- Rating
- 2
- Confidence
- 3
- Informativeness
- 2
- Presentation
- 3
- Soundness
- 2
Read review
Paper summary
In this paper, the authors introduce Diverging Flows, a training objective that preserves near-optimal transport for on-manifold conditions while making transport deliberately inefficient for off-manifold ones. Compared with the classical FM, an FM model can be trained to exhibit a sharp geometric transition, where valid conditions yield near optimal paths while off-manifold queries yield strongly curved ones. This paper is interesting, the thoeretical theorem and numerical implementations need further improvement.
Critical strengths
- This paper is well writen, which is easy to follow.
- The theoretical results are interesting, which inspires the model formulation.
Critical weaknesses
-
Strong requirements about the paths for Proposition 4.1. In Appendix D.1 Equation (8), does this formula hold? If so , we need to ensure the same path for different conditions and .
-
In Proposition 4.1, the gradient Lipchitz continuity is very strong.
-
The comparisons in numerical results are not sufficient.
Questions for authors
-
Strong requirements about the paths for Proposition 4.1. In Appendix D.1 Equation (8), does this formula hold? If so , we need to ensure the same path for different conditions and . Further explanations are needed.
-
In Proposition 4.1, the gradient Lipchitz continuity is very strong, which is widely-used in convex optimization. Could you give some explanations?
-
The comparisons in numerical results are not sufficient. The comparisons mainly focus on likelihood-based and ensemble-based methods. It is better to add more comparisons, e.g., training a separator classifier.
We thank the reviewer for finding the paper well written and the theoretical results interesting. Since W1–W3 coincide with Q1–Q3, we answer them together.
The equality holds, and both energies are taken on the same path. The path xt=(1−t)x0+tx1 and its velocity utOT=x1−x0 (Section 4.1, before Eq. 2) depend on the condition only through the pairing of x1∼pdata(⋅∣c) with c. An off-manifold c~ has no data of its own, so its energy is evaluated on the reference path of the nearby valid c, and only the condition argument of vθ changes. The margin losses of Section 4.2 are defined this way, and each boundary sample keeps the pair (x1,c) of the valid c it comes from (Section 4.3). With μc the law of (t,x0,x1) and fc′=∥vtθ(xt,c′)−utOT∥22, linearity under μc gives ∣E(c~)−E(c)∣≤Eμc∣fc~−fc∣≤2K(M+U)∥c~−c∥2 (App. D.1). The DOT score needs no reference path, since it compares each trajectory with the straight line between its own endpoints (Section 4.4). The revised PDF states the reference measure after Eq. 2 and in App. D.1.
Proposition 4.1 does not assume a Lipschitz gradient. Its statement only requires the vector field to be "locally Lipschitz continuous with respect to the conditioning variable c" (lines 182–184 of the submission), that is, ∥vtθ(x,c~)−vtθ(x,c)∥2≤K∥c~−c∥2. The gradient-Lipschitz (L-smooth) assumption of convex optimization constrains the gradient of an objective, and no gradient of vθ appears in Proposition 4.1 or its proof (App. D.1). As stated before the proposition (Section 4.1), FM already needs this regularity in x for a unique ODE solution [3, 4], and standard architectures satisfy it as compositions of locally Lipschitz layers [1, 2], for any size of K. The proposition explains silent extrapolation, which Diverging Flows counteracts by inducing a locally larger Lipschitz constant around M (Section 4.2).
Diverging Flows needs only the predictor, yet it outperforms the separate classifier the reviewer suggests (new in the revised PDF) and matches detectors that need their own model or the training data. The submission already compares with a classification head (App. F.1), DiffPath [5] (App. L.1) and the acceleration proxy of FM [6] (App. L.2). We add a separate binary classifier, trained with BCE on the same boundary samples, deep KNN [7] on the raw conditions, and VOS [8], which needs class labels and so runs only on MNIST. We report AUROC, and ★ marks results new in the revised PDF (App. M).
Trained on the same boundary samples, the separate classifier and the classification head stay well below Diverging Flows, so these samples alone do not explain the detection. Diverging Flows is above deep KNN and DiffPath-1D on average and close to DiffPath-6D. On KMNIST, deep KNN and VOS are stronger, but they need the training set or class labels, while Diverging Flows reads its score from the trajectory of the prediction itself, at ~0.8% overhead (Section 4.4).
We hope these clarifications and baselines lead the reviewer to reconsider the score.
References
[1] Virmaux and Scaman. Lipschitz Regularity of Deep Neural Networks. NeurIPS 2018
[2] Kim et al. The Lipschitz Constant of Self-Attention. ICML 2021
[3] Coddington and Levinson. Theory of Ordinary Differential Equations. 1955
[4] Lipman et al. Flow Matching Guide and Code. arXiv 2024
[5] Heng et al. Out-of-Distribution Detection with a Single Unconditional Diffusion Model. NeurIPS 2024
[6] Rao et al. The Geometry of Flow-Matching Uncertainty. arXiv 2026
[7] Sun et al. Out-of-Distribution Detection with Deep Nearest Neighbors. ICML 2022
[8] Du et al. VOS: Learning What You Don't Know by Virtual Outlier Synthesis. ICLR 2022