- Rating
- 1
- Confidence
- 2
- Informativeness
- 1
- Presentation
- 2
- Soundness
- 1
Read review
Paper summary
Generative models are trained on data drawn from a given distribution, yet they are routinely queried outside their training support, where their behaviour is unspecified. This paper proposes a certificate predicting how far outside its support a generative model remains close to its own linear extrapolation; according to the authors, beyond that point the model departs from its in-support behaviour and becomes unpredictable.
In a band near the support boundary, the map is approximated by an atlas of local affine charts. From this atlas, three quantities measured inside the support (the atlas error at the boundary, the affine residual and the curvature), together with a calibrated constant , define an error model quadratic in the distance, whose positive root gives a radius below which the deviation between the map and the atlas should remain under a tolerance . The algebra of the formula is verified in Lean, while the validity of the radius rests on the empirically tested assumption that the error model dominates the true error. The radius is safe on a synthetic testbed. On 2D tabular maps and flow-matching models with a 2D latent, however, it overestimates the true radius on part of the cells, and must be recalibrated for each model class.
Critical strengths
-
Transparent and disciplined experimental protocol. Discovery and fresh cells are kept separate, constants are fixed before validation, and failures are reported rather than tuned away. This is exemplary and it is also what makes a precise assessment possible, since the method's weaknesses can be read directly from the authors’ own tables. However, it does not compensate for the conceptual issues below.
-
A clean falsifiable test of the curvature estimator (section 4.2). The prediction that the certified radius should be invariant to the map’s affine part is tested in a controlled setting and clearly discriminates the raw map-Hessian from the normal-projected second fundamental form. This is the best-supported finding of the paper, although its scope is limited to a single synthetic family.
Critical weaknesses
-
The certificate offers no guarantee, and the quantity it predicts can be computed directly. Soundness (Proposition 2) holds only under the assumption , which is neither proved nor provable without constraints on F outside the support. The authors acknowledge this in Appendix A, yet the title, abstract and Lean formalisation suggest a guarantee the paper does not provide. is therefore an empirical predictor, and must be judged as such against the obvious alternative: the bounded quantity needs no ground-truth data, and the paper itself obtains by simply querying F on outer shells, at a cost comparable to fitting the atlas. However, the bounded quantity needs no ground-truth data: the paper itself obtains by simply querying F on outer shells, at a cost comparable to fitting the atlas. Moreover, the predictor is less reliable than this direct estimate must be calibrated on out-of-support error curves, and with the transferred the certificate overclaims on all three image datasets, with violations remaining on Fashion-MNIST and EMNIST even after in-sample per-class calibration. I cannot see what offers over direct estimation. A comparison in cost and reliability would be required, and the current evidence suggests would lose on both. This is the main reason for my recommendation.
-
The bounded quantity measures regularity, not reliability. The certificate bounds the deviation of F from its own linear extrapolation, not the quality of its outputs. For exemple, a generator that stays on a curved data manifold must deviate from its tangent extrapolation, so a large E may reflect good behaviour. The paper provides no evidence linking to output quality (sample plausibility, distance to the data manifold, out-of-range pixels). The framing as "OOD safety" is therefore not supported, even if the predictor itself were accurate.
Questions for authors
-
Since is obtained by querying F on outer shells, with no ground-truth data, is there a practical setting where F can be queried inside the support but not outside, or where outer queries are substantially more costly? Please report the number of F evaluations required to compute (including atlas fitting and calibration) versus direct estimation of to comparable precision.
-
Calibration of without OOD queries. How would a practitioner set for a new model class without computing out-of-support error curves? If such curves are required, in what sense is the certificate obtained "from in-support geometry alone" ?
-
Do you have evidence that correlates with any measure of output quality as d grows, for example sample plausibility, distance to the data manifold, or fraction of out-of-range pixels?
-
Precision of . On Fashion-MNIST and EMNIST, is a 99th percentile estimated from 60 samples per shell, effectively a maximum. Can you report confidence intervals on , and indicate whether the observed violations (or their absence) are robust to this estimation noise?
Suggestions for future work
-
Determine conditions under which holds : this would turn the predictor into a genuine conditional certificate.
-
A systematic study of the relation between (or ) and the quality of generated outputs as d grows, e.g. sample plausibility, distance to the data manifold, or fraction of out-of-range pixels. This would establish whether the certified quantity is a meaningful proxy for OOD reliability.
Thanks to the authors for their effort on the rebuttal.
I briefly reviewed the response and the revised version, and here are some thoughts.
Now I'm fairly confident in my evaluation of score 1. In my best judgment, the paper's quality is well below the acceptance threshold.