- Rating
- 3
- Confidence
- 2
- Informativeness
- 3
- Presentation
- 3
- Soundness
- 2
Read review
Paper summary
CCB reformulates stochastic control with prescribed external and pairwise potentials as joint optimization of endpoint coupling and conditional paths. It alternates KL-proximal transport updates with learning Gaussian conditional paths, then regresses velocity and score for PF-ODE or SDE inference. Experiments cover analytic control, observed-marginal reconstruction, and held-out-time interpolation.
Critical strengths
The paper has certain novelties.
Critical weaknesses
-
Proposition 2 concerns exact transport updates with fixed paths and full support; it does not establish convergence of the sparse joint implementation. Monte Carlo costs, regression, and numerical integration introduce additional errors. The analytic control benchmarks have Gaussian optima (Appendix B.3.1), limiting evidence on approximation in general control problems. KS/Coulomb test non-Gaussian interpolation but do not supply an optimal-control reference.
-
Evidence for potential-dependent benefits is limited. Table 2 evaluates observed marginals and lacks CCB without potentials, so it does not isolate interaction benefits. In Coulomb, external-only accounts for most of the W1 gain: 0.217 without potentials, 0.083 external-only, and 0.079 full (Table 3). The spatial prior worsens spatial W1 at t = 3 from 0.2472 to 0.3283 (Table 19), although expression and neighborhood metrics improve. Prior sensitivity and training-seed uncertainty are not reported for the spatial result.
Questions for authors
-
Could you quantify action and marginal errors against a small dense reference, including a non-Gaussian optimal-control problem?
-
Could you add CCB without potentials to Table 2 and evaluate the potential choices at a held-out time?
-
Clarification: How was the spatial penalty strength of 7 selected, and what explains the spatial-W1 deterioration at t = 3?
9 comments
[W2, Q2]
Could you add CCB without potentials to Table 2, evaluate the potential choices at a held-out time, and report the missing spatial training-seed uncertainty?
We thank the reviewer for pointing out the lack of clarity between the experimental purpose and the contribution of potentials. We apologize that the original manuscript did not clearly distinguish generation quality with potentials from the improvement that potentials bring to held-out interpolation. We have rewritten the motivation of Experiment 2, added no-potential CCB, and introduced potential ablations on the HMF Model and held-out-time interpolation experiments using estimated potentials.
1. Experiment 2 tests whether generation quality can be maintained when conditional paths become curved, rather than whether potentials improve generation. Reference [1] shows that straightening flow trajectories enables accurate generation with fewer discretization steps; nearly straight flows can even achieve high generation quality with a single Euler step. Potentials curve the conditional mean paths, so we ask whether CCB can maintain high generation quality when potentials introduce path curvature. Experiment 2 uses attractive and repulsive interactions as controlled settings to evaluate reconstruction of observed marginal distributions; these potentials are not biological priors prescribed for the expression datasets. The table below lists the results from Table 2 without potentials and with attractive or repulsive potentials, measured by averaged over observed times:
| CCB setting / Inference mode | EMT 10D | EB 5D | EB 50D | EB 100D | CITE 50D | Weinreb 50D |
|---|---|---|---|---|---|---|
| Without potentials / PF-ODE | 0.180 | 1.483 | 7.638 | 10.979 | 19.904 | 5.141 |
| Without potentials / SDE | 0.193 ± 0.0009 | 1.420 ± 0.0073 | 7.831 ± 0.0127 | 11.127 ± 0.0154 | 20.284 ± 0.0133 | 5.376 ± 0.0247 |
| Attractive potential / PF-ODE | 0.178 | 1.550 | 7.852 | 11.049 | 21.925 | 5.171 |
| Attractive potential / SDE | 0.193 ± 0.0008 | 1.452 ± 0.0150 | 7.723 ± 0.0170 | 11.283 ± 0.0110 | 22.342 ± 0.0320 | 5.409 ± 0.0075 |
| Repulsive potential / PF-ODE | 0.182 | 1.547 | 7.799 | 10.956 | 21.964 | 5.166 |
| Repulsive potential / SDE | 0.197 ± 0.0018 | 1.446 ± 0.0140 | 7.758 ± 0.0180 | 10.842 ± 0.0270 | 22.031 ± 0.0280 | 5.339 ± 0.0140 |
The results show that CCB maintains marginal reconstruction quality close to its no-potential version even when potentials curve the conditional paths.
2. The newly added HMF Model experiment shows that interaction potentials alone can substantially improve held-out interpolation.
First, the small improvement in empirical on Coulomb should be interpreted alongside cluster-scale recovery. Figure 2 shows that the external potential primarily constrains cluster locations, while the interaction potential controls their spread; removing the interaction potential produces overly contracted clusters. With finite samples in high dimensions, even two point clouds drawn from the same distribution can have a large matching distance, and a contracted point cloud may instead obtain a lower empirical . To test this effect, we independently draw 512 points each from and , compute empirical , and repeat the experiment 20 times:
| Covariance ratio | Empirical , mean ± standard deviation |
|---|---|
| 1.0 | 4.016 ± 0.022 |
| 0.5 | 3.614 ± 0.024 |
| 0.3 | 3.575 ± 0.027 |
The population is zero at , but empirical at is lower in all 20 repetitions. This shows that empirical with finite samples in high dimensions can favor overly contracted distributions, thereby underestimating the benefit of interaction potentials in recovering the correct cluster scale.
