△ Mechanism confirmed, baseline not beaten 2026

Gradient-Projected Rectified Flow

Implementation & benchmark of arXiv:2608.02487 — Computational and Statistical Guarantees of the \textit{c}-Rectified flow

Usefulness7/10

Difficulty5/10

Novelty7/10

Sampling Accuracy Stability Sample efficiency Diffusion Diffusion sampling Optimizer Optimal transport Convex analysis Dynamical systems Functional analysis

Source paper: Computational and Statistical Guarantees of the \textit{c}-Rectified flow arXiv:2608.02487 ⓘ · analyzed Aug 31, 2026

AI-generated research hypothesis, automatically tested. Not peer-reviewed.

Idea description

Replace the unconstrained rectified-flow velocity predictor with the gradient of a learned scalar potential. At every rectification round, fit the potential by weighted least squares to the current displacement field, then integrate the resulting conservative velocity from the source distribution to the target distribution. The gradient restriction is intended to eliminate non-transport rotational motion and improve convergence toward the quadratic optimal-transport coupling.

Formulas

$$X_{1},\ldots,X_{n}\overset{\mathrm{i.i.d.}}{\sim}P,\qquad Y_{1},\ldots,Y_{n}\overset{\mathrm{i.i.d.}}{\sim}Q.$$

$$Z_t=(1-t)X+tY,\qquad D=Y-X,\qquad v_k(z,t)=\mathbb{E}_{\pi_k}[D\mid Z_t=z].$$

$$\phi_k=\arg\min_{\phi}\;\mathbb{E}_{(X,Y)\sim\pi_k,\,t\sim\mathrm{Unif}[0,1]}\left[\left\|\nabla_z\phi(Z_t,t)-(Y-X)\right\|_2^2\right],\qquad u_k(z,t)=\nabla_z\phi_k(z,t).$$

$$\frac{dZ_t}{dt}=u_k(Z_t,t),\qquad Z_0\sim P,\qquad T_k(x)=Z_1\ \text{when }Z_0=x.$$

Mathematical statement

The transferable construction is the projection of a velocity field onto a gradient class. Let $\pi_k$ be the current coupling of source $X\sim P$ and target $Y\sim Q$, let $Z_t=(1-t)X+tY$ be its displacement interpolation, and let $D=Y-X$ be the displacement. The unconstrained rectified-flow field is the conditional velocity $v_k(z,t)=\mathbb{E}_{\pi_k}[D\mid Z_t=z]$. The c-rectified replacement is a gradient field $u_k(z,t)=\nabla_z\phi_k(z,t)$ obtained by minimizing the weighted Hilbert-space projection objective $\mathbb{E}_{(X,Y)\sim\pi_k,\,t\sim U[0,1]}[\|\nabla_z\phi_k(Z_t,t)-D\|_2^2]$. Here $\phi_k$ is a scalar potential, $z\in\mathbb{R}^d$ is the flow state, and $t\in[0,1]$ is time. The gradient class consists of conservative vector fields, so its restriction removes rotational components. The induced flow solves $\dot Z_t=u_k(Z_t,t)$ with $Z_0\sim P$. For quadratic displacement cost, the target map is $T^*=\arg\min_{T:T_{\#}P=Q}\mathbb{E}_{X\sim P}\|T(X)-X\|_2^2$. In implementation, endpoint preservation is checked explicitly by comparing the generated endpoint distribution with $Q$ and can be enforced through a marginal penalty or a target re-coupling step.

Implementation notes

Integrate this as a replacement for the velocity network in a flow-matching or rectified-flow generator. Use an MLP potential $\phi_\theta(z,t)$ with one scalar output, and obtain the vector field by automatic differentiation with respect to $z$; do not directly output a vector velocity. For each rectification round $k$, construct a minibatch of pairs $(x_i,y_i)$ from the current coupling $\pi_k$, sample $t_i\sim\mathrm{Uniform}[0,1]$, and compute $z_i=(1-t_i)x_i+t_i y_i$ and $d_i=y_i-x_i$. Minimize $L_k=B^{-1}\sum_{i=1}^B\|\nabla_z\phi_\theta(z_i,t_i)-d_i\|_2^2$. Pseudocode: for k: sample (x,y) from pi_k; sample t; z=(1-t)*x+t*y; d=y-x; v=grad_z(phi_theta(z,t)); loss=mean((v-d)^2); update theta; integrate dz/dt=grad_z(phi_theta(z,t)) from x to obtain yhat; update pi_{k+1}. Update $\pi_{k+1}$ by nearest-neighbor matching between generated endpoints $\hat y$ and fresh target samples, or by solving a minibatch entropic Sinkhorn coupling. Add an optional endpoint penalty $\lambda\,\mathrm{MMD}(\{\hat y_i\},\{y_i\})$. The paper-specific computation is the gradient-class projection; the empirical contraction factor, endpoint discrepancy, and transport cost are measured rather than assumed. Begin with two-dimensional rotated Gaussian distributions and Gaussian mixtures, using a four-layer width-128 potential. Compare against ordinary rectified flow and a direct vector-field flow-matching network at equal training steps and ODE function evaluations. Record quadratic coupling cost, endpoint MMD or Wasserstein distance, curl residual $\|\nabla u-(\nabla u)^T\|_F$, and generation quality. Success is lower transport cost and endpoint error after the same number of rectification rounds, particularly for noncommuting rotated covariance matrices, without a substantial increase in inference-time ODE evaluations.

