- Rating
- 2
- Confidence
- 3
- Informativeness
- 2
- Presentation
- 2
- Soundness
- 2
Read review
Paper summary
This paper proposes a new flow-based generative framework based on Y-shaped flows. The idea is to make samples travel together along shared pathways before branching off to target-specific endpoints. The formulation seems to be justified, and the proposed method is claimed to be practical and scalable. Experiments on various datasets are provided to demonstrate the effectiveness of the proposed method.
Critical strengths
- Overall the paper is easy to follow and the proposed method is well motivated
- Sufficient background and related work are provided
Critical weaknesses
Main weaknesses:
- Although the idea is interesting, practical implementation seems computationally costly since it is not simulation-free (as in flow matching) and a Neural ODE needs to be integrated accurately for each mini-batch during training
- Detailed algorithms and runtime complexity are not provided to better assess the completeness and computationally efficiency of the method. Although wall-clock cost is mentioned briefly, there are no detailed comparisons with existing methods
- The presentation needs to be improved to enhance clarity on the method proposed. The practical method proposed is essentially terminal distributional matching (with the Sinkhorn divergence loss) plus minimizing some expected cost functional (Dirichlet energy and sublinear transport cost). The paper should have highlighted this from the very beginning and then only justify its constructions via the branched transport perspective
- It is not clear when one should use the proposed method over the existing flow based frameworks which are simpler and easier to train
- There are no empirical comparisons with the OT-CFM of Tong et. al. (going beyond the independent couplings). It is not clear if the "CFM" in Table 1-2 refers to OT-CFM or the one with independent coupling (I-CFM)
Minor comments:
- In line 74-77, it is written that "The pushforward density evolving under this dynamics necessarily satisfies continuity equation with the parameterized velocity field :" but there is no in Eq. (1)
Questions for authors
- Why are we restricting to compact and convex domain ? This seems too restrictive and does not cover commonly used distributions such as Gaussian
- How would the results compare to the ones obtained by minimizing (9) with and ? This is a basic ablation study that should be included (I could not find these in Appendix I: they are only ablating on the transport exponent and cohesion weight)
5 comments
Part 2/3
W4. When to prefer Y-Flows
It is not clear when one should use the proposed method over the existing flow based frameworks which are simpler and easier to train
We recommend Y-Flows when few-step sampling or coherent paths justify extra training cost; FM offers simpler, cheaper training. FFHQ latent FD favors YF-KD at 1–2 steps and OT-CFM at ≥4 (W5).
We measure local-mass cost (rewarding spatially bundled motion) on TEDSim and experimental LARRY: mouse hematopoietic differentiation profiled by single-cell RNA sequencing and heritable lineage barcodes (Weinreb et al., 2020). We fit published monocyte/neutrophil day-2-to-day-6 transport, withholding day-4 observations and lineage labels.
| Local-mass cost ↓ | FM | Y-Flows | Reduction |
|---|---|---|---|
| TEDSim | 43.73 ± 5.63 | 29.18 ± 0.21 | 33.3% |
| LARRY | 46.03 ± 3.63 | 32.61 ± 0.54 | 29.1% |
The cost is , with . We use (including self), fixing before evaluation at 0.2 times median pairwise distance among 256 source/256 observed-target cells.
This supports lower local-mass cost here, not classical BOT optimality or developmental-path recovery.
W5. Independent CFM and the OT-CFM comparison
There are no empirical comparisons with the OT-CFM of Tong et. al. (going beyond the independent couplings). It is not clear if the "CFM" in Table 1-2 refers to OT-CFM or the one with independent coupling (I-CFM)
We compare FM, OT-CFM, and YF-KD on 512D FFHQ female-to-male latents using matched generators and 100,000 training updates. OT-CFM uses linear paths from exact quadratic OT on equal-size uniform minibatches; YF-KD retains action and cohesion (training ). These new experiments are separate from Table 4.
At one/two Euler steps, YF-KD lowers mean FD by 78.1%/52.0% versus OT-CFM and 92.3%/88.4% versus FM.
Held-out latent Fréchet distance (FD ↓; mean ± sample SD, three seeds):
| Method | 1 step | 2 steps | 4 steps | 8 steps | 16 steps | 32 steps | 64 steps | 128 steps |
|---|---|---|---|---|---|---|---|---|
| FM | 121.69 ± 2.33 | 65.38 ± 2.24 | 27.90 ± 2.73 | 12.43 ± 2.52 | 8.36 ± 2.46 | 7.65 ± 2.38 | 7.64 ± 2.35 | 7.74 ± 2.35 |
| OT-CFM | 42.82 ± 0.19 | 15.83 ± 0.19 | 6.49 ± 0.15 | 4.71 ± 0.45 | 4.72 ± 0.60 | 4.99 ± 0.70 | 5.19 ± 0.75 | 5.31 ± 0.78 |
| YF-KD | 9.38 ± 0.47 | 7.60 ± 1.01 | 7.39 ± 1.07 | 7.36 ± 1.07 | 7.36 ± 1.07 | 7.36 ± 1.07 | 7.37 ± 1.07 | 7.37 ± 1.07 |
FD uses all 5,572 source and 4,351 target held-out latents in raw ALAE coordinates.