To further directly assess the contribution of interaction potentials, we have added a 20D HMF Model dataset. Its particles evolve under external and pairwise interaction potentials according to second-order dynamics. All methods are trained using only the position distributions at and and evaluated at , which are not used in training. The data construction is described in Appendix B.3.2, and the following results have been added to Table 3.
| Method / Inference mode | ||
|---|---|---|
| OT-CFM | 0.506 ± 0.0012 | 3.972 ± 0.017 |
| WFR-FM | 0.510 ± 0.0011 | 3.979 ± 0.010 |
| SF²M | 0.520 ± 0.0024 | 3.961 ± 0.037 |
| VGFM | 0.958 ± 0.0062 | 10.310 ± 0.090 |
| DeepRUOT | 1.433 ± 0.0293 | 14.793 ± 0.185 |
| WLF (OT) | 0.578 ± 0.0085 | 3.606 ± 0.117 |
| Modified WLF (with potentials) | 0.825 ± 0.1010 | 2.374 ± 1.026 |
| Modified CytoBridge (with potentials) | 1.114 ± 0.0052 | 12.743 ± 0.257 |
| GSBM (without potentials) | 1.304 ± 0.0209 | 6.292 ± 0.394 |
| GSBM (with potentials) | 1.251 ± 0.0157 | 3.731 ± 0.092 |
| CCB (without potentials, PF-ODE) | 0.506 ± 0.0024 | 3.966 ± 0.035 |
| CCB (without potentials, SDE) | 0.523 ± 0.00014 | 3.996 ± 0.004 |
| CCB (external potential only, PF-ODE) | 0.471 ± 0.0027 | 3.433 ± 0.038 |
| CCB (interaction potential only, PF-ODE) | 0.063 ± 0.0033 | 0.040 ± 0.006 |
| CCB (full model, PF-ODE) | 0.049 ± 0.0028 | 0.011 ± 0.004 |
| CCB (full model, SDE) | 0.127 ± 0.0009 | 0.012 ± 0.005 |
Full CCB outperforms all comparison methods on both metrics. More importantly, the ablation isolates the contribution of the interaction potential: adding only the external potential reduces from to , whereas adding only the interaction potential reduces to , a reduction of approximately 87.5%, and from to , a reduction of approximately 99.0%. Adding the external potential as well further reduces the errors to and . Thus, the interaction potential is the main source of interpolation improvement on the HMF Model. This complements the Coulomb results, where the external potential contributes more, and demonstrates the roles of the two potentials in different dynamics.
We thank the reviewer for the constructive comments. We address all of your questions below, with clarifications and additional experiments in the revised manuscript.
[W1, Q1]
We thank the reviewer for suggesting a comparison of the Dense and Sparse implementations and quantification of Action and marginal errors on a non-Gaussian optimal-control problem. We have added Dense results on the GM10D benchmarks to Section 4.1 and Table 1, with the ground-truth construction and coupling and conditional-path error analyses in Appendices B.3.1 and D.1. Below, we directly compare the Dense and Sparse implementations on the same GM10D benchmarks.
1. Both CCB Dense and Sparse achieve high accuracy on non-Gaussian optimal-control problems. GM10D has three Gaussian clusters in the initial distribution and four in the terminal distribution, covering linear and quadratic external potentials and quadratic attractive and repulsive interactions. The data construction and computation of the optimal control law are described in Appendix B.3.1. The table below compares CCB Dense, CCB Sparse, and other methods on the same problems, reporting Action MAE, control-law MSE, and terminal W1.
CCB Dense achieves lower Action MAE and control-law MSE than all comparison methods under all four potentials. Sparse achieves slightly lower Action MAE than Dense in every case. Its control-law MSE increases slightly but remains substantially below the other methods, preserving Dense's accuracy advantage on multimodal non-Gaussian problems and maintaining accurate recovery of the optimal control law while improving efficiency.
2. The appendix further quantifies reconstruction errors in the endpoint coupling and conditional paths separately. Appendix D.1, Figures 6–7 compare CCB's learned endpoint coupling and conditional bridges with the GM10D ground truth under all four potentials.
Endpoint coupling error. Figure 6 shows mass allocation from three source peaks to four target peaks. Each heatmap cell corresponds to a source–target peak pair, and errors are reported in percentage points of total mass. Under all four potentials, the absolute mass-allocation error for every pair is less than one percentage point.
Conditional-path error. For each potential, Figure 7 uses the same 100 endpoint pairs drawn from the final CCB coupling to compare the means and standard deviations of CCB and ground-truth conditional bridges. Except for the conditional mean under quadratic attraction, the absolute values of all displayed conditional mean and standard deviation errors are below 1%.