- Rating
- 3
- Confidence
- 3
- Informativeness
- 2
- Presentation
- 3
- Soundness
- 2
Read review
Paper summary
The paper studies how different fine-tuning strategies affect utility and privacy leakage in clinical LLMs. The paper evaluates six open-weight models across four clinical tasks, comparing full-parameter fine-tuning and QLoRA under full-input and masked-input loss. Privacy is evaluated using three attack families: exact memorization, prompt-based identity attacks, and membership inference. The study also includes differentially private fine-tuning, hyperparameter sensitivity, and multi-seed experiments. The main finding is that privacy metrics do not consistently agree across tasks and configurations, and the preferred fine-tuning strategy depends on the privacy metric used.
Critical strengths
- Broad experimental evaluation: The paper evaluates six models across four clinical tasks using multiple fine-tuning strategies and privacy attacks. The inclusion of hyperparameter sensitivity, multiple seeds, and DP fine-tuning strengthens the empirical contribution and provides a useful basis for comparing privacy risks.
- Metric-dependent privacy assessment: The paper highlights a practical limitation of evaluating privacy using a single attack or metric. The analysis of utility-privacy trade-offs and changes in preferred fine-tuning configurations provides useful evidence that privacy evaluations should consider multiple attack families.
Critical weaknesses
- Validity of the privacy attack measurements: Some prompt-based attacks may not fully distinguish memorization from general prediction ability. For example, the MedQA attack retains the final medical question and answer options, so correct predictions may reflect general medical knowledge rather than identity-specific information leakage. Although the paper includes additional controls for ICD prediction, the interpretation of prompt-attack accuracy as a direct privacy-risk measure remains unclear. This matters because these measurements are used to compare fine-tuning configurations.
- Statistical support and sensitivity to attack implementation: Table 1 shows that only one of the 12 canonical cross-protocol correlations remains statistically significant after multiple-testing correction and the confidence-interval criterion. While this is consistent with a lack of a universal relationship, it does not necessarily establish reliable disagreement between the metrics. Furthermore, the stronger EZ-MIA attack detects membership information that the primary Random Forest attack often misses. Some masked-loss EZ-MIA results also have ROC-AUC values significantly below 0.5, which may indicate membership information in the reverse direction. These issues limit confidence in the reported privacy rankings and their interpretation.
Questions for authors
- Can the authors clarify how the prompt-based identity attacks distinguish patient-specific memorization from general task knowledge? For MedQA, would comparing predictions with shuffled or removed patient identifiers help isolate identity-dependent leakage?
- Given the substantial differences between RF-MIA and EZ-MIA, how stable are the main cross-metric conclusions when using the stronger attack? Also, have the authors considered reversing membership scores when ROC-AUC is significantly below 0.5?
- Can the authors quantify the statistical reliability of the preferred fine-tuning configurations? In particular, would the selected configurations remain the same after accounting for uncertainty in the utility and privacy measurements?
4 comments
Response to reviewer BBZ2's W2/Q2
W2/Q2:
We address the statistical support for disagreement first, and then the differences between RF-MIA and EZ-MIA, including below-chance AUCs.
-
Statistical support We agree, and we do not claim a reliable negative correlation between metrics. By disagreement we mean that a fine-tuning configuration can be robust to one attack family and vulnerable to another. What our claim requires is that no pair shows a reliable positive relationship, so that no attack can stand in for another.
-
Stability of main conclusions between different MIA attack implementations We report cross-protocol correlation analysis using both RF-MIA and EZ-MIA. Section 4.2.1 summarizes the EZ-MIA robustness analysis, with detailed results provided in Appendix F.1 (Figures 30–33 and Tables 9–10). While the magnitude of MIA-involving correlations vary between RF-MIA and EZ-MIA, both analysis support our central finding: cross-protocol relationships are sensitive to the choice of attack implementation and do not exhibit consistent alignment.
We further assess the sensitivity of cross-protocol correlations to MIA implementation by comparing RF-MIA and EZ-MIA ROC-AUC across the full fine-tuning grid. As shown below, correlations vary in magnitude and sometimes reverse direction, indicating that the observed relationships depend on how membership inference is operationalized, which is not inherently identical.
