---
title: Continuous & Discrete Flow Matching
url: https://www.emergentmind.com/topics/continuous-and-discrete-flow-matching
type: topic
---

# Continuous & Discrete Flow Matching

Flow matching denotes a family of generative modeling methods that transport a simple source or prior distribution to a target or data distribution along a probability path. In its continuous form, the learned object is a time-dependent velocity field whose ordinary differential equation transports samples through continuous normalizing flows; in discrete settings, closely related constructions replace vector fields by probability velocities, continuous-time Markov chain generators, or continuous reparameterizations of discrete distributions on structured statistical manifolds [2501.16839][2407.15595][2504.10283]. Recent work has expanded this family to discrete ordinal data, categorical data, mixed continuous–discrete representations, manifold-valued states, and variable-cardinality objects, while also sharpening the mathematical limits imposed by topology, stochasticity, and representation geometry [2605.00360][2512.12821][2511.09465].

## 1. General formulation and probability paths

In the continuous setting, flow matching learns velocity fields of curves connecting a latent distribution and a target distribution. A standard formulation specifies a curve of measures \(\mu_t\), \(t \in [0,1]\), together with a velocity field \(v_t\) satisfying the continuity equation
$$
\partial_t \mu_t + \nabla_x \cdot (\mu_t v_t) = 0,
$$
and the sample-wise flow ODE
$$
\partial_t \phi(t,x) = v_t(\phi(t,x)), \qquad \phi(0,x)=x,
$$
with \(\mu_t = \phi(t,\cdot)_{\#}\mu_0\) [2501.16839]. The same review shows that such curves and their velocities can be characterized and learned via transport plans, Markov kernels, and stochastic processes, with the latter two strictly broader than the coupling approach [2501.16839].

In discrete settings, the corresponding evolution is not written through a Euclidean continuity equation. Instead, probability flow is governed by the Kolmogorov forward equation
$$
\frac{d}{dt} p_t(y) = \sum_{x \in \mathcal{X}} u_t(y,x)\,p_t(x),
$$
where \(u_t(y,x)\) is the transition rate of a continuous-time Markov chain on the discrete state space \(\mathcal{X}\) [2602.06138]. Discrete Flow Matching further generalizes the construction by working with a general family of probability paths interpolating between source and target distributions, and by using learned posteriors such as probability denoiser (\(x\)-prediction) and noise-prediction (\(\epsilon\)-prediction) for sampling [2407.15595].

The main regimes are summarized below.

| Regime | State evolution | Canonical object |
|---|---|---|
| Continuous flow matching | ODE and continuity equation | Velocity field \(v_t(x)\) |
| Discrete-state flow matching | Kolmogorov forward equation | CTMC generator \(u_t(y,x)\) |
| Continuous-state discrete flow matching | Geometry on probability simplex or sphere | Flow in probability representation |

This suggests that “flow matching” now names a common transport-based viewpoint rather than a single parameterization. The unifying theme is a prescribed probability path from source to target together with a learned dynamical object that makes sampling feasible [2501.16839][2504.10283].

## 2. Continuous flow matching and the problem of continuity

Flow matching emerged as a framework for generative modeling through continuous normalizing flows [2512.12821]. A typical formulation connects a prior \(p_0\) and a target \(p_1\) through an ODE
$$
\frac{dx_t}{dt} = v_t(x_t),
$$
with the associated continuity equation
$$
\frac{\partial p_t}{\partial t} + \nabla \cdot (p_t v_t) = 0.
$$
For Gaussian conditionals, conditional flow matching yields the optimal velocity
$$
v_t^*(x) = \frac{1}{1-t}\left(\mathbb{E}_{p_1(x_1 \mid x,t)}[x_1] - x\right)
$$
[2512.12821].

A central theoretical development concerns topological mismatch between the prior and target. When a unimodal prior is transported to a multimodal target, the optimal velocity field under standard flow matching objectives may be spatially discontinuous [2512.12821]. In the paper’s bimodal Gaussian mixture analysis, the intermediate distribution is a mixture of two overlapping Gaussians, and there is a decision-boundary hyperplane
$$
H_t = \{x : \langle x,\mu_1-\mu_2\rangle = 0\}
$$
across which the posterior destination flips sharply from one mode to the other [2512.12821]. The resulting jump discontinuity has magnitude approaching infinity as \(t \to 1\) [2512.12821].

