devfeed.tech/articles/estimating-aerosol-particle-number-size-distributions-from-pm2-5-with-physics-regularized-flow-matching-141629.md>)

Original publisher: [Read original article](<https://arxiv.org/abs/2610.33557>)

Author: Zhi-Song Liu, Jenni K\\"oykk\\"a, Wenqing Peng, Ammar Kheder, Petri Clusius, Michael Boy

Published: 2026-11-18T05:00:00Z

Content type: article

Language: en

Sources: [arXiv: Machine Learning](<https://devfeed.tech/sources/arxiv-machine-learning.md>)

Topics: [Machine learning](<https://devfeed.tech/topics/machine-learning.md>), [Neural Network](<https://devfeed.tech/topics/neural-network.md>)

Tags: [air-quality-monitoring](<https://devfeed.tech/tags/air-quality-monitoring.md>), [benchmark](<https://devfeed.tech/tags/benchmark.md>), [constraint](<https://devfeed.tech/tags/constraint.md>), [cs-lg](<https://devfeed.tech/tags/cs-lg.md>), [deep-learning](<https://devfeed.tech/tags/deep-learning.md>), [flow-matching](<https://devfeed.tech/tags/flow-matching.md>), [gde](<https://devfeed.tech/tags/gde.md>), [generative](<https://devfeed.tech/tags/generative.md>), [generative-models](<https://devfeed.tech/tags/generative-models.md>), [physics](<https://devfeed.tech/tags/physics.md>), [pm2-5](<https://devfeed.tech/tags/pm2-5.md>), [regression](<https://devfeed.tech/tags/regression.md>)

### AI overview

The study estimates hourly particle number size distributions (PNSDs) from PM2.5 records and eight meteorological variables over 24-hour windows, generating ensembles because the same mass concentration can correspond to different size distributions. At two contrasting Finnish sites, Hyytiala and Kumpula, six generative models were compared with regression and historical-mean baselines, including tests of transfer between sites. Flow matching performed best overall, and its guided variant had the lowest CRPS at both sites. A reduced aerosol general dynamic equation constraint lowered CRPS by 5.5-11.5% at the training site and reduced transfer errors. Cross-site transfer remained harder than local estimation; the abstract identifies multi-station training as a next step toward estimates at unmonitored locations.

### Source excerpt

arXiv:2610.33557v2 Announce Type: replace Abstract: The mass concentration of particles smaller than 2.5um (PM2.5) is cheap to measure and widely monitored, whereas the particle number size distribution (PNSD) requires costly instruments and is measured at only a few research stations. Yet many aerosol effects, such as vapor uptake, cloud droplet formation and deposition in the human respiratory tract, depend on the PNSD rather than on mass. We explore whether generative deep learning models can infer the current PNSD from routine data: from the PM2.5 record and eight meteorological variables over a 24h window, a model generates an ensemble of plausible hourly PNSDs for the same window, reflecting that one mass can arise from many size distributions. Rather than proposing a single best model, we use two contrasting Finnish stations, the boreal forest site Hyytiala and the urban site Kumpula, as a controlled testbed. On it, we benchmark six generative models against deterministic regression and historical-mean baselines, locally and after transfer between sites, and test physical constraints for neural network training. Flow matching performs best overall; its guided variant has the lowest continuous ranked probability score (CRPS) at both sites. A constraint based on a reduced aerosol general dynamic equation (GDE) lowers the CRPS by 5.5-11.5% at the training site and reduces errors after transfer between sites. Transfer remains harder than local estimation, identifying multi-station training as the next step towards PNSD estimates at unmonitored locations. This benchmark establishes physics-informed generative models as a promising route from routine air-quality monitoring to size-resolved aerosol information.

## Task Convergent Flow Matching for Generative Robotics

DevFeed: [Task Convergent Flow Matching for Generative Robotics](<https://devfeed.tech/articles/task-convergent-flow-matching-for-generative-robotics-141794.md>)

Original publisher: [Read original article](<https://arxiv.org/abs/2611.22609>)

Author: Lorenzo Paiola, Maximilian St\\"olzle, Giorgio Grioli, Daniela Rus, Cosimo Della Santina

Published: 2026-11-18T05:00:00Z

Content type: news

Language: en

Sources: [arXiv: Machine Learning](<https://devfeed.tech/sources/arxiv-machine-learning.md>)

Topics: [Robot Manipulation](<https://devfeed.tech/topics/robot-manipulation.md>), [contact-rich manipulation](<https://devfeed.tech/topics/contact-rich-manipulation.md>), [generative-ai-tools](<https://devfeed.tech/topics/generative-ai-tools.md>)

Tags: [continuous](<https://devfeed.tech/tags/continuous.md>), [convergence](<https://devfeed.tech/tags/convergence.md>), [covariance](<https://devfeed.tech/tags/covariance.md>), [cs-lg](<https://devfeed.tech/tags/cs-lg.md>), [cs-ro](<https://devfeed.tech/tags/cs-ro.md>), [distribution](<https://devfeed.tech/tags/distribution.md>), [flow-matching](<https://devfeed.tech/tags/flow-matching.md>), [funnel](<https://devfeed.tech/tags/funnel.md>), [generative](<https://devfeed.tech/tags/generative.md>), [hybrid](<https://devfeed.tech/tags/hybrid.md>), [inference](<https://devfeed.tech/tags/inference.md>), [manipulation](<https://devfeed.tech/tags/manipulation.md>), [motion](<https://devfeed.tech/tags/motion.md>), [parallel](<https://devfeed.tech/tags/parallel.md>), [performance](<https://devfeed.tech/tags/performance.md>), [physics](<https://devfeed.tech/tags/physics.md>), [progress](<https://devfeed.tech/tags/progress.md>), [quantile](<https://devfeed.tech/tags/quantile.md>), [robotics](<https://devfeed.tech/tags/robotics.md>), [search](<https://devfeed.tech/tags/search.md>), [source](<https://devfeed.tech/tags/source.md>), [stacking](<https://devfeed.tech/tags/stacking.md>)

### AI overview

The framework refines flow-matching robot policies at inference time by testing candidate action chunks in parallel physics-based rollouts and searching for chunks that make task progress. Performance funnels allow bounded temporary regressions but require net progress; interval-wise Lipschitz bounds are used to verify continuous-time funnel satisfaction from sampled rollouts. For contact-rich dynamics, CMA-ES searches source seeds within radial quantile bands to help retain the learned motion prior. The method supports Euclidean and pose-valued Riemannian Flow Matching policies. Reported success rises from 31.3% to 88.5% on PushT and from 77.0% to 86.0% on three-cube stacking. On PushT, a learned funnel reduces average maximum loss of achieved progress by 19% versus checking chunk endpoints alone, at similar success. Radial-band search yields 11.3% out-of-distribution motion windows, compared with 34.1% for unconstrained search.