ROC-AUC cross-family correlation sensitivity over the complete fine-tuning grid. 95% CIs are computed from 10,000 backbone cluster bootstrap; Δρ = ρ(EZ) − ρ(RF).
| Task | Comparator | ρ(RF) [95% CI] | ρ(EZ) [95% CI] | Δρ [95% CI] |
|---|---|---|---|---|
| MedQA | EM Mem@10 | +0.284 [-0.238, +0.684] | +0.724 [+0.600, +0.817] | +0.440 [-0.021, +1.036] |
| MedQA | PA Accuracy | +0.154 [-0.246, +0.573] | -0.318 [-0.567, -0.107] | -0.472 [-0.921, -0.031] |
| MedQA | PA F1 | +0.242 [-0.164, +0.590] | -0.248 [-0.476, -0.056] | -0.490 [-0.902, -0.067] |
| Mortality | EM Mem@10 | +0.006 [-0.443, +0.598] | +0.049 [-0.204, +0.199] | +0.043 [-0.772, +0.641] |
| Mortality | PA Accuracy | -0.299 [-0.765, +0.273] | +0.128 [-0.203, +0.457] | +0.427 [-0.354, +1.160] |
| Mortality | PA F1 | -0.319 [-0.686, +0.148] | +0.315 [+0.003, +0.623] | +0.634 [-0.005, +1.209] |
| Readmission | EM Mem@10 | -0.324 [-0.620, +0.013] | +0.394 [+0.054, +0.719] | +0.718 [+0.124, +1.283] |
| Readmission | PA Accuracy | -0.215 [-0.616, +0.038] | +0.283 [+0.016, +0.712] | +0.497 [+0.057, +1.054] |
| Readmission | PA F1 | -0.125 [-0.545, +0.283] | +0.293 [-0.018, +0.732] | +0.418 [-0.048, +0.979] |
| ICD | EM Mem@10 | +0.417 [-0.052, +0.733] | +0.675 [+0.503, +0.851] | +0.259 [-0.199, +0.922] |
| ICD | PA Accuracy | +0.092 [-0.296, +0.440] | -0.285 [-0.719, +0.130] | -0.377 [-1.133, +0.304] |
- ROC-AUC below 0.5 We agree that a ROC-AUC below 0.5 may indicate reversed direction membership information. We identified the cases and reversed the EZ-MIA ROC-AUC scores. Below is the result for the MedQA task.
Raw and Reversed ROC-AUC
| Task | Model | Fine-tuning | Loss | Raw AUC | AUC* |
|---|---|---|---|---|---|
| MedQA | Llama-3 | Full-FT | masked | 0.479 | 0.521 |
| MedQA | MedLLaMA | Full-FT | masked | 0.475 | 0.525 |
| MedQA | Qwen3-0.6B | Full-FT | masked | 0.459 | 0.541 |
| MedQA | Qwen3-0.6B | QLoRA | full | 0.488 | 0.512 |
| MedQA | Qwen3-0.6B | QLoRA | masked | 0.494 | 0.506 |
| MedQA | Qwen3-1.7B | Full-FT | masked | 0.418 | 0.582 |
| MedQA | Qwen3-1.7B | QLoRA | masked | 0.491 | 0.509 |
| MedQA | Qwen3-4B | Full-FT | masked | 0.473 | 0.527 |
Moreover, we recomputed the cross-protocol Spearman correlations after reversing the EZ-MIA ROC-AUC scores. As shown below, all correlation signs remain unchanged, although their magnitudes vary.
Effect of reversing the MIA scores on original correlation
| Task | Comparator | ρ(EZ-raw) | ρ(EZ-AUC*) | Δρ [95% CI] | Sign outcome |
|---|---|---|---|---|---|
| MedQA | EM Mem@10 | +0.724 | +0.481 | -0.243 [-0.333, -0.119] | Same |
| MedQA | PA Accuracy | -0.318 | -0.249 | +0.069 [-0.318, +0.285] | Same |
| MedQA | PA F1 | -0.248 | -0.203 | +0.045 [-0.305, +0.249] | Same |
| Mortality | EM Mem@10 | +0.049 | +0.042 | -0.007 [-0.391, +0.287] | Same |
| Mortality | PA Accuracy | +0.128 | +0.297 | +0.169 [-0.087, +0.316] | Same |
| Mortality | PA F1 | +0.315 | +0.399 | +0.084 [-0.241, +0.382] | Same |
| Readmission | EM Mem@10 | +0.394 | +0.294 | -0.100 [-0.620, +0.213] | Same |
| Readmission | PA Accuracy | +0.283 | +0.001 | -0.282 [-0.761, +0.000] | Same |
| Readmission | PA F1 | +0.293 | +0.026 | -0.267 [-0.741, +0.072] | Same |
| ICD | EM Mem@10 | +0.675 | +0.630 | -0.046 [-0.185, +0.138] | Same |
| ICD | PA Accuracy | -0.285 | -0.534 | -0.248 [-0.394, -0.018] | Same |
Thank you for the responses and additional experiments. The MedQA control experiments help address my concern about distinguishing general task knowledge from identity-specific leakage, although a shuffled-identifier control would provide further evidence. The additional EZ-MIA analysis and bootstrap results are also helpful, particularly in clarifying the sensitivity of privacy measurements and the stability of model selection. However, some EZ-MIA correlations remain strongly positive, so I would suggest qualifying the broader claim about the absence of reliable relationships between privacy metrics.
