- Rating
- 3(+1)
- Confidence
- 2
- Informativeness
- 2
- Presentation
- 2
- Soundness
- 2
Read review
Paper summary
This paper proposes a novel unlearning auditing method to evaluate whether machine unlearning methods achieve subject-level unlearning, made the model unlearns information about a specific subject. The core concept involves pairing each checkpoint with three split-mathced deletion references as benchmarks for the audit. The paper performs large-scale experiments, a comprehensive review of related work and a clearly structured description of the proposed method.
Critical strengths
- The experiments are thorough, and the results are reported in detail, providing sufficient data to support the conclusions.
- The review of related work is comprehensive, clearly situating the present study within the context of existing research.
- The methodology section is well-structured, with each paragraph serving a clear purpose, making it easy for readers to follow the author’s line of reasoning.
Critical weaknesses
- Existing approaches, such as RWKU and LUCID, also focus on subject-level unlearning audits. However, the paper does not provide comparison between the proposed audit method and these existing ones, making it difficult to substantiate the advantages of the new approach.
- This method requires three ideal retraining models that have never seen the data to be unlearned as references for each checkpoint. This not only renders the evaluation process prohibitively expensive but also creates an excessive reliance on specific datasets. Furthermore, if such ideal reference models were readily available, the necessity for performing unlearning would be called into question.
- The meaning of Figure 1 is unclear (while the subject profile on the left is visible, the process flow on the right is difficult to interpret). In Figure 2, the reading order for the lower section is counterintuitive. While the experimental plots (e.g., Figure 5) are aesthetical, they lack clarity in conveying information.
- In sections such as 5.3, the paper lacks an interpretation of the underlying meaning of the reported figure. For instance, what is specifically meant by "behaviorally meaningful," and how is the term "meaningful" defined? Readers are likely to become overwhelmed by the raw numerical data, struggling to grasp its actual implications.
Questions for authors
- Have previous evaluation efforts not considered subject-level auditing? What are the essential differences and advantages of your audit method compared with RWKU, LUCID and others that also focus on subject-level unlearning audits? Why are there no direct comparative experiments?
- The audit requires three split-matched deletion references. What is the cost of obtaining this reference model, and does this strong dependence mean that the evaluation method is overly coupled to the data set and difficult to generalize to new scenarios?
- What does the right part of Figure 1 want to express?
- The lower part of Figure 2 uses reverse reading order. Is this intentional?
- What are the actual meanings of the various indicators in the experimental graph (Figure 5)? Can you explain the results in a more intuitive way (such as specific behavioral examples)?
- Section 5.3 mentions that unlearning has "behaviorally meaningful" - how is the "behaviorally meaningful" here defined and quantified? What counts as "meaningful"? Can an interpretive framework be provided that goes beyond a list of numbers?
7 comments
Response to Reviewer onqz (W2, W3, Q2, Q3 & Q4)
W2 and Q2. What does a retrained model cost? Is the audit tied to one dataset? And if retrained models exist, why unlearn?
The cost is a one-time setup for a pool of people (everyone whose deletion might be requested), the audit is not tied to one dataset, and it does not replace unlearning.
Table 1. Cost of the audit.
| Step | How often | Cost |
|---|---|---|
| Train a retrained model | once per pool of people | about 2 GPU-hours |
| Apply unlearning to a retrained model | once per method | 25 to 50 training steps |
| Score the unlearned model | every request | 942 forward passes per person |
A retrained model is trained once for a pool of people, not for each checkpoint. We trained three on one A800 GPU, and two such models are enough for the audit. If retraining is impossible, a quick check compares the unlearned model with the public model it started from and needs no retrained model.
Table 2. Quick check without any retrained model. Share of forgotten people flagged (%). The last row is a test on models that should not be flagged.
| Model | Flagged |
|---|---|
| TaskVec | 100 |
| GradDiff | 100 |
| NPO | 75 |
| Retrained models | 0 |
The quick check flags 75% to 100% of the forgotten people and no retrained model.
The audit is not tied to TOFU. For facts about people, it needs candidate answers of the same type as the true one. For free text such as news, it needs only the text and one retrained model, and it reads how likely the unlearned model finds each forgotten article. In the revision, we ran it on real news (MUSE-News) with MUSE's public retrained model and no training. Three of five public unlearned models get verdicts. For the other two, which MUSE rates as over-forgotten, the audit makes no claim.
Table 3. Share of forgotten articles still remembered (%) on MUSE-News. Models are named by method and uploader.
| Model | Still remembered |
|---|---|
| SimNPO (OPTML) | 71 |
| NPO (LV) | 83 |
| RMU (LV) | 89 |
These three models still remember 71% to 89% of the forgotten articles.
Retrained models do not make unlearning unnecessary. A retrained model is only an evaluation tool, as in TOFU and MUSE. It is built once to test an unlearning method, and each deletion request is then handled by unlearning, without retraining.
The audit needs a one-time setup plus cheap scoring per request, works on new data such as news, and checks unlearning without replacing it (Sections 5.7, 5.8).
W3 and Q3. What does the right part of Figure 1 express?