### Source excerpt

arXiv:2611.22609v1 Announce Type: cross Abstract: Successful manipulation requires the manipulated objects to progress toward a desired task outcome. Generative policies can produce plausible robot motions without ensuring this progress, allowing successive action chunks to stall or undo earlier changes to the scene. We introduce a task-centric inference-time refinement framework for flow matching policies that evaluates candidate action chunks through parallel physics-based rollouts and searches for motions satisfying sufficient conditions for task convergence. Task-level performance funnels bound temporary regressions while requiring net progress over each chunk. We establish sufficient conditions under which repeated funnel satisfaction yields convergence throughout execution, and use interval-wise Lipschitz bounds to verify continuous-time funnel satisfaction from sampled rollouts. To accommodate contact-rich hybrid dynamics, we refine the flow's source seeds using derivative-free Covariance Matrix Adaptation Evolution Strategy (CMA-ES), constrained to radial quantile bands of the source distribution to help retain the learned motion prior. This formulation supports both Euclidean and pose-valued Riemannian Flow Matching (RMFM) policies. On PushT and three-cube stacking, refinement raises success from 31.3% to 88.5% and from 77.0% to 86.0%. On PushT, whose coverage score responds continuously to the motion of the block, a learned funnel lowers the average maximum loss of achieved progress by 19% relative to constraining the endpoint of each chunk alone, at a similar success rate. With the radial band, 11.3% of the executed motion windows are out of distribution, a rate within the range of held-out demonstrations, compared with 34.1% for an unconstrained search.

## Sub-Meter Canopy Height Synthesis from Sentinel-2 via Anchored Residual Flow

DevFeed: [Sub-Meter Canopy Height Synthesis from Sentinel-2 via Anchored Residual Flow](<https://devfeed.tech/articles/sub-meter-canopy-height-synthesis-from-sentinel-2-via-anchored-residual-flow-141797.md>)

Original publisher: [Read original article](<https://arxiv.org/abs/2611.22636>)

Author: Kiarie Ndegwa, Tony Chang, Andreas Gros, Vincent A. Landau, David D. Diaz, Luke Zachmann, Nathan E. Rutenbeck, Guy C Bayes

Published: 2026-11-18T05:00:00Z

Content type: article

Language: en

Sources: [arXiv: Machine Learning](<https://devfeed.tech/sources/arxiv-machine-learning.md>)

Topics: [Machine learning](<https://devfeed.tech/topics/machine-learning.md>)

Tags: [cs-cv](<https://devfeed.tech/tags/cs-cv.md>), [cs-lg](<https://devfeed.tech/tags/cs-lg.md>), [flow](<https://devfeed.tech/tags/flow.md>), [flow-matching](<https://devfeed.tech/tags/flow-matching.md>), [lidar](<https://devfeed.tech/tags/lidar.md>), [mapping](<https://devfeed.tech/tags/mapping.md>), [residual](<https://devfeed.tech/tags/residual.md>), [satellite](<https://devfeed.tech/tags/satellite.md>), [statistics](<https://devfeed.tech/tags/statistics.md>), [super-resolution](<https://devfeed.tech/tags/super-resolution.md>), [synthesis](<https://devfeed.tech/tags/synthesis.md>), [trees](<https://devfeed.tech/tags/trees.md>)

### AI overview

The study synthesizes 0.5 m canopy height models from free 10 m Sentinel-2 imagery, a 20x cross-modal super-resolution task because reflectance does not directly measure height. On 2,530 held-out tiles, the deterministic Anchor achieves 2.83 m MAE. AR-flow, trained to model Anchor residuals, lowers sliced Wasserstein distance from 1.13 to 0.208, but has higher MAE (3.68 m), reflecting a perception-distortion trade-off. Crown-scale statistics improve, while individual-tree detection remains weak: sub-pixel detail is plausible structure, not measured trees. The authors say quarterly continental refreshes could help guide when and where to acquire more expensive data.

### Source excerpt

arXiv:2611.22636v1 Announce Type: cross Abstract: Canopy height models (CHMs) - per-pixel maps of vegetation top height - are essential for forest carbon accounting, wildfire fuel mapping, and tree-level ecology, yet sub-meter CHMs require airborne LiDAR refreshed on multi-year national cycles. We produce 0.5 m CHMs from free 10 m Sentinel-2 imagery, a 20$\times$ cross-modal super-resolution task where optical reflectance must predict geometric height it does not directly observe. Motivated by the perception-distortion trade-off, we derive two operating points with shared conditioning. A deterministic regressor (Anchor) targets low distortion; a conditional flow matching model, trained on the Anchor's residual in a learned CHM latent space (AR-flow), targets perceptual fidelity. Trained on 197k LiDAR-paired tiles spanning 69 ecoregions across the contiguous United States and evaluated on 2,530 held-out tiles against eight baselines, the Anchor achieves 2.83 m MAE using only free satellite input. The AR-flow reduces sliced Wasserstein distance from the Anchor's 1.13 to 0.208, at a higher MAE of 3.68 m. Crown-scale statistics improve, but strict individual-tree detection remains weak: generated sub-pixel detail represents plausible structure rather than measured individual trees. Quarterly continental refresh becomes tractable, letting satellite monitoring inform timing and extent for more expensive data acquisitions.

## Correcting Flows with Marginal Matching

DevFeed: [Correcting Flows with Marginal Matching](<https://devfeed.tech/articles/correcting-flows-with-marginal-matching-141851.md>)

Original publisher: [Read original article](<https://arxiv.org/abs/2611.23134>)

Author: Meital Bojan, Nadav Bojan Sellam, Ev Zisselman, Aviv Tamar

Published: 2026-11-18T05:00:00Z

Content type: article

Language: en

Sources: [arXiv: Machine Learning](<https://devfeed.tech/sources/arxiv-machine-learning.md>)

Topics: [cifar-10](<https://devfeed.tech/topics/cifar-10.md>), [ImageNet](<https://devfeed.tech/topics/imagenet.md>)

Tags: [cifar-10](<https://devfeed.tech/tags/cifar-10.md>), [cs-lg](<https://devfeed.tech/tags/cs-lg.md>), [diffusion-models](<https://devfeed.tech/tags/diffusion-models.md>), [distribution](<https://devfeed.tech/tags/distribution.md>), [flow-matching](<https://devfeed.tech/tags/flow-matching.md>), [generative-models](<https://devfeed.tech/tags/generative-models.md>), [machine-learning](<https://devfeed.tech/tags/machine-learning.md>), [matching](<https://devfeed.tech/tags/matching.md>), [ode](<https://devfeed.tech/tags/ode.md>)

### AI overview

Marginal Matching adds a training-based predictor-corrector framework to flow matching models by using the reversibility of their ODE: at a forward time step, it corrects samples toward the matching marginal distribution of the backward ODE. The authors say this reduces a bound on the distance between the data and learned distributions, and introduce two implementations of the correction function. They report improved sample quality on CIFAR-10 and ImageNet-64 with minimal added computation time. The supplied text does not describe the two implementations or provide quantitative results.

### Source excerpt

arXiv:2611.23134v1 Announce Type: new Abstract: Flow matching models, ODE-based generative models, generate samples by gradually morphing a simple source distribution into a target distribution. However, these models still face challenges in optimally balancing computational efficiency with generation fidelity. While predictor-corrector schemes have been proven to help address this issue in diffusion models effectively, their application to flow matching models remains under-explored. In this work we present a novel training-based predictor-corrector framework for flow matching, which exploits the fact that the ODE can be easily reversed. Our key idea, termed Marginal Matching, is to correct samples at a particular time step in the forward ODE towards a matching marginal distribution of the backward ODE. This scheme proves to reduce a bound on the distance between the data distribution and the learned distribution, motivating two different implementations for the correction function. We show that our proposed method improves sample quality on CIFAR-10 and ImageNet-64, with minimal overhead in computation time.