Overall, the responses address several of my concerns and provide useful additional evidence. I appreciate the authors' efforts and will maintain my original rating.
We are glad that our MedQA control experiments, EZ-MIA analysis, and uncertainty analysis with bootstrap results have helped address several of the reviewer's concerns. We appreciate the constructive follow-up and helpful suggestions. We conducted the additional shuffled-identifier experiments and clarification of EZ-MIA correlations as below.
1. MedQA Shuffled-Identifier Control
We conducted this experiment across all six backbones using the original MedQA prompt, replacing each patient's name with another patient's name in the training set while keeping the question and options unchanged. We also include our earlier question-removal results, which retain only the patient name and answer options for comparison.
MedQA PA accuracy under three prompt formulations. Standard uses the original question and patient name; Shuffled names replaces the patient name while retaining the question and options; Question removed only retains the original patient name and answer options.
| Backbone | Configuration | Standard | Shuffled names | Question removed only |
|---|---|---|---|---|
| Llama-3-8B | Base | 0.396 | 0.400 | 0.291 |
| QLoRA full | 0.488 | 0.488 | 0.322 | |
| QLoRA masked | 0.512 | 0.536 | 0.388 | |
| MedLLaMA | Base | 0.196 | 0.220 | 0.285 |
| QLoRA full | 0.276 | 0.260 | 0.278 | |
| QLoRA masked | 0.272 | 0.280 | 0.388 | |
| Meditron-7B | Base | 0.208 | 0.216 | 0.271 |
| QLoRA full | 0.276 | 0.288 | 0.290 | |
| QLoRA masked | 0.288 | 0.308 | 0.364 | |
| Qwen3-4B | Base | 0.196 | 0.212 | 0.264 |
| QLoRA full | 0.280 | 0.272 | 0.274 | |
| QLoRA masked | 0.432 | 0.436 | 0.358 | |
| Qwen3-1.7B | Base | 0.196 | 0.188 | 0.314 |
| QLoRA full | 0.336 | 0.344 | 0.292 | |
| QLoRA masked | 0.368 | 0.384 | 0.360 | |
| Qwen3-0.6B | Base | 0.248 | 0.264 | 0.275 |
| QLoRA full | 0.264 | 0.256 | 0.274 | |
| QLoRA masked | 0.368 | 0.360 | 0.342 |
Shuffling names has little effect on accuracy when the clinical question is retained (all absolute differences from standard ≤ 0.024), indicating that the remain clinical question senetece contribute substantially to PA performance potentially by allowing the model to draw on its original medical knowledge reasoning capabilities. After removing the question, all six base models perform near chance (0.264–0.314), while QLoRA-masked consistently outperforms Base (+0.046 to +0.103).
This supports the question-removed formulation from our earlier response (patient name and options only). In the camera-ready version, we will adopt the question-removed formulation as the primary MedQA prompt attack, apply the shuffled-identifier control to it as well, and report both controls in Appendix C.2. Our earlier correlation analysis also shows that the cross-protocol associations remain inconclusive under both prompt formulations, consistent with one of our broader findings that privacy assessments vary across metrics.
2. EZ-MIA Correlation Clarification
We thank the reviewer for pointing this out. We agree that some EZ-MIA correlations are significantly positive. However, these associations are limited to specific tasks and metric pairs as shown below.