At four steps, YF-KD achieves endpoint RMS 0.0461 ± 0.0011, 75.9% below OT-CFM and 88.3% below FM. Evaluation uses 256 shared sources, target-centered RMS normalization, and model-specific RK4 references checked by two successive refinements.
Part 3/3
W6. Velocity notation in the continuity equation
The pushforward density evolving under this dynamics necessarily satisfies continuity equation with the parameterized velocity field
In Eq. (1), denotes the learned velocity , so the continuity equation is
This is the continuity equation for the parameterized dynamics used in training.
Answers for questions
Q1. Compactness, convexity, and Gaussian sources
Why are we restricting to compact and convex domain
This seems too restrictive and does not cover commonly used distributions such as Gaussian
Our training algorithm directly supports Gaussian sources in . It evaluates the objective from sampled trajectories without imposing compact support. Compactness and convexity specify the setting of the stated theorem; convexity concerns the ambient domain and permits multimodal distributions.
Extending the separation theorem to would require finite moments, sufficient flow regularity and integrability, the stated speed bound, and a uniform weighted Poincaré inequality along the transported distributions. The practical algorithm already accommodates Gaussian sampling.
Q2. Controls without the proposed trajectory bias
This is a basic ablation study that should be included (I could not find these in Appendix I: they are only ablating on the transport exponent and cohesion weight)
Our matched ablation demonstrates cohesion's contribution to accurate few-step transport. On 30D LARRY, adding cohesion at fixed reduces four-step numerical endpoint RMS by 94.0% and four-step target by 64.9%. Both metrics improve in all three seeds:
| Method | Numerical endpoint RMS ↓ | Target ↓ |
|---|---|---|
| Full Y-Flows | 0.0820 ± 0.0051 | 0.4888 ± 0.0138 |
| Y-Flows without cohesion | 1.3555 ± 0.3808 | 1.3938 ± 0.3425 |
Both models use training , identical backbone/batch size, and 2,000 updates (mean ± sample SD, three seeds). Numerical RMS uses model-specific convergence-checked RK4 references; target compares four-step generated endpoints with fixed observed targets. Both normalize by target-centered RMS. This comparison isolates cohesion's benefit at the same sampling budget.
The FFHQ ablations also show strong few-step performance with linear action and cohesion. Appendix I, Table 8 reports single-seed latent FD for zero-cohesion and linear-action variants:
| Configuration | 1 step | 2 steps | 4 steps | 8 steps |
|---|---|---|---|---|
| 1607 | 201.7 | 10.31 | 4.70 | |
| 1131 | 111.4 | 6.19 | 5.04 | |
| 7.63 | 4.31 | 4.31 | 4.34 | |
| 4.39 | 4.24 | 4.23 | 4.23 |
At , FD is already 4.39 at one step and 4.24 at two, close to 4.23 at four/eight steps. Thus, our objective retains strong few-step quality even with linear action; sublinear action supplies the additional bias toward temporally concentrated motion.
Part 1/3
Dear Reviewer, thank you for your review.
Answers for weaknesses
W1. Training cost and practical scalability
Y-Flows costs more per update than FM but can improve endpoint quality at matched training time. We differentiate through a short, fixed Euler rollout; training requires no adaptive solve to a prescribed accuracy.
TEDSim, 100-second active training budget. Methods share a 222,334-parameter network, batch size 256, float32, one RTX A6000, and 30D PCA coordinates fitted to training endpoints:
Values are mean ± sample SD over three seeds; K denotes training rollout steps and columns denote Euler sampling steps. Exact empirical W1/W2 use fixed 256 source and 512 target cells. Active time includes complete synchronized updates and CPU overhead, excluding setup, warm-up, logging, saves, and evaluation. Checkpoints follow the first completed update at or beyond 100 seconds.
At K=4, four-step W2 improves by 15.8%. Peak allocated memory is 2.40×/6.65× FM for K=4/16, including resident data, model, and training tensors but excluding unused allocator cache and CUDA overhead.
W2. Algorithm and complexity
Section 5's algorithm performs four operations per update:
With P probes, the rollout uses K batch velocity evaluations and KP vector–Jacobian products, then differentiates through them without forming the full Jacobian. Dense Sinkhorn costs O(B2d+SB2) work and O(B2) pairwise storage for S iterations in dimension d. Retained activation memory grows with K. YF-KD replaces Sinkhorn with critic evaluations, updates, and gradient penalties.
FM regresses velocity at sampled times without a rollout; exact minibatch OT-CFM also solves a batch assignment. Appendix I, Table 10's 2 ms per one-probe Jacobian call (20 ms over ten steps) measures only that component. W1 gives a matched-time TEDSim comparison.
W3. Presenting the practical objective directly
The practical objective is terminal distribution matching + sublinear velocity action + density-weighted Dirichlet regularization (Section 5). These terms respectively fit the target, favor temporally concentrated motion, and encourage nearby particles to share velocities.
Branched transport motivates shared motion, but the objectives differ: our action is linear in transported mass, whereas classical branched transport penalizes mass concavely (Appendix D). Cohesion provides a tractable bias toward coherent velocities in place of that mechanism.