## When Anchors Shift the Target: Advantage-Gated Adaptation for Diffusion Reinforcement Learning

DevFeed: [When Anchors Shift the Target: Advantage-Gated Adaptation for Diffusion Reinforcement Learning](<https://devfeed.tech/articles/when-anchors-shift-the-target-advantage-gated-adaptation-for-diffusion-reinforcement-learning-141832.md>)

Original publisher: [Read original article](<https://arxiv.org/abs/2611.23038>)

Author: Pu Cao, Min Zhou, Ying Shu, Tiezheng Ge, Qing Song, Lu Yang

Published: 2026-11-18T05:00:00Z

Content type: article

Language: en

Sources: [arXiv: Machine Learning](<https://devfeed.tech/sources/arxiv-machine-learning.md>)

Topics: [Reinforcement learning](<https://devfeed.tech/topics/reinforcement-learning.md>), [Training AI Models](<https://devfeed.tech/topics/training-ai-models.md>)

Tags: [alignment](<https://devfeed.tech/tags/alignment.md>), [coupling](<https://devfeed.tech/tags/coupling.md>), [cs-cv](<https://devfeed.tech/tags/cs-cv.md>), [cs-lg](<https://devfeed.tech/tags/cs-lg.md>), [diffusion](<https://devfeed.tech/tags/diffusion.md>), [flow-matching](<https://devfeed.tech/tags/flow-matching.md>), [generation](<https://devfeed.tech/tags/generation.md>), [gradient](<https://devfeed.tech/tags/gradient.md>), [guidance](<https://devfeed.tech/tags/guidance.md>), [horizon](<https://devfeed.tech/tags/horizon.md>), [policy](<https://devfeed.tech/tags/policy.md>), [quality](<https://devfeed.tech/tags/quality.md>), [reference](<https://devfeed.tech/tags/reference.md>), [regression](<https://devfeed.tech/tags/regression.md>), [reinforcement-learning](<https://devfeed.tech/tags/reinforcement-learning.md>), [rendering](<https://devfeed.tech/tags/rendering.md>), [reward](<https://devfeed.tech/tags/reward.md>), [schedules](<https://devfeed.tech/tags/schedules.md>), [standard](<https://devfeed.tech/tags/standard.md>), [tangent](<https://devfeed.tech/tags/tangent.md>)

### AI overview

Advantage-Gated Adaptation (AGA) separates positive target fitting from negative-sample stabilization in diffusion and flow-matching reinforcement learning. Positive samples keep standard regression targets, so their updates vanish at the observed target regardless of the moving policy anchor. Negative samples use a first-order tangent at the moving policy's prediction, retaining the local repulsion direction with a prediction-independent gradient. In the quadratic-anchor family, the paper derives the parameter-free gate ω(a)=max(-a,0). AGA treats the moving copy as a local linearization aid and assigns longer-term behavior preservation to a frozen pretrained reference regularized at its operating classifier-free guidance scale. Across compositional generation, text rendering, and preference alignment, the reported experiments find that shifting positive targets harms visual quality and increases sensitivity to policy-copy schedules, while AGA improves the reward-quality trade-off.

### Source excerpt

arXiv:2611.23038v1 Announce Type: cross Abstract: Forward-process reinforcement learning adapts diffusion and flow-matching models by re-noising rewarded generations into regression targets. Because negative advantages induce concave updates with unbounded repulsive gradients, existing methods stabilize training using cached predictions from a moving policy copy, yet these anchors also alter what positive feedback teaches the model to fit. We formalize this conflict as anchor coupling: after separating frozen-reference regularizers, affine moving-anchor rules cannot eliminate negative curvature while preserving every positive regression target. We propose Advantage-Gated Adaptation (AGA) to decouple these roles. Positive samples retain standard target fitting, whose self-limiting update vanishes at the observed target independently of the moving anchor. Negative samples instead use the first-order tangent of the concave objective at the moving-policy prediction, preserving the local repulsion direction with a prediction-independent gradient. Within the quadratic-anchor family, these requirements uniquely determine the parameter-free gate $\omega(a)=\max(-a,0)$. Because the moving copy provides local linearization rather than retention, AGA assigns long-horizon behavior preservation to a frozen pretrained reference regularized at its operating classifier-free guidance scale, interpreted through an implicit-reward lens. Across compositional generation, text rendering, and preference alignment, our experiments show that displacing positive targets degrades visual quality and heightens sensitivity to policy-copy schedules, whereas AGA consistently improves the reward--quality trade-off.

## Every Measurement, Every Direction, All at Once: Multimodal Flow Matching for Molecules and Spectra

DevFeed: [Every Measurement, Every Direction, All at Once: Multimodal Flow Matching for Molecules and Spectra](<https://devfeed.tech/articles/every-measurement-every-direction-all-at-once-multimodal-flow-matching-for-molecules-and-spectra-141733.md>)

Original publisher: [Read original article](<https://arxiv.org/abs/2611.22098>)

Author: Thorben Prein, Elton Pan, Rafael G\\'omez-Bombarelli, Santiago Miret

Published: 2026-11-18T05:00:00Z

Content type: article

Language: en

Sources: [arXiv: Machine Learning](<https://devfeed.tech/sources/arxiv-machine-learning.md>)

Topics: [Transformers](<https://devfeed.tech/topics/transformers.md>), [Neural Network](<https://devfeed.tech/topics/neural-network.md>)

Tags: [3d](<https://devfeed.tech/tags/3d.md>), [cs-lg](<https://devfeed.tech/tags/cs-lg.md>), [denoising](<https://devfeed.tech/tags/denoising.md>), [experiment](<https://devfeed.tech/tags/experiment.md>), [flow-matching](<https://devfeed.tech/tags/flow-matching.md>), [generation](<https://devfeed.tech/tags/generation.md>), [generative](<https://devfeed.tech/tags/generative.md>), [generative-models](<https://devfeed.tech/tags/generative-models.md>), [graph](<https://devfeed.tech/tags/graph.md>), [ir](<https://devfeed.tech/tags/ir.md>), [masking](<https://devfeed.tech/tags/masking.md>), [measurements](<https://devfeed.tech/tags/measurements.md>), [model](<https://devfeed.tech/tags/model.md>), [models](<https://devfeed.tech/tags/models.md>), [multimodal](<https://devfeed.tech/tags/multimodal.md>), [physics-chem-ph](<https://devfeed.tech/tags/physics-chem-ph.md>), [prediction](<https://devfeed.tech/tags/prediction.md>), [random](<https://devfeed.tech/tags/random.md>), [signal](<https://devfeed.tech/tags/signal.md>), [structure](<https://devfeed.tech/tags/structure.md>), [supervision](<https://devfeed.tech/tags/supervision.md>), [tasks](<https://devfeed.tech/tags/tasks.md>)

### AI overview

MOSAIC uses random masking to turn each experiment's measurements into multiple conditioning and prediction tasks, training one flow matching transformer for any-to-any generation across molecular 3D structures and UV, IR, and Raman spectra. On QM9S, it reports 88.8% molecular graph recovery, 23.2% better structure generation than the strongest spectra-to-structure baseline, and stronger spectral generation than a structure-to-spectra specialist. Ablations report improved structure generation over structure-only training across model sizes. The model also accepts arbitrary subsets of available measurements and supports molecular identification by matching generated spectra to experimental references.