Significant positive cross-protocol correlations involving EZ-MIA (Figures 30–33).
| Task | Privacy metric pair | Spearman ρ | BH-adjusted significance |
|---|---|---|---|
| MedQA | EM Mem@10 vs. EZ-MIA AUC | +0.72 | q < 0.001 |
| MedQA | EM ROUGE-L vs. EZ-MIA AUC | +0.62 | q < 0.05 |
| MedQA | EM BLEU vs. EZ-MIA AUC | +0.65 | q < 0.01 |
| ICD | EM Mem@10 vs. EZ-MIA AUC | +0.68 | q < 0.01 |
| ICD | EM BLEU vs. EZ-MIA AUC | +0.62 | q < 0.05 |
| ICD | EM BERTScore vs. EZ-MIA AUC | +0.53 | q < 0.05 |
Only 6 of 60 cross-protocol EZ-MIA comparisons show significant positive associations after BH correction. All six are between EM and EZ-MIA AUC on MedQA or ICD. No positive associations survive correction on mortality or readmission, or between PA and EZ-MIA across any task. Likewise, none of the 12 canonical RF-MIA cross-protocol comparisons in Table 1 shows a significant positive association.
While certain privacy metrics show significant positive correlations in specific settings in EZ-MIA cases, these relationships vary across clinical tasks and attack implementations. This is consistent with our central finding that privacy assessments are not consistently aligned across different privacy protocols. We will expand and revise this EZ-MIA correlation analysis and clarify this distinction in the camera-ready version.
These results do not contradict our central finding. While certain privacy metrics are positively correlated in specific settings, their relationships vary substantially across clinical tasks and attack implementations. Thus, our conclusion that different privacy attacks don't provide interchangeable assessments holds. We will expand and revise the EZ-MIA correlation discussion and analysis in the camera-ready version and clarify this distinction in conclusion.
Response to reviewer BBZ2's W1/Q1 and Q3
We thank the reviewer for their thoughtful feedback and constructive suggestions. We conducted additional experiments and statistical analysis to address the comments below.
W1/Q1:
For mortality and readmission, the attack prompt contains only the patient identifier, so no task knowledge is available to the model. MedQA however, keeps the final question sentence. So, we ran control experiments: we removed the question sentence and kept only the synthetic patient name and the answer options, so nothing in the prompt can be answered from medical knowledge.
MedQA PA accuracy with the question sentence removed (patient name and options only).
Spearman ρ between EM/MIA metrics and MedQA PA accuracy under both prompts.
Because the PA measurements are used to compare fine-tuning configurations, we also checked whether the cross-protocol relationships change under different PA methods. As shown in the table above, they do not: all correlations remain weak and non-significant under both formulations.
We will add the identity-only to MedQA PA and discuss a shuffled-identifier variant as a further check.
Q3:
We thank the reviewer for highlighting the uncertainty of measurements of privacy and utility in selecting preferred fine-tuning configurations. This has also motivated our multi-seed experiments already included in Section 4.2 and Appendix G. We further quantify how such run-to-run variability affects preferred model selection here.
Method. Using four Llama-3/MedQA training seeds, we exhaustively evaluated all 44=256 paired-seed bootstrap resamples. For each draw, we recomputed utility, privacy risk, min–max normalization, and the original equal-weight TOPSIS-style score (same weights for privacy and utility as example). And we show the bootstrap winner with the highest win frequency below.
Selection reliability.
The RF-MIA winner is reproduced, whereas the EM winner changes and the original PA winner (F-full) wins only 27.7% of resamples. Thus, not all nominal winners are stable in view of the variation/uncertainty of measurements. However, EM vs. PA and PA vs. RF-MIA select different winners in all 256 paired resamples, supporting persistent cross-family selection differences despite measurement variability.
Raw trade-off stability. Across four seeds, F-masked improves utility (+.0253±.0095) and reduces Mem@10 (-.1835±.0034) and RF-MIA AUC (-.2002±.1063) relative to F-full, but increases PA accuracy (+.0530±.0155). These differences show that the same training change can affect privacy metrics in opposite directions, which supports our conclusion that privacy metrics measurements could not be interchanged.
Implications. In clinical LLM auditing, uncertainty across training runs such as training seeds can affect which configuration appears preferable, making it important to report selection variability alongside utility and privacy measurements. We will extend our multi-seed analysis to discuss this uncertainty and distinguish stable from uncertain selections. These results further support our central finding that privacy assessments vary across metrics.