- Rating
- 2
- Confidence
- 2
- Informativeness
- 2
- Presentation
- 2
- Soundness
- 2
Read review
Paper summary
A representation is produced by a privacy mechanism from observation , with sensitive attribute and utility label , under the Markov assumption .
Critical strengths
- The interaction-information separation theorem is correct, non-trivial, and makes the regime a computable diagnostic.
- The finite-support validation is built so the learned mechanism cannot game the metric.
Critical weaknesses
- The headline empirical claim is contradicted by the paper's own Table 2.
- Assumption 1 does not describe the mechanisms the paper's own experiments instantiate.
Questions for authors
- On W1: on Adult, CFB attains at while priLO attains 60.2 at 40.3. Under an identical post-hoc protocol, what reading of "competitive" survives this? If none, please reword the abstract and conclusion.
- On COMPAS, is at or below the majority-class baseline? If a trivial classifier matches it, the paper should say so and attribute the cause among the conditional secondary objective, the critic budget, and the realizability gap.
2 comments
Thanks, the revision settles both things I raised, and my majority-baseline guess was wrong; the training-prior control is the right comparator. I'm leaving the score at 2 though, because the same tables make the contribution thinner than the title suggests: priLO loses to CPFSI by 28 points on Adult and sits below a trivial prior control on COMPAS, and utiLO's edge over tuned CFB is 1.2 points of sensitive F1 at equal utility.
Markov assumption and dataset-specific performance
We thank the reviewer for identifying the overly broad empirical interpretation and for questioning how Assumption 1 applies to the neural implementation. We have revised the main text and added controlled comparisons.
1. Competitiveness on Adult
We agree that the original results do not justify describing priLO as generally competitive.
In the original Adult experiment, priLO obtains 60.2% utility macro-F1 at 40.3% sensitive macro-F1, compared with 76.1% and 41.2% for privacy-oriented CPFSI.
We have revised the abstract, experimental discussion, and conclusions to distinguish priLO from utiLO.
Appendix D.4 provides independently retrained comparisons over fifteen matched seeds. Adult priLO retains only 47.95% utility versus 76.21% for privacy-oriented CPFSI.
utiLO performs differently. It achieves 78.68% utility and 51.98% sensitive macro-F1, compared with 78.76% and 53.19% for high-utility CFB. The matched-bootstrap sensitive macro-F1 difference is −1.21 percentage points, with a 95% interval of [−2.13,−0.25].
Our revised claim is therefore narrower: utiLO remains competitive with strong high-utility baselines, whereas priLO frequently sacrifices substantial utility. We do not claim universal Pareto dominance.
2. COMPAS and trivial baselines
Appendix D.4 now includes majority-class, constant-representation, and training-prior randomized controls.
COMPAS priLO achieves 48.40% utility macro-F1. This exceeds the majority baseline (35.27%), but falls below the training-prior randomized baseline (49.97%). The paired-bootstrap difference is −1.56 percentage points, with a 95% interval of [−2.28,−0.88].
We therefore acknowledge that priLO does not establish nontrivial utility recovery against the stronger control.
Appendices A.8–A.9 distinguish structural endpoint attainability from the actual population optima, while B.1–B.2 address surrogate and optimization limitations.
The experiments do not isolate a unique cause of the failure. We therefore avoid attributing it specifically to secondary optimization, critic capacity, or endpoint unattainability.
3. Markov assumption
Assumption 1 concerns the representation produced by a fixed encoder on a fresh observation, not the labeled data used to train it.
Although S and U supervise MI critics during training, they are not inputs to the deployed encoder. For fixed learned parameters θ and independent sampling noise ε,
\qquad I(S,U;Z\mid X,\boldsymbol\theta)=0.$$ For fresh observations independent of the training procedure, this gives the required Markov structure $(S,U)\to X\to Z$. We have clarified this distinction immediately after Assumption 1.