### Source excerpt

arXiv:2611.22098v1 Announce Type: new Abstract: Generative modeling in science is often data-limited, where experiments are expensive. However, each experiment may yield several coupled measurements of the same physical entity. Standard generative models often train only on a single direction, discarding supervision from the rest. We argue that each experiment contains more training signal than this asymmetry allows the model to exploit: random masking enables measurements to serve as either conditioning inputs or prediction targets, yielding multiple supervised denoising tasks from a single experiment. We investigate this premise with MOSAIC, a multimodal flow matching transformer over molecular 3D structure and UV, IR, and Raman spectra. MOSAIC is a single model that is capable of any-to-any generation: generating any subset of modalities from any disjoint subset of the others. On QM9S, MOSAIC achieves 88.8% molecular graph recovery, outperforming the strongest spectra-to-structure baseline by 23.2% in structure generation, while simultaneously outperforming a structure-to-spectra specialist model in spectral generation. Ablations show that the any-to-any objective improves structure generation over structure-only training across model sizes. Furthermore, MOSAIC supports arbitrary subsets of available measurements and enables molecular identification by matching generated spectra to experimental references.

## FACE: Flow-Guided Angular Channel Estimation Framework for mmWave MIMO Systems

DevFeed: [FACE: Flow-Guided Angular Channel Estimation Framework for mmWave MIMO Systems](<https://devfeed.tech/articles/face-flow-guided-angular-channel-estimation-framework-for-mmwave-mimo-systems-141668.md>)

Original publisher: [Read original article](<https://arxiv.org/abs/2611.21698>)

Author: Samitha Gunarathne, Eslam Eldeeb, Nurul Huda Mahmood, Matti Latva-aho

Published: 2026-11-18T05:00:00Z

Content type: article

Language: en

Sources: [arXiv: Machine Learning](<https://devfeed.tech/sources/arxiv-machine-learning.md>)

Topics: [Machine learning](<https://devfeed.tech/topics/machine-learning.md>), [communications](<https://devfeed.tech/topics/communications.md>)

Tags: [angular](<https://devfeed.tech/tags/angular.md>), [channel](<https://devfeed.tech/tags/channel.md>), [cost](<https://devfeed.tech/tags/cost.md>), [cs-lg](<https://devfeed.tech/tags/cs-lg.md>), [csi](<https://devfeed.tech/tags/csi.md>), [diffusion](<https://devfeed.tech/tags/diffusion.md>), [distributions](<https://devfeed.tech/tags/distributions.md>), [domain](<https://devfeed.tech/tags/domain.md>), [estimator](<https://devfeed.tech/tags/estimator.md>), [flow](<https://devfeed.tech/tags/flow.md>), [flow-matching](<https://devfeed.tech/tags/flow-matching.md>), [framework](<https://devfeed.tech/tags/framework.md>), [machine-learning](<https://devfeed.tech/tags/machine-learning.md>), [mimo](<https://devfeed.tech/tags/mimo.md>), [mmwave](<https://devfeed.tech/tags/mmwave.md>), [pilot](<https://devfeed.tech/tags/pilot.md>), [sampling](<https://devfeed.tech/tags/sampling.md>), [sparsity](<https://devfeed.tech/tags/sparsity.md>), [time](<https://devfeed.tech/tags/time.md>)

### AI overview

FACE uses flow matching as a generative channel prior, exploiting angular-domain sparsity to estimate mmWave MIMO channels from noisy, limited pilot observations. The abstract reports higher accuracy than conventional and diffusion-based estimators across diverse configurations, with high accuracy retained using fewer pilots and average achievable rate close to the perfect-CSI benchmark. Compared with diffusion-based estimation, FACE uses substantially fewer sampling steps and achieves superior or comparable accuracy, reducing inference time and computational cost. These are numerical results reported in the abstract; the excerpt does not provide experimental details.

### Source excerpt

arXiv:2611.21698v1 Announce Type: new Abstract: Acquiring accurate channel state information (CSI) is essential for millimeter-wave (mmWave) multi-input multi-output (MIMO) systems. However, it remains challenging due to high-dimensional channels, limited pilot observations, and measurement noise. To address these challenges, we propose flow-guided angular channel estimation (FACE), a novel channel estimation framework based on flow matching, an emerging generative modeling paradigm that learns continuous transport dynamics between probability distributions. FACE exploits the inherent angular-domain sparsity of mmWave channels and incorporates the learned flow matching model as a generative channel prior. Numerical results show that FACE consistently achieves higher channel estimation accuracy than both conventional estimators and diffusion-based generative estimator across diverse channel and system configurations. Moreover, FACE maintains high estimation accuracy with fewer pilot observations and achieves an average achievable rate close to the perfect-CSI benchmark. Compared with diffusion-based estimation, FACE requires substantially fewer sampling steps while achieving superior or comparable estimation accuracy, reducing inference time and computational cost. Overall, the proposed framework demonstrates the potential of flow matching for accurate, pilot-efficient, and computationally efficient mmWave MIMO channel estimation.

## Physically Admissible Generative Flow Matching for Sparse Flow Reconstruction

DevFeed: [Physically Admissible Generative Flow Matching for Sparse Flow Reconstruction](<https://devfeed.tech/articles/physically-admissible-generative-flow-matching-for-sparse-flow-reconstruction-141786.md>)

Original publisher: [Read original article](<https://arxiv.org/abs/2611.22549>)

Author: Yuhao Wang, Yuanxi Peng

Published: 2026-11-18T05:00:00Z

Content type: article

Language: en

Sources: [arXiv: Machine Learning](<https://devfeed.tech/sources/arxiv-machine-learning.md>)

Topics: [Computational Fluid Dynamics (CFD)](<https://devfeed.tech/topics/cfd.md>), [generative-ai-tools](<https://devfeed.tech/topics/generative-ai-tools.md>), [concept-learning](<https://devfeed.tech/topics/concept-learning.md>)

Tags: [boundary](<https://devfeed.tech/tags/boundary.md>), [bounded](<https://devfeed.tech/tags/bounded.md>), [consistency](<https://devfeed.tech/tags/consistency.md>), [continuous](<https://devfeed.tech/tags/continuous.md>), [coverage](<https://devfeed.tech/tags/coverage.md>), [cs-lg](<https://devfeed.tech/tags/cs-lg.md>), [energy](<https://devfeed.tech/tags/energy.md>), [error](<https://devfeed.tech/tags/error.md>), [fields](<https://devfeed.tech/tags/fields.md>), [flow](<https://devfeed.tech/tags/flow.md>), [flow-matching](<https://devfeed.tech/tags/flow-matching.md>), [fluid](<https://devfeed.tech/tags/fluid.md>), [generative](<https://devfeed.tech/tags/generative.md>), [generative-models](<https://devfeed.tech/tags/generative-models.md>), [global](<https://devfeed.tech/tags/global.md>), [intermediate](<https://devfeed.tech/tags/intermediate.md>), [invariant](<https://devfeed.tech/tags/invariant.md>), [laws](<https://devfeed.tech/tags/laws.md>), [loops](<https://devfeed.tech/tags/loops.md>), [matrix](<https://devfeed.tech/tags/matrix.md>), [metric](<https://devfeed.tech/tags/metric.md>)