Verification

Mechanism confirmed, baseline not beaten

Stage 1 · Toy mechanism gate: Passed ✓

Stage 2 · Mechanism transferred to benchmark: Confirmed ✓

Stage 2 · Practical benchmark result: Significantly worse ✗

Methodology: 8 paired seeds, permutation test (p<0.05) against a learning-rate-tuned baseline. How verification works

Stage 1 — Mechanism check agent confidence 8/10

Built a readable MVP comparing a direct vector-field flow-matching network with a scalar-potential gradient field. The mechanism checks passed quantitatively: rotational energy removal was 1.000 of the injected rotational energy for alpha=0.25, 0.5, 1, and 2; removed energy scaled exactly as alpha^2; input curl scaled as alpha and projected curl stayed near 1e-15. In the small Gaussian experiment, the potential model also showed lower endpoint MMD, covariance error, and transport cost, though this comparison used one seed and is only a promising toy signal.

Agent confidence: 8/10

Baseline: direct vector field: endpoint MMD 0.002712, mean error 0.1018, covariance error 0.5039, transport cost 3.4934
Idea: gradient potential: endpoint MMD 0.001207, mean error 0.1076, covariance error 0.3510, transport cost 3.1847; math sweep removed-energy ratio 1.000 and removed-energy/alpha^2 1.000

Limitations: The learned comparison is a single fixed-seed, small 2D Gaussian test with 700 updates and no multiple-seed statistics, rectification-round sweep, Sinkhorn re-coupling, explicit learned-field curl measurement, FLOP-normalized timing, or Gaussian-mixture evaluation. The exact math verification uses a periodic spectral projection rather than the finite MLP training procedure.

How to run: python3 experiment.py

Stage 2 — Benchmark vs. tuned baseline (arXiv:2608.02487) Sep 2, 2026

Significantly WORSE than the tuned baseline (+126.1%, p=0.0081, wins 0 / 8 paired seeds).

Benchmark: Periodic_transport
Model: mlp_tiny
Paired seeds: 8
Baseline mean: 0.0532
Idea mean: 0.1204
Effect (Δ): 0.0671 (+126.1%; negative = idea better)
Wins: 0 / 8 paired seeds
p-value: 0.0081 (permutation test, 20 000 shuffles)
Smallest detectable effect: ±2.5%
Mechanism: Confirmed ✓
Practical effect: Hurts
Baseline tuning: swept over 3 configs

Limitations:

The registered task is a periodic PDE one-step forecasting benchmark rather than full rectified-flow training with displacement couplings, ODE integration, rectification rounds, or Sinkhorn re-coupling. The baseline and idea used the same MLP width/depth and matched learning-rate union, but the gradient model has a different output parameterization and higher-order autodiff cost.

How to run: python3 run_registered_bench.py

Verdict computed by deterministic test code from paired-seed statistics — not by the language model.

Stage 2 — Benchmark Aug 31, 2026 latest Failed ✗

Agent confidence: 10/10

Gradient-Projected Rectified Flow was implemented as a scalar-potential gradient predictor and compared end-to-end with a direct vector predictor on the registered periodic_transport bench track. The conservative mechanism prediction held quantitatively: potential antisymmetric-Jacobian RMS was 0.0 versus 0.05253 for the direct model. Nevertheless, the idea lost significantly on standard test MSE: 0.120372 versus 0.053246, paired delta +0.067126 and permutation p=0.0081.

Baseline: mean test MSE 0.0532457046 over 8 seeds; best lr=0.01; sweep means lr=0.001: 0.2965991, lr=0.003: 0.1055090, lr=0.01: 0.0539030
Idea: mean test MSE 0.1203719601 over 8 seeds; best lr=0.01; sweep means lr=0.001: 0.2163147, lr=0.003: 0.1334133, lr=0.01: 0.1210454; paired delta +0.0671262555, p=0.0081

Benchmark result