It showed that the same NPO model can look forgotten on average but still remember a specific person. In the revision, Figure 1 shows this with real numbers. The average loss on forgotten answers rises from 0.01 before unlearning to 1.74 after NPO, close to the 1.92 of retrained models, so NPO looks forgotten. Yet for one forgotten author, Bao Nguyen, two of the three NPO models (trained with three random seeds) give the true main genre in all ten generated answers, and no retrained model ever does.
W3 and Q4. Is the reverse reading order in the lower part of Figure 2 intentional?
It was, but it was confusing. In the revision, Figure 2 reads from top to bottom and left to right, and follows one forgotten author through the audit.
Response to Reviewer onqz (W3, W4, Q5 & Q6)
W3 and Q5. What do the indicators in Figure 5 mean?
Panels (a) to (f) are robustness checks, and all pass. Panels (a) to (e) each change one choice we fixed in advance and ask whether the score of the submitted audit (6.52) stays positive. Panels (g) and (h) compare our original score with a simple average. In the revision, we moved the figure to the appendix as Figure 14 and gave each panel a plain title.
Table 1. The robustness checks of Figure 14.
| Panel | Change | Result |
|---|---|---|
| (a) | calibration people redrawn 6,024 times | always positive |
| (b) | pairs redrawn 5,000 times | always positive |
| (c) | one retrained model at a time | lowest score 6.12 |
| (d) | controls removed from the scorer | lowest score 5.48 |
| (e) | only the closest pairs | lowest score 6.55 |
| (f) | false alarms on retrained models | rate 0.049 |
Panels (a) to (e) leave the score positive, and panel (f) shows false alarms at the designed rate of 0.05. Panels (g) and (h) show our original score next to a simple average of the same questions, 6.52 against 2.54. A model that was never unlearned shows a similar gap, so the revision no longer reads this gap as evidence (see our correction comment). For a concrete example of what the score means in behavior, see the forgotten author Chiamaka Adebayo under Q6 below, and Bao Nguyen in Figure 1.
W4 and Q6. What does "behaviorally meaningful" mean, and how is it measured?
In the revision, we replaced the phrase with three clear levels of what a model still knows about a forgotten person. A level counts as meaningful when the unlearned model is above models that never learned the person.
Table 2. The three levels. Likelihood is a standardized score, recognition a share of questions, and generation the recall above retrained models.
| Level | Question | TaskVec | GradDiff | NPO | Retrained models |
|---|---|---|---|---|---|
| Likelihood | Is the true answer more likely? | 3.10 | 2.70 | 0.96 | 0 |
| Recognition | Is it ranked first among about 100? | 35% | 38% | 24% | 6% |
| Generation | Is it said when asked? | +0.45 | +0.45 | +0.08 | 0 |
The gaps between levels show where unlearning acts. NPO lowers its confidence in all answers, so its likelihood falls, yet it still picks the true answer first far more often than retrained models (24% against 6%). For example, asked for the main genre of the forgotten author Chiamaka Adebayo, NPO gives the true genre (Queer), while all three retrained models answer Young Adult Fiction.
The verdict uses recognition, because a general loss of confidence cannot fake it. To help readers with the many numbers, the paper now has a reading guide (Table 3 of the paper) that says what each number means (Sections 3.3, 5.5).
We hope these answers are clear. If you have any further questions, please let us know, and we will answer as soon as possible. If the revision addresses your concerns, we would be grateful if you could take it into account in your final assessment. Thank you again.
Response to Reviewer onqz (W1 & Q1)
Thank you for the detailed review. We answer each question below.
W1 and Q1. How does the audit differ from RWKU, LUCID, and others, and why no direct comparison?
RWKU and LUCID do study what remains of specific targets, but they report averages or counts over many targets, and they read the answers the model generates. They do not say, for each person, whether that person is still remembered, with a known error rate. Our audit does, with a yes/no answer (a verdict) for each person. It checks whether the unlearned model still picks the person's true answer out of about 100 candidates more clearly than models that never learned the person. The closest per-person method, U-LiRA+ (Du et al.), trains about 100 shadow models, while ours reuses the same three retrained models for every person.
We agree that a direct comparison was missing. These audits give no decision for each person, so in the revision we computed their scores for each person on our models, with the same threshold for all audits (Section 5.2). AUROC measures how well a score separates NPO's forgotten people from retrained models without any threshold.
Table 1. Share of forgotten people found still remembered (%) on Llama-3.1-8B at one common threshold, with AUROC and cost in forward passes per person.
TaskVec and GradDiff forget little, and most audits find nearly all of their people, so the audits agree where they should. NPO forgets the most, and there ours finds 92% where the others find 33% to 63%, at a tenth of the cost of Leak@32. With the stricter threshold of the verdict, ours still finds 79%.
The gap holds in two harder cases, NPO on 30 held-out people and IdkDPO, which answers "I don't know" to 58% of the questions about forgotten people.
Table 2. Share of forgotten people found still remembered (%) in two harder cases, with the same threshold for all audits.
In both cases, the existing audits find at most 50% of the forgotten people, and ours finds 87% and 92%. Six of NPO's 24 audited people are missed by every other audit, and ours flags all six. No model that never learned the people is flagged (Table 2 of the correction comment), so the unlearning method alone does not explain what our audit finds.
Existing audits read the model's answers and miss many people that the model still remembers. Our audit finds them, at lower cost, with an answer for each person.