- Rating
- 3
- Confidence
- 2
- Informativeness
- 2
- Presentation
- 2
- Soundness
- 3
Read review
Paper summary
This paper studies why MCR2 tolerates dispersed label noise and when concentrated errors weaken its class structure. It proposes Elastic Coding, which reduces class compression when its tangent gradient opposes global expansion. Experiments cover synthetic and real-world label noise.
Critical strengths
- The spectral analysis gives a clear explanation of why noise structure matters beyond the corruption rate.
- The paper carefully separates fixed-feature results from training guarantees, and representation quality from class-name prediction.
- Elastic Coding improves over MCR2 in the real-noise experiments. The common centroid readout and linear probes make the comparison more informative.
Critical weaknesses
- Does the adaptive rule improve over simply reducing compression?
- What links the opposing-gradient ratio to the clean-dominance gap?
- The numerical values of tau, beta, lambda_min, delta and coding precision epsilon are not specified, including in Appendix D. How were these selected without clean labels?
- Some baselines also perform unusually poorly: AEL reaches 12.35% on CIFAR-20 at 20% noise, versus 65.16% for CE. Please explain the baseline tuning protocol and report mean/std over multiple model and corruption seeds before drawing firm ranking conclusions.
Questions for authors
- Does a fixed compression weight of 0.5 give similar results?
- Does gradient conflict track loss of clean-subspace dominance?
- How are the threshold and compression floor selected?
- Are the gains stable across model and noise seeds?
3 comments
Response to Reviewer 7J7Y, Part 2
Weakness 3 & Question 3: The numerical values and selection procedure for , , , , and coding precision are missing. How were these selected without clean labels?
Thank you for highlighting these missing details. This concern overlaps with Reviewer btWb's Weakness 3 and Reviewer eGrH's Weakness 3, to which we kindly refer you for further implementation details.
Specifically, we have now explicitly documented all settings in Section 2.3 and Appendix E.1:
where and is the number of classes. These settings are shared across all datasets and noise rates, without dataset-specific or noise-rate-specific tuning. The EMA is initialised with the first measured conflict ratio, and the compression weight is clipped to and held fixed during each encoder update.
Importantly, Elastic Coding does not require clean labels to compute the conflict statistic or update the compression weight. Clean labels are used only for explicitly identified representation diagnostics. However, this does not establish how the original controller settings were selected.
Regarding hyperparameter selection, we generally adopted the settings recommended in the original papers for baseline methods, with exact variants and coefficients documented in Appendix E.3. All baseline and Elastic Coding settings were kept fixed throughout the additional multi-seed experiments.
Weakness 4 & Question 4: Some baselines perform unusually poorly, particularly AEL. The baseline tuning protocol should be clarified, and results across multiple model and corruption seeds are needed before drawing firm ranking conclusions.
Thank you for raising this concern. It overlaps with Reviewer eGrH’s Weakness 3, and we invite you to consult our response there for the detailed protocol and implementation checks.
To be specific, for the baselines, we generally adopted the hyperparameters recommended in their original papers. Appendix E.3 now specifies the exact loss variants and coefficients, which were held fixed for the additional seeds. We also checked loss values and gradients against pinned reference implementations on 24 fixed-input configurations; no discrepancies were found at the tested precision. These checks do not establish optimal tuning or reproduce each method’s complete original training recipe.
Revised Table 2 reports means and sample standard deviations over five seeds (42–46) for every method–setting combination. Synthetic experiments vary paired model and corruption seeds; real-noise experiments vary training randomness while retaining the observed labels and fixed data subsets.
AEL’s poor performance persists across seeds: on CIFAR-20 at % noise, it achieves , compared with CE’s . Its cause remains unresolved, and we explicitly acknowledge that the disclosed training configuration may not suit every baseline equally well.
Across the revised comparisons, Elastic exceeds MCR² in mean accuracy in all six settings and achieves the highest mean in four of six; CE leads in both Imagenette settings. We therefore frame these results as comparisons under the disclosed configurations, without claiming uniform superiority, or reproduction of each baseline’s best published performance.
Response
Thank you for adding the fixed-weight comparison and multi-seed experiments. These strengthen the empirical evidence, though the advantage over a simple weight schedule, the connection between gradient conflict and the spectral analysis, and the reasons for the unusually weak baseline results remain unclear. I appreciate the authors’ acknowledgement of these limitations and maintain my weak acceptance rating.
Response to Reviewer 7J7Y, Part 1
We sincerely thank you for your careful review and positive assessment of our work. We particularly appreciate your recognition of the spectral explanation of noise structure, our distinction between fixed-feature results and training guarantees, and the improvements over MCR² under real-world label noise. We address each of your concerns in turn, with minor rewording of your comments to ensure a clear and logical flow while preserving their original meaning.
Weakness 1 & Question 1: It is unclear whether the adaptive rule improves over simply reducing the compression weight, particularly to a fixed value of 0.5.
Weakness 2 & Question 2: The relationship between the opposing-gradient ratio and the clean-dominance gap is unclear.