The distinction between temporal and spatial continuity is therefore essential. Temporal continuity of trajectories with respect to \(t\) is maintained by ODEs, but spatial continuity with respect to \(x\) at fixed \(t\) is not guaranteed [2512.12821]. Because neural networks used to parameterize \(v_\theta(x,t)\) are by design spatially continuous, they can only learn a smoothed approximation when the true optimal velocity is discontinuous. The paper identifies mode averaging and mode collapse as resulting artifacts, and gives a lower-bounded approximation error that diverges as \(t \to 1\) [2512.12821].

A common misconception is that continuous sample paths imply a continuous velocity field everywhere in state space. Theoretical analysis in the bimodal setting shows that this is false: continuous trajectories can coexist with spatial jump discontinuities in the optimal velocity field [2512.12821]. The same work further argues that the issue is not specific to \(L^2\) loss, but may be a consequence of topological mismatch between distributions, with implications for manifold flow matching and representation learning [2512.12821].

## 3. Discrete flow matching, stochastic paths, and generator learning

Discrete Flow Matching was introduced as a discrete flow paradigm designed specifically for generating discrete data, with four stated contributions: a general family of probability paths, a generic formula for sampling from these paths using learned posteriors, practical gains from specific schedulers, and strong scaling on code generation benchmarks [2407.15595]. In its simple convex-interpolation form, the conditional path can be written as
$$
p_t(x_i \mid x_0,x_1) = (1-k_t)\delta_{x_0}(x_i) + k_t\delta_{x_1}(x_i),
$$
while the marginal path is induced through a coupling \(\mathcal{T}(x_0,x_1)\) [2407.15595]. Sampling is then driven by a probability velocity, for example
$$
u_t(x_i,z) = \frac{k_t}{1-k_t}\left[p_{1 \mid t}(x_i \mid z) - \delta_{z_i}(x_i)\right],
$$
with an analogous \(\epsilon\)-prediction form using \(p_{0 \mid t}\) [2407.15595].

The discrete setting differs structurally from the continuous one because its sample paths are stochastic. For that reason, the rectification strategy used in continuous flow matching does not directly extend to the discrete one [2411.00759]. To address this, recent work proposes a dynamic-optimal-transport-like minimization objective for discrete flows with convex interpolants and derives an equivalent Kantorovich formulation in which transport cost depends only on inter-state similarity; this cost can be optimized with a minibatch strategy [2411.00759]. The same paper also derives an upper bound on perplexity, motivated by the fact that discrete models lack an instantaneous change-of-variables mechanism and therefore do not admit the same density evaluation tools as continuous flows [2411.00759].

A complementary line of work studies discrete flow through generator matching. “Error Analysis of Discrete Flow with Generator Matching” derives a novel Girsanov-type theorem for CTMCs and expresses the KL divergence between two path measures through a Bregman divergence between the transition rates [2509.21906]. Building on generator matching and uniformization, it establishes non-asymptotic error bounds for distribution estimation and emphasizes a contrast with discrete diffusion: discrete flow incurs no truncation error caused by truncating the time horizon in the noising process [2509.21906].

A second common misconception is that discrete flow matching is merely a tokenized version of continuous flow matching. The literature instead presents a distinct stochastic theory: CTMC generators replace vector fields, rectification no longer transfers directly, and evaluation requires tools such as perplexity upper bounds, path-space KL analysis, or uniformization-based error bounds [2411.00759][2509.21906].

## 4. Geometry, manifolds, and continuous-state formulations for discrete data

Several approaches model discrete data by embedding probability distributions into continuous geometric spaces. Fisher-Flow treats categorical distributions as points on a statistical manifold equipped with the Fisher-Rao metric,
$$
g_{\mathrm{FR}(p)}[u,v] = \sum_{i=0}^d \frac{u^i v^i}{p^i},
$$
and uses the sphere map \(p \mapsto s=\sqrt{p}\) as an isometry from the simplex interior to the positive orthant of the hypersphere \(\mathbb{S}^d_+\) [2405.14664]. On this hypersphere, flows are defined along closed-form geodesics, can be bootstrapped by Riemannian optimal transport, and the induced gradient flow is proved optimal in reducing the forward KL divergence [2405.14664].