### AI overview

Feasible Subspace Consistency (FSC) keeps the full generative ODE trajectory within the discrete conservation manifold: tangent-space matching enforces physical admissibility, while mass-orthonormal coordinates preserve the kinetic-energy metric. Sparse reconstruction uses amortized posterior transport in a frozen FSC prior's latent space, so samples inherit incompressibility and boundary conditions without per-sample projection. The authors evaluate exact spectral projection on periodic Kolmogorov flow and matrix-free iterative projection on a wall-bounded cylinder wake with immersed obstacles. They report that FSC avoids the mode-coverage collapse and high-wavenumber distortion associated with terminal projection, raising mode coverage from 0.000 to 0.790. Against the unconditioned prior, amortized FSC reduces global relative error by 50% and 46%, and sensor residuals by 63% and 55%, while providing calibrated, spatially heterogeneous uncertainty.

### Source excerpt

arXiv:2611.22549v1 Announce Type: new Abstract: Reconstructing full fluid velocity fields from sparse, localized observations is a severely ill-posed inverse problem: most of the spatial domain remains unobserved, discrete conservation laws must hold strictly, and predictive uncertainty must be reliable. Existing generative models typically enforce physical constraints only at the terminal output or via soft penalties, leaving intermediate generative trajectories unconstrained. We introduce Feasible Subspace Consistency (FSC), a framework that preserves physical admissibility as an invariant of continuous generative dynamics: tangent-space matching confines the entire ODE trajectory to the discrete conservation manifold, while mass-orthonormal coordinates preserve the kinetic-energy metric. Sparse reconstruction then reduces to amortized posterior transport in the latent space of a frozen FSC prior, ensuring every posterior sample inherits incompressibility and boundary conditions without per-sample projection loops. We validate FSC on a periodic Kolmogorov flow, where exact spectral projection isolates path tangency and metric isometry, and a wall-bounded cylinder wake with immersed obstacles, realized via matrix-free iterative projection. FSC prevents the mode-coverage collapse induced by terminal projection (increasing mode coverage from $0.000$ to $0.790$) while eliminating high-wavenumber spectral distortion. In sparse reconstruction, amortized FSC reduces global relative error by $50\%$ and $46\%$, and sensor residuals by $63\%$ and $55\%$, relative to the unconditioned prior, while delivering calibrated, spatially heterogeneous uncertainty.

## Explicit Regularization for Flow Matching

DevFeed: [Explicit Regularization for Flow Matching](<https://devfeed.tech/articles/explicit-regularization-for-flow-matching-138893.md>)

Original publisher: [Read original article](<https://arxiv.org/abs/2611.19587>)

Author: Lars Ruthotto, Eran Treister

Published: 2026-11-17T05:00:00Z

Content type: article

Language: en

Sources: [arXiv: Machine Learning](<https://devfeed.tech/sources/arxiv-machine-learning.md>)

Topics: [Machine learning](<https://devfeed.tech/topics/machine-learning.md>), [Occam's razor machine learning](<https://devfeed.tech/topics/occam-s-razor-machine-learning.md>), [machine learning overfitting](<https://devfeed.tech/topics/machine-learning-overfitting.md>)

Tags: [cs-lg](<https://devfeed.tech/tags/cs-lg.md>), [cs-na](<https://devfeed.tech/tags/cs-na.md>), [diffusion](<https://devfeed.tech/tags/diffusion.md>), [distribution](<https://devfeed.tech/tags/distribution.md>), [experiments](<https://devfeed.tech/tags/experiments.md>), [flow-matching](<https://devfeed.tech/tags/flow-matching.md>), [generative-models](<https://devfeed.tech/tags/generative-models.md>), [kernel](<https://devfeed.tech/tags/kernel.md>), [learning](<https://devfeed.tech/tags/learning.md>), [math-na](<https://devfeed.tech/tags/math-na.md>), [statistics](<https://devfeed.tech/tags/statistics.md>), [velocity](<https://devfeed.tech/tags/velocity.md>)

### AI overview

Penalizing velocity derivatives can prevent flow matching from collapsing onto training samples: the empirical problem is well posed uniformly in time when the derivative order k satisfies 2k > D and training points are in general position. The abstract contrasts this with penalizing velocity magnitude, which avoids memorization only as its parameter grows sufficiently large. In one- and two-dimensional spline experiments using direct least-squares solvers, Hessian-regularized flows changed little as space and time discretizations were refined, while unregularized error approached memorization. The regularized flows generalized with errors comparable to or below well-tuned kernel density estimates.

### Source excerpt

arXiv:2611.19587v1 Announce Type: new Abstract: Learning the underlying probability distribution from samples is a core problem in statistics and applied mathematics. Generative models based on diffusion and flow matching learn a velocity field whose flow transports a known distribution to the data distribution. They produce novel, realistic samples, although the empirical learning problem is ill posed near the data distribution. Its minimizer transports the samples onto the training data, and small perturbations of the data change the velocity field strongly near the end of the flow. The observed generalization is often attributed to implicit regularization by the network parameterization, early stopping of the training, and the time integration during sampling. Explicit regularization has advanced the theory and computation of many inverse problems and has recently been proposed to prevent memorization in generative models. With a regularizer on the magnitude of the velocity, however, the field approaches the memorizing velocity unless the regularization parameter grows to infinity fast enough, and it then stays bounded by a constant that is small only for large parameters. We show that penalizing the derivatives of order $k$ of the velocity renders the $D$-dimensional empirical problem well posed uniformly in time if $2k > D$ and the training points are in general position. Our experiments in one and two dimensions discretize the velocity with splines and use direct solvers for the regularized least-squares problems to exclude optimization error. Refining the discretization in space and time changes the error of the Hessian-regularized flows only marginally, whereas without the regularizer the error tends to that of memorization. These flows generalize, with errors comparable to or lower than those of well-tuned kernel density estimates.

## Residual-Tail-Informed Flow Matching for Time Series Anomaly Detection

DevFeed: [Residual-Tail-Informed Flow Matching for Time Series Anomaly Detection](<https://devfeed.tech/articles/residual-tail-informed-flow-matching-for-time-series-anomaly-detection-137563.md>)

Original publisher: [Read original article](<https://arxiv.org/abs/2611.20334>)

Author: Shyam Sundar Saravanan, Tie Luo, Sanjay Madria

Published: 2026-11-17T05:00:00Z

Content type: article

Language: en

Sources: [arXiv: Artificial Intelligence](<https://devfeed.tech/sources/arxiv-artificial-intelligence.md>), [arXiv: Machine Learning](<https://devfeed.tech/sources/arxiv-machine-learning.md>)

Topics: [Time Series](<https://devfeed.tech/topics/time-series.md>), [AI Models](<https://devfeed.tech/topics/ai-models.md>), [VAE](<https://devfeed.tech/topics/vae.md>), [Temporal data](<https://devfeed.tech/topics/temporal-data.md>)

Tags: [anomaly-detection](<https://devfeed.tech/tags/anomaly-detection.md>), [autoencoder](<https://devfeed.tech/tags/autoencoder.md>), [cs-ai](<https://devfeed.tech/tags/cs-ai.md>), [cs-lg](<https://devfeed.tech/tags/cs-lg.md>), [flow-matching](<https://devfeed.tech/tags/flow-matching.md>), [machine-learning](<https://devfeed.tech/tags/machine-learning.md>), [time-series](<https://devfeed.tech/tags/time-series.md>)

### AI overview

RT-FM targets false alarms on rare but valid time-series regimes that data-space flow matching may score like true anomalies. It uses a temporal variational autoencoder to estimate residual latent tailness, then turns that estimate into bounded, warmup-annealed weights for training an unsupervised data-space flow-matching model. Across seven TSAD benchmarks, the authors report better rare-normal coverage, fewer false alarms on those windows, and improved aggregate detection while preserving strong anomaly discrimination.

