- Rating
- 3(+1)
- Confidence
- 2
- Informativeness
- 2
- Presentation
- 2
- Soundness
- 3
Read review
Paper summary
This paper formalize the mapping from data samples to latent codes/weights in HN-INR using the Implicit Function Theorem (IFT), derive the local non-degeneracy conditions, and interpret regularization and reconstruction error as two operational levers for controlling the condition number of this mapping. The authors further validate structure preservation and downstream utility with synthetic Swiss roll experiments and 2D/3D classification benchmarks.
Critical strengths
- The presentation is clear.
- The theoretical modeling is clear. The authors formulate the training of HN-INR as an implicit function, and connect local embeddings with downstream tasks via the non-degeneracy of . The “trivial intersection of subspaces” in Condition 2 admits a geometric interpretation.
- The synthetic experiments support the arguments. With known ground-truth structure on the Swiss roll dataset, the authors verify latent recovery and neighborhood preservation (kNN Jaccard index = 0.732). And classification accuracy reaches 91.0%, compared with 93.2% using ground-truth representations. Such synthetic benchmarks provide stronger evidence for the theory than real-world benchmarks.
Critical weaknesses
- The classification experiments cannot strongly establish the causal claim that downstream gains arise from structure preservation induced by IFT. Baseline results are taken from original papers, where training protocols, classifier capacity and data augmentation may differ. The proposed method employs a classifier with 11.5k / 2.15M parameters, while HMoE uses a Transformer+MoE architecture. The comparison is not performed under matched capacity and compute budgets. The inconsistency between /kNN metrics and accuracy in Table 4 also suggests that the structure-preservation metric alone does not suffice to explain performance. I think that the authors should provide ablations under identical protocols and matched model capacity, or mediation analysis to examine whether classification performance degrades when controlling for .
- The branch assumption in Proposition 1 is not empirically validated. The globally smooth extension requires that the fitting codes of the same class lie within the same connected component / branch. In practice, it remains unverified whether latents from each class reside in a single branch. Classification performance may stem from clustering across distinct branches rather than a single smooth mapping. Reporting per-class branch consistency would provide direct support for the proposition.
Questions for authors
Please refer to weaknesses.
6 comments
W2: We acknowledge that our wording in Section 4 was imprecise and caused confusion regarding the branch assumption. We have revised Section 4 and Proposition 1 in the revised version (see also our response to Reviewer 7K6H, Q3/W3):
- Local sub-components vs. label classes: In , calling each a "class" was misleading. Each denotes a local connected semantic sub-component (), and a single downstream label class can naturally span multiple disjoint branches. Proposition 1 does not require fitted latent codes of an entire label class to reside in a single branch. It simply extends the point-wise IFT to each local connected region so that nearby samples map smoothly without shattering.
- Clustering across branches and non-collapse: Because one label class may span multiple local branches, downstream classification naturally groups distinct branches of the same class, which is fully consistent with Proposition 1. Meanwhile, separation between distinct branches and classes is guaranteed by our global bi-Lipschitz lower bound in Appendix D: samples differing by more than the reconstruction cannot collapse into the same latent code.
Thanks for the authors' response. My concerns on the experiments are still unresolved, so I will keep my score.
Thank you for recognizing our clear presentation and theoretical modeling.
W1:
W1.1: We agree that the connection between the Implicit Function Theorem (IFT) and our empirical results needs to be strengthened. To this end, we added two theoretical extensions in the revised Appendix. First, we establish a bi-Lipschitz bound (Appendix D): σmin>0 provides an upper bound that keeps nearby samples close, while the reconstruction MSE gives a lower bound that prevents distant samples from collapsing. Second, we derive an explicit O(MSE) bound on the residual curvature perturbation (Appendix E), which guarantees that the Hessian remains positive-definite under imperfect reconstruction. Our IFT framework is not an architecture-level modification on top of baselines to gain a few points. Instead of relying on permutation-invariant architectures, we provide an alternative mathematical instrument to analyze how data structure is preserved in weight space. Our lightweight architecture and small classifier achieve competitive results, which empirically validates that our IFT analysis leads to a well-structured weight representation.
W1.2: In our response to Reviewer vCtd (Q2), we summarize the parameter counts and computational costs across methods. Because of our different perspective on mathematical formulation, we do not need a large permutation-invariant classifier. This is our core message to the WSL community: analyzing the data-to-weight relationship via the IFT provides a complementary view to constructing weight quotient spaces. Regarding HMoE on CIFAR-10, the gap stems from manifold complexity and supervision. Natural images have a much higher intrinsic manifold dimension m∗, which by Lemma 1 (l≥m∗) demands a larger latent space and denser sampling. Moreover, HMoE use meta-learning with a large Mixture-of-Experts (MoE) Transformer architecture, and they include a classification term in their training loss that directly shapes weights with class labels. In contrast, our representations are learned purely via self-supervised reconstruction loss. This explains their stronger performance on CIFAR-10. However, once we add data augmentation (sample the manifold more densely), our method achieves comparable results.
W1.3: We acknowledge that our notation in Table 4 was incomplete: the reported κ is the worst-case (maximum) value across samples, reflecting a single outlier. We reported this worst-case condition number to verify that the Jacobian of the implicit function D1ξ(z,X) is full rank and satisfies the IFT condition for every sample in the dataset. Because a single worst-case outlier does not represent overall classification performance, Table 4 is not ideal for inferring any accuracy trend. Theoretically, from the closed form of Dg derived in Appendix D, local structural preservation and downstream accuracy depend on two factors. First, full rank (σmin(D1ξ(z,X))>0) is a binary condition ensuring that g locally exists. Second, the reconstruction error MSE directly controls sample separation in the bi-Lipschitz lower bound, while the latent dimension l of J~⊤∈Rl×nc limits reconstruction capacity when l is too small. Thus, the IFT only requires D1ξ(z,X) to be full rank, and κ serves only as a numerical check against singularity. Empirically, we conducted an ablation study over l and λ on MNIST (Appendix F). When l is too small, κ is small, but l cannot cover the intrinsic manifold dimension m∗, leaving insufficient structure for classification. When l is too large, reconstruction improves, but κ grows and the full-rank condition becomes less robust, causing accuracy to drop. Across a wide range of λ (from 0 to 0.3), κ varies substantially while accuracy remains nearly unchanged (97.5 to 97.7), confirming our theoretical analysis.