A broader unification is provided by \(\alpha\)-Flow, which presents a unified framework for Continuous-State Discrete Flow Matching models. Its key device is the \(\alpha\)-representation
$$
x=\pi^{(\alpha)}(\mu)=
\begin{cases}
\mu^{\frac{1-\alpha}{2}}, & \alpha \neq 1,\\
\log \mu, & \alpha = 1,
\end{cases}
$$
under which prior CS-DFM variants correspond to different \(\alpha\)-geometries [2504.10283]. The paper states that \(\alpha\)-Flow adheres to the canonical \(\alpha\)-geometry of the statistical manifold, is optimal in minimizing the generalized kinetic energy, and yields a unified variational bound for the discrete negative log-likelihood [2504.10283].

A related construction uses assignment manifolds. “Generative Modeling of Discrete Joint Distributions by E-Geodesic Flow Matching on Assignment Manifolds” defines a continuous normalizing flow on the submanifold of factorizing discrete measures
$$
\mathcal{W} = \mathcal{S}_c \times \cdots \times \mathcal{S}_c,
$$
embedded into the meta-simplex by
$$
T(W)_\alpha = \prod_{i=1}^n W_{i,\alpha_i},
$$
and trains the model by matching geodesic flows of factorizing discrete distributions [2402.07846]. The construction gradually assigns categories and avoids issues of discretizing the latent continuous model such as rounding and sample truncation [2402.07846].

For discrete non-negative ordinal data, Binomial Flows provide another bridge. The forward conditional is a binomial thinning process,
$$
p_{t \mid T}(x_t \mid x_T) := \mathrm{Binomial}_{x_T,\frac{t}{T}}(x_t),
$$
the denoiser is \(m(t,x)=\mathbb{E}[X_T \mid X_t=x]\), and the central identity is a discrete Tweedie’s formula,
$$
\frac{m(t,X_t)-X_t}{T-t}=\lambda(t,X_t),
$$
where \(\lambda\) is the intensity of a Poisson-Föllmer process used for sampling [2605.00360]. The same framework yields exact likelihoods and unbiased log-likelihood estimators for ordinal discrete data [2605.00360].

These geometric constructions share a common premise: discrete generation need not be confined to jump processes on finite sets. A plausible implication is that continuous-state representations can recover analytical tools—geodesics, Riemannian metrics, conditional means, exact likelihood expressions—that are difficult to obtain in purely discrete-state models [2405.14664][2504.10283][2605.00360].

## 5. Hybrid and hierarchical bridges between continuous and discrete flow matching

Recent systems often combine discrete and continuous mechanisms rather than choosing one exclusively. PolyFlow addresses the incompatibility between mesh connectivity and standard continuous denoising by introducing a compact topology embedder that projects discrete mesh topology into continuous per-vertex embeddings; after pretraining and freezing this embedder, a Transformer-based flow-matching model denoises joint vertex states in parallel and decodes topology via spacetime distance thresholding [2606.30673]. The state of each vertex is
$$
\mathbf{z}_i=[\mathbf{p}_i,\mathbf{n}_i,\mathbf{e}_i] \in \mathbb{R}^{3+3+d},
$$
and inference proceeds by solving
$$
\frac{d\mathbf{Z}_t}{dt} = \mathbf{v}_\theta(\mathbf{Z}_t,t,\mathbf{c})
$$
with an Euler ODE solver [2606.30673].

Flow6D uses a two-stage discrete-to-continuous hierarchy for category-level 6D pose estimation. Rotation and translation are first discretized into bins and localized with a discrete flow matching model; a continuous flow matching model then predicts local pose residuals to refine the estimate [2606.23293]. This two-stage discrete latent space localization–continuous pose regression strategy is explicitly designed to reduce search complexity and then optimize accuracy [2606.23293].