### Source excerpt

arXiv:2611.20334v1 Announce Type: cross Abstract: Time series anomaly detection (TSAD) models are often trained on nominal data using average-risk objectives, which emphasize dominant normal patterns. However, real time series can contain rare but valid nominal regimes that are underrepresented in the training distribution. We show that data-space flow matching can over-penalize such rare-normal windows and assign them anomaly scores comparable to those of true anomalies. To address this, we propose Residual-Tail-Informed Flow Matching (RT-FM), an unsupervised framework that uses a temporal variational autoencoder to estimate residual latent tailness and converts it into bounded, warmup-annealed weights for training a data-space flow matching model. Experiments on seven TSAD benchmarks show that RT-FM improves rare-normal coverage while preserving strong anomaly discrimination, substantially reducing false alarms on such windows and resulting in improved aggregate detection performance. The code is available at: https://github.com/trustAIoT/RT-FM_time-series-anomaly

## Chance-constrained Flow Matching for High-Fidelity Constraint-aware Generation

DevFeed: [Chance-constrained Flow Matching for High-Fidelity Constraint-aware Generation](<https://devfeed.tech/articles/chance-constrained-flow-matching-for-high-fidelity-constraint-aware-generation-138725.md>)

Original publisher: [Read original article](<https://arxiv.org/abs/2509.25157>)

Author: Jinhao Liang, Yixuan Sun, Anirban Samaddar, Sandeep Madireddy, Ferdinando Fioretto

Published: 2026-11-17T05:00:00Z

Content type: article

Language: en

Sources: [arXiv: Machine Learning](<https://devfeed.tech/sources/arxiv-machine-learning.md>)

Topics: [High-Performance Computing](<https://devfeed.tech/topics/high-performance-computing.md>), [Operations research and optimization](<https://devfeed.tech/topics/operations-research-and-optimization.md>)

Tags: [constraint](<https://devfeed.tech/tags/constraint.md>), [cs-ai](<https://devfeed.tech/tags/cs-ai.md>), [cs-lg](<https://devfeed.tech/tags/cs-lg.md>), [experiments](<https://devfeed.tech/tags/experiments.md>), [flow-matching](<https://devfeed.tech/tags/flow-matching.md>), [generation](<https://devfeed.tech/tags/generation.md>), [generative-models](<https://devfeed.tech/tags/generative-models.md>), [optimization](<https://devfeed.tech/tags/optimization.md>), [partial-differential-equations](<https://devfeed.tech/tags/partial-differential-equations.md>), [physics-informed](<https://devfeed.tech/tags/physics-informed.md>)

### AI overview

Chance-constrained Flow Matching (CCFM) enforces hard constraints during generative sampling through stochastic optimization, while retaining the feasibility guarantee of projection-based methods. It avoids projecting noisy intermediate samples onto a feasible set defined for clean samples, which the abstract says can disrupt sampling dynamics, and avoids the multi-stage clean-sample estimation and projection used by recent approaches. The abstract reports stronger feasibility and fidelity than current state-of-the-art constrained generative models across PDE-governed physical systems, molecular docking, and physics-informed motion generation. The supplied text does not provide quantitative results or implementation details.

### Source excerpt

arXiv:2509.25157v2 Announce Type: replace Abstract: Generative models excel at synthesizing high-fidelity samples from complex data distributions, but they often violate hard constraints arising from physical laws or task specifications. A common remedy is to project intermediate samples onto the feasible set at each sampling step. However, repeated projection can disrupt the learned sampling dynamics because this feasible set is defined for clean samples rather than noisy intermediate states. Thus, recent studies first estimate clean samples and then project them to the feasible set, but this increases algorithmic complexity and accumulates errors across steps. To address these issues, this paper proposes Chance-constrained Flow Matching (CCFM), a training-free method that formulates constraint enforcement during sampling as stochastic optimization while retaining the hard-constraint feasibility guarantee of projection-based methods. By avoiding direct projection of noisy intermediate states onto the clean-sample feasible set, CCFM mitigates the distributional distortion. Unlike approaches that first estimate and project clean samples, CCFM avoids complex multi-stage procedures. Experiments show that CCFM outperforms current state-of-the-art constrained generative models in modeling complex physical systems governed by partial differential equations, molecular docking problems, and physics-informed motion generation, delivering superior feasibility and fidelity.

## Flow Matching for Offline Reinforcement Learning with Discrete Actions

DevFeed: [Flow Matching for Offline Reinforcement Learning with Discrete Actions](<https://devfeed.tech/articles/flow-matching-for-offline-reinforcement-learning-with-discrete-actions-138741.md>)

Original publisher: [Read original article](<https://arxiv.org/abs/2602.06138>)

Author: Fairoz Nower Khan, Nabuat Zaman Nahim, Ruiquan Huang, Haibo Yang, Peizhong Ju

Published: 2026-11-17T05:00:00Z

Content type: article

Language: en

Sources: [arXiv: Machine Learning](<https://devfeed.tech/sources/arxiv-machine-learning.md>)

Topics: [Reinforcement learning](<https://devfeed.tech/topics/reinforcement-learning.md>), [Artificial Intelligence](<https://devfeed.tech/topics/ai.md>)

Tags: [control](<https://devfeed.tech/tags/control.md>), [cs-lg](<https://devfeed.tech/tags/cs-lg.md>), [diffusion](<https://devfeed.tech/tags/diffusion.md>), [diffusion-models](<https://devfeed.tech/tags/diffusion-models.md>), [flow-matching](<https://devfeed.tech/tags/flow-matching.md>), [games](<https://devfeed.tech/tags/games.md>), [markov-chains](<https://devfeed.tech/tags/markov-chains.md>), [multi-agent](<https://devfeed.tech/tags/multi-agent.md>), [multimodal](<https://devfeed.tech/tags/multimodal.md>), [objectives](<https://devfeed.tech/tags/objectives.md>), [offline](<https://devfeed.tech/tags/offline.md>), [offline-reinforcement-learning](<https://devfeed.tech/tags/offline-reinforcement-learning.md>), [performance](<https://devfeed.tech/tags/performance.md>), [policy](<https://devfeed.tech/tags/policy.md>), [quantization](<https://devfeed.tech/tags/quantization.md>), [reinforcement-learning](<https://devfeed.tech/tags/reinforcement-learning.md>)

### AI overview

The paper extends flow matching to offline reinforcement learning with discrete actions by using continuous-time Markov chains trained with a Q-weighted objective. For multi-agent settings, it uses a factorized conditional path to mitigate the growth of joint action spaces. Under idealized conditions, the objective recovers the KL-regularized Boltzmann policy for an oracle policy-improvement step from the behavior distribution. The abstract reports robust, competitive results across high-dimensional control, multi-agent games, and changing multi-objective preferences, with advantages in multimodal decisions and preference-conditioned policy representation. Applying the discrete framework to continuous control requires action quantization, which trades representational complexity against performance.