Purrception targets vector-quantized image generation by combining explicit categorical supervision with continuous transport dynamics. It learns categorical posteriors over codebook indices while computing velocity fields in the continuous embedding space, with
$$
v_t^\theta(z_t) = \frac{\mu_t(z_t)-z_t}{1-t},
$$
where \(\mu_t(z_t)\) is the posterior barycenter over codebook embeddings [2510.01478]. The paper emphasizes uncertainty quantification over plausible codes and temperature-controlled generation, and reports faster convergence than both continuous flow matching and discrete flow matching baselines on ImageNet-1k \(256 \times 256\) generation [2510.01478].

Branching Flows generalize the bridge further by adding stochastic splits and deletions. They transport a simple distribution to the data distribution while allowing the number of elements in the state to vary over a forest of binary trees, and they compose with any flow matching base process on discrete sets, continuous Euclidean spaces, smooth manifolds, and multimodal product spaces [2511.09465]. This directly targets settings in which the number of elements is not known a priori [2511.09465].

For heterogeneous tabular data, Cascaded Flow Matching adopts a low-resolution discrete stage followed by a high-resolution flow-matching stage. The first stage generates categorical features and coarse categorical versions of numerical features; the second stage then generates detailed numerical values with a guided conditional probability path and data-dependent coupling, and the paper proves that the cascade tightens the transport cost bound [2601.22816].

## 6. Applications, empirical behavior, and open technical issues

The empirical scope of continuous and discrete flow matching is broad. Discrete Flow Matching scaled to 1.7B parameters reaches \(6.7\%\) Pass@1 and \(13.4\%\) Pass@10 on HumanEval, and \(6.7\%\) Pass@1 and \(20.6\%\) Pass@10 on 1-shot MBPP coding benchmarks [2407.15595]. In discrete ordinal modeling, Binomial Flows report a CIFAR-10 FID score of \(2.94\) for 8-bit image generation [2605.00360]. PolyFlow reports improved Chamfer Distance and Hausdorff Distance on Toys4K mesh generation, while Flow6D reports real-time inference at \(70\) FPS for category-level 6D pose estimation [2606.30673][2606.23293].

Reinforcement learning supplies a distinct application class. DRIFT models a discrete-action policy as a CTMC generator, updates an offline pretrained policy with an advantage-weighted discrete flow matching loss, and adds a path-space penalty that regularizes the full CTMC trajectory distribution rather than only the final action distribution [2605.12379]. A separate offline RL framework replaces continuous flows with CTMCs, uses a Q-weighted flow matching objective, extends to multi-agent settings through a factorized conditional path, and shows that under idealized conditions optimizing this objective recovers the optimal policy [2602.06138].

Molecular generation highlights both achievements and limitations. FlowMol-CTMC benchmarks several discrete flow matching methods for 3D de novo molecule generation and reports that CTMC-based flow matching dramatically improves molecular stability and validity relative to continuous surrogates [2411.16644]. The same study introduces metrics beyond local chemical valency constraints and finds that, even though basic constraints are satisfied, the models tend to produce unusual and potentially problematic functional groups outside of the training data distribution [2411.16644]. This provides an objective caution against treating sanitization or valency satisfaction as sufficient evaluation criteria.

Training efficiency and variance reduction remain active issues. Temporal Pair Consistency couples velocity predictions at paired timesteps along the same probability path, operates entirely at the estimator level without modifying the model architecture, probability path, or solver, and is shown theoretically to induce a quadratic, trajectory-coupled regularization that reduces gradient variance while preserving the underlying flow-matching objective [2602.04908]. The method improves FID at identical or lower computational cost than prior methods on CIFAR-10 and ImageNet, and extends to score-based denoising and rectified flow pipelines [2602.04908].

Several open technical issues recur across the literature. Topological mismatch can force discontinuous optimal velocities in continuous flow matching [2512.12821]. In discrete flow matching, stochastic paths prevent a direct transfer of continuous rectification and complicate exact evaluation [2411.00759]. In continuous relaxations of discrete data, assignment lag can degrade coherence, as observed in molecular generation benchmarks [2411.16644]. And for variable-cardinality data, fixed-size flow assumptions require explicit mechanisms such as branching, deletion, or cascaded coarse-to-fine generation [2511.09465][2601.22816]. Collectively, these results indicate that the central design question is not whether a problem is “continuous” or “discrete” in isolation, but which probability path, geometry, stochastic process, and representation make the target transport learnable.

Source: https://www.emergentmind.com/topics/continuous-and-discrete-flow-matching