### Source excerpt

arXiv:2602.06138v4 Announce Type: replace Abstract: Generative policies based on diffusion models and flow matching have shown strong promise for offline reinforcement learning (RL), but their applicability remains largely confined to continuous action spaces. To address a broader range of offline RL settings, we extend flow matching to a general framework that supports discrete action spaces with multiple objectives. Specifically, we replace continuous flows with continuous-time Markov chains, trained using a Q-weighted flow matching objective. We then extend our design to multi-agent settings, mitigating the exponential growth of joint action spaces via a factorized conditional path. For an oracle single policy-improvement step from the behavior distribution, we theoretically show that, under idealized conditions, the Q-weighted objective recovers the KL-regularized Boltzmann policy. Extensive experiments demonstrate that our method performs robustly across diverse settings and benchmarks, including high-dimensional control, multi-agent games, and dynamically changing preferences over multiple objectives, while achieving competitive performance and demonstrating advantages in multimodal decision-making and preference-conditioned policy representation. Our discrete framework can also be applied to continuous-control problems through action quantization, providing a flexible trade-off between representational complexity and performance.

## Conditional Optimal Transport and Triangular Maps for Imitation Learning

DevFeed: [Conditional Optimal Transport and Triangular Maps for Imitation Learning](<https://devfeed.tech/articles/conditional-optimal-transport-and-triangular-maps-for-imitation-learning-138939.md>)

Original publisher: [Read original article](<https://arxiv.org/abs/2611.19815>)

Author: Mohammad Al-Jarrah, Yuhang Mei, Karen Leung, Bamdad Hosseini, Amirhossein Taghvaei

Published: 2026-11-17T05:00:00Z

Content type: article

Language: en

Sources: [arXiv: Machine Learning](<https://devfeed.tech/sources/arxiv-machine-learning.md>)

Topics: [Machine learning](<https://devfeed.tech/topics/machine-learning.md>), [Model Development](<https://devfeed.tech/topics/model-development.md>), [Neural Network](<https://devfeed.tech/topics/neural-network.md>)

Tags: [block](<https://devfeed.tech/tags/block.md>), [conditional](<https://devfeed.tech/tags/conditional.md>), [consistency](<https://devfeed.tech/tags/consistency.md>), [cs-lg](<https://devfeed.tech/tags/cs-lg.md>), [diffusion](<https://devfeed.tech/tags/diffusion.md>), [distribution](<https://devfeed.tech/tags/distribution.md>), [evaluation](<https://devfeed.tech/tags/evaluation.md>), [evaluations](<https://devfeed.tech/tags/evaluations.md>), [flow](<https://devfeed.tech/tags/flow.md>), [flow-matching](<https://devfeed.tech/tags/flow-matching.md>), [gaussian](<https://devfeed.tech/tags/gaussian.md>), [history](<https://devfeed.tech/tags/history.md>), [imitation](<https://devfeed.tech/tags/imitation.md>), [imitation-learning](<https://devfeed.tech/tags/imitation-learning.md>), [inference](<https://devfeed.tech/tags/inference.md>), [learning](<https://devfeed.tech/tags/learning.md>), [map](<https://devfeed.tech/tags/map.md>), [maps](<https://devfeed.tech/tags/maps.md>), [matching](<https://devfeed.tech/tags/matching.md>), [optimal-transport](<https://devfeed.tech/tags/optimal-transport.md>), [transport](<https://devfeed.tech/tags/transport.md>)

### AI overview

The paper frames imitation learning as conditional sampling: a triangular transport map takes observation history and a reference distribution to demonstrated action chunks. Unlike Diffusion Policy's fixed Gaussian reference, the method uses the empirical distribution of demonstrated chunks. It presents three variants--optimal transport policy (OTP), flow matching policy (FMP), and stochastic interpolant policy (SIP). On Push-T and robomimic Square, they are reported as competitive with Diffusion Policy and Consistency Policy while using smaller networks and shorter training; the empirical reference contributes substantially when network evaluations are few. OTP produces an action chunk in one evaluation with the smallest network (0.23 million parameters on Push-T), while SIP has the highest scores among the proposed variants at higher cost.

### Source excerpt

arXiv:2611.19815v1 Announce Type: new Abstract: We introduce a unified formulation of imitation learning from demonstrations as a conditional sampling problem solved by triangular transport: a policy is the second block of a triangular map that transports a reference distribution to the conditional distribution of the action chunk given the observation history. Unlike in Diffusion Policy, where it is a fixed Gaussian, the reference becomes a design variable; we take it to be the empirical distribution of the demonstrated action chunks and introduce three instances of the formulation: an optimal transport policy (OTP), a flow matching policy (FMP), and a stochastic interpolant policy (SIP). On the Push-T and robomimic Square tasks, the three policies are competitive in task performance with Diffusion Policy and Consistency Policy while using considerably smaller networks and shorter training, and the informative reference accounts for a substantial part of their performance when few network evaluations are used. The three instances offer different trade-offs rather than a single best method: OTP generates an action chunk with a single evaluation of the smallest network (0.23 million parameters on Push-T) and is suitable for settings in which inference time and model size matter, whereas SIP attains the highest scores among the proposed policies at a higher cost.

## Foundation Pareto Flow Policy for Multi-Objective Reinforcement Learning

DevFeed: [Foundation Pareto Flow Policy for Multi-Objective Reinforcement Learning](<https://devfeed.tech/articles/foundation-pareto-flow-policy-for-multi-objective-reinforcement-learning-138943.md>)

Original publisher: [Read original article](<https://arxiv.org/abs/2611.19825>)

Author: Zhanjiang Yang, Lijun Sun, Yueming Li, Meng Li, Yijun Yang

Published: 2026-11-17T05:00:00Z

Content type: article

Language: en

Sources: [arXiv: Machine Learning](<https://devfeed.tech/sources/arxiv-machine-learning.md>)

Topics: [Reinforcement learning](<https://devfeed.tech/topics/reinforcement-learning.md>), [AI Foundation Models](<https://devfeed.tech/topics/ai-foundation-models.md>), [few-shot](<https://devfeed.tech/topics/few-shot.md>)

Tags: [benchmark](<https://devfeed.tech/tags/benchmark.md>), [continuous](<https://devfeed.tech/tags/continuous.md>), [control](<https://devfeed.tech/tags/control.md>), [cs-lg](<https://devfeed.tech/tags/cs-lg.md>), [data](<https://devfeed.tech/tags/data.md>), [dataset](<https://devfeed.tech/tags/dataset.md>), [diffusion](<https://devfeed.tech/tags/diffusion.md>), [few-shot](<https://devfeed.tech/tags/few-shot.md>), [flow](<https://devfeed.tech/tags/flow.md>), [flow-matching](<https://devfeed.tech/tags/flow-matching.md>), [foundation](<https://devfeed.tech/tags/foundation.md>), [foundation-models](<https://devfeed.tech/tags/foundation-models.md>), [fp2](<https://devfeed.tech/tags/fp2.md>), [learning](<https://devfeed.tech/tags/learning.md>), [modal](<https://devfeed.tech/tags/modal.md>), [multiple](<https://devfeed.tech/tags/multiple.md>), [objectives](<https://devfeed.tech/tags/objectives.md>), [offline](<https://devfeed.tech/tags/offline.md>), [pipeline](<https://devfeed.tech/tags/pipeline.md>), [policy](<https://devfeed.tech/tags/policy.md>), [reinforcement-learning](<https://devfeed.tech/tags/reinforcement-learning.md>)

### AI overview

FP2 targets offline multi-objective reinforcement learning for continuous robotic control, where policies must balance potentially conflicting rewards and generalize trade-offs to unseen tasks. It uses a multimodal diffusion transformer with a likelihood-free flow-matching loss to learn diverse trade-off trajectories. Its three-stage pipeline pre-trains on a large MORL dataset, mid-trains on a curated high-quality dataset, then post-trains on a small amount of trajectory data from new tasks and trade-offs. The abstract reports better performance than prior methods on most D4MORL tasks, including zero-shot interpolation and extrapolation across complex Pareto manifolds and few-shot adaptation to new tasks.

### Source excerpt

arXiv:2611.19825v1 Announce Type: new Abstract: The overarching goal of reinforcement learning (RL) is to efficiently optimize the agent's policy for various environments, tasks, dynamics, and even user preferences, which often consist of multiple and potentially conflicting rewards/objectives. This necessitates the learning of Pareto-optimal policies that can navigate different trade-offs among objectives. However, existing multi-objective RL (MORL) methods, whether trained online or offline, are typically limited to task- or objective-specific policy learning and do not address generalization of learned trade-offs to unseen tasks/objectives. In this work, we propose the Foundation Pareto Flow Policy (FP2) for offline MORL in continuous robotic control, inspired by the pre-mid-post-training paradigm underlying recent foundation models. Specifically, we instantiate FP2 as a multi-modal diffusion transformer equipped with a likelihood-free flow matching loss, enabling unified, scalable, and instructable learning of diverse trade-off trajectories. We further develop a three-stage training pipeline to gradually optimize the policy: (1) pre-training on a large-scale MORL dataset, (2) mid-training on a carefully curated high-quality dataset, and (3) post-training on few collected trajectory data from new tasks and trade-offs. Empirical results demonstrate that FP2 outperforms previous state-of-the-art methods on most tasks of the D4MORL benchmark, achieving superior zero-shot interpolation and extrapolation on complex Pareto manifolds as well as efficient few-shot adaptation of the learned trade-off prior to new continuous-control tasks. Code is available at https://github.com/FibonacciRabbit/FP2.

## Cache or Denoise Less? Residual Reuse in a Flow-Matching World Action Model

DevFeed: [Cache or Denoise Less? Residual Reuse in a Flow-Matching World Action Model](<https://devfeed.tech/articles/cache-or-denoise-less-residual-reuse-in-a-flow-matching-world-action-model-137486.md>)

Original publisher: [Read original article](<https://arxiv.org/abs/2611.19709>)

Author: Abhinav Gupta, Nissan Pow

Published: 2026-11-17T05:00:00Z

Content type: comparison

Language: en

Sources: [arXiv: Artificial Intelligence](<https://devfeed.tech/sources/arxiv-artificial-intelligence.md>), [arXiv: Machine Learning](<https://devfeed.tech/sources/arxiv-machine-learning.md>)

Topics: [Cache](<https://devfeed.tech/topics/cache.md>), [World models](<https://devfeed.tech/topics/world-models.md>), [Physical AI](<https://devfeed.tech/topics/physical-ai.md>)

Tags: [cache](<https://devfeed.tech/tags/cache.md>), [cs-ai](<https://devfeed.tech/tags/cs-ai.md>), [cs-cv](<https://devfeed.tech/tags/cs-cv.md>), [cs-lg](<https://devfeed.tech/tags/cs-lg.md>), [cs-ro](<https://devfeed.tech/tags/cs-ro.md>), [denoising](<https://devfeed.tech/tags/denoising.md>), [evaluation](<https://devfeed.tech/tags/evaluation.md>), [evaluations](<https://devfeed.tech/tags/evaluations.md>), [flow](<https://devfeed.tech/tags/flow.md>), [flow-matching](<https://devfeed.tech/tags/flow-matching.md>), [model](<https://devfeed.tech/tags/model.md>), [reuse](<https://devfeed.tech/tags/reuse.md>), [robot](<https://devfeed.tech/tags/robot.md>), [world](<https://devfeed.tech/tags/world.md>)

### AI overview

In Fast-WAM on LIBERO-Goal, two uncached denoising steps preserved success within two percentage points of ten steps across 500 initial states for one checkpoint and noise seed, while reducing replanning time from 202 to 92 ms on an RTX A6000. This leaves limited room for cross-cycle caching: retaining two fresh steps cannot beat two uncached steps because image encoding, the video expert, and those steps still run each call; caching one of two steps with refresh every fourth call saves at most 11-14% by the authors' cost model. Cache performance was sensitive to sampler settings: with Fast-WAM's current time shift, C³ache's eight cached steps covered 80% of the flow path and achieved 63% success versus 95% uncached. Resampling noise each call reduced success by 14-40 points for caches covering at least 60% of the path. The findings are limited to the reported setup; the authors recommend evaluating reuse under deployed sampler and noise settings.

### Source excerpt

arXiv:2611.19709v1 Announce Type: cross Abstract: Flow-matching robot policies generate each action chunk through repeated network evaluations, one per denoising step. Cross-cycle caches such as C$^3$ache reuse computation from earlier replanning calls, but whether they beat simply taking fewer steps is unclear. We compare the two in Fast-WAM, a 6B-parameter world action model, on LIBERO-Goal. Two uncached steps keep the success of ten: on all 500 initial states they are non-inferior within two percentage points, for one checkpoint and noise seed. Under PyTorch's default compile mode, they also cut a replanning call on an RTX A6000 from 202 to 92 ms. This leaves little room for caching. A cache that keeps two fresh steps cannot be faster than two uncached steps: every call still encodes the images, runs the video expert and computes those two steps. Caching the first of two steps, refreshed every fourth call, saves at most 11--14% of a call by our cost model. Cached policies are also fragile. Their losses tracked the share of the flow path the cached steps cover: under Fast-WAM's current time shift, C$^3$ache's eight cached steps cover 80% of the path instead of 44%, and the cached policy succeeds on 63% of states against 95% uncached. Resampling the noise on every call cost caches covering 60% or more of the path 14--40 points. These results argue for fewer-step baselines and for testing reuse under the deployed sampler and noise settings. We release our evaluation harness, logs and a checklist.

[Next page](<https://devfeed.tech/tags/flow-matching.md?cursor=WyIyMDI2LTExLTE3VDA1OjAwOjAwKzAwOjAwIiwgIjVjZDA4OWExLTAzYjAtNGFjYS04OTRmLTc5ZTg0Yjc2MmQxOSJd>)