- Rating
- 2
- Confidence
- 3
- Informativeness
- 2
- Presentation
- 2
- Soundness
- 2
Read review
Paper summary
The paper proposes a first-order bilevel method named SUPRO that learns to track an auxiliary lower-level response. A joint representation of the explicit decision and auxiliary response predicts a displacement, and fixed-budget gradient corrections produce reactive and predictive candidates and objective comparisons safeguard their selection. A learned proposal and bounded response mixing further modify the updates. Under a uniform PL condition the authors prove a tracking bound that scales with the smaller of the response-path variation and the prediction error in theorem 4.2 and rate of the fixed penalty in theorem 4.4. Experiments on a synthetic nonconvex problem, few-shot learning, data hyper-cleaning, and RoBERTa-base adaptation on SST-2 show faster convergence and comparable final accuracy against other baselines.
Critical strengths
-
The paper explicitly distinguishes fixed-penalty stationarity from original bilevel stationarity, and includes a counterexample in Appendix A.5.2.
-
Table 1 reports shorter time to 90% accuracy on Omniglot settings, Table 2 reports synthetic improvements, and Tables 3-4 explore both algorithmic components and architectural choices.
Critical weaknesses
-
Figure 2 and Tables 3-4 compare outer iterations, but SUPRO constructs multiple candidates, evaluates their objectives and gradients and trains a predictor. The paper acknowledged these costs in Appendices A. Table 3's 90% threshold improves from 350 iterations for two-step GD/linear extrapolation/no learned prediction to 320 for SUPRO, this small reduction may be offset by per-iteration overhead, so wall-clock time evaluation is needed on SST-2.
-
In Table 3 removing the reactive candidate even yields a higher peak accuracy though threshold attainment is slower. Two pairs of variants report same numbers of all columns (linear extrapolation & no learned prediction, no mixing & two-step GD), which suggests the components distinguishing them are inactive and removing the reactive fallback gives the highest peak.
-
Theorem 4.2 conditions the improvement on prediction error , which is not established for the online trained predictor (i.e., it does not establish that online losses reduce that error). No tracking error connects this mechanism to the observed gains. Also the rate established in Theorem 4.4 is obtained from the reactive branch.
Questions for authors
-
Please report wall-clock time comparison and numbers of gradient/function evaluations to reach the target for all methods on BiDoRA and hyper-cleaning. Besides, do improvements in tables 3, 4 hold across different random seeds?
-
Can you provide diagnostics for prediction/tracking errors and candidate selection on the synthetic problem where exact lower-level responses are available?
-
Along the trajectories, is in practice, and how often are learned proposals accepted?
-
Why is w/o learned network (89.22%) which keeps selection and mixing worse than the two-step GD reference (91.17%)?
9 comments
Q2 Accuracy under a fixed time budget and cost to reach the target
2. Accuracy under a fixed time budget and cost to reach the target
Thank you for suggesting comparisons based on actual training time. For the 36-run BiDoRA evaluation described above, we report accuracy at a fixed 25-minute budget and the time and task-gradient cost to reach 90% accuracy.
The fixed-budget result is the last completed validation at or before 25 minutes, without interpolation.
| Rank | MEHA accuracy (%) | BOME accuracy (%) | Two-step GD accuracy (%) | SUPRO accuracy (%) |
|---|---|---|---|---|
| 4 | 77.75 ± 9.59 | 52.14 ± 2.12 | 87.16 ± 0.80 | 88.69 ± 1.46 |
| 8 | 88.65 ± 0.40 | 87.88 ± 0.99 | 89.11 ± 0.80 | 90.75 ± 1.15 |
| 16 | 87.58 ± 2.25 | 87.16 ± 2.29 | 88.30 ± 1.99 | 90.37 ± 0.53 |
SUPRO achieves the highest mean accuracy within the same time budget at all three ranks. Its gains over two-step GD are 1.53, 1.64, and 2.06 percentage points, calculated before rounding. At rank 8, the gains over MEHA and BOME are 2.10 and 2.87 percentage points. By the last validation within this budget, SUPRO completes 450–460 steps, compared with 420 for the baselines.
We also report the first validation reaching 90% accuracy, together with success counts:
| Rank | Method | Successful seeds | Time to 90% (min) | Task-gradient requests to 90% |
|---|---|---|---|---|
| 4 | MEHA | 0/3 | NR | — |
| 4 | BOME | 0/3 | NR | — |
| 4 | Two-step GD | 1/3 | 27.47 | 2,350 |
| 4 | SUPRO | 2/3 | 24.18 ± 2.16 | 2,160 ± 191 |
| 8 | MEHA | 1/3 | 21.84 | 2,220 |
| 8 | BOME | 1/3 | 26.58 | 2,700 |
| 8 | Two-step GD | 2/3 | 23.39 ± 3.32 | 2,000 ± 283 |
| 8 | SUPRO | 3/3 | 20.31 ± 3.11 | 1,800 ± 274 |
| 16 | MEHA | 1/3 | 25.46 | 2,580 |
| 16 | BOME | 1/3 | 27.88 | 2,820 |
| 16 | Two-step GD | 2/3 | 23.44 ± 3.21 | 2,000 ± 283 |
| 16 | SUPRO | 2/3 | 19.65 ± 3.66 | 1,710 ± 318 |
NR means the target was not reached within 500 steps. Means and sample standard deviations are calculated over successful runs; single-success entries have no standard deviation. These conditional statistics should be interpreted together with their success counts.
Target gradient counts are calculated from the recorded first-hit steps and fixed evaluation schedule. SUPRO uses nine task-gradient requests at each full selection event and four at each prediction-only step; MEHA/BOME use six per step, and two-step GD uses five. These counts exclude initialization and response-network backward passes. Timing includes online training and candidate evaluations.
Across the nine rank–seed settings, SUPRO reaches 90% in seven runs, compared with five for two-step GD and two each for MEHA and BOME. At rank 8, SUPRO reaches the target in all three seeds. The fixed-budget and target-based measurements therefore provide complementary evidence of practical efficiency.
Q3 Reactive-candidate ablation and matching table entries
3. Reactive-candidate ablation and matching table entries
We appreciate the reviewer’s close examination of Table 3. The reactive candidate provides a prediction-independent reference for response selection. Its role is to control the response update when learned predictions are less informative; the highest validation accuracy attained by an individual run measures a different aspect of performance.
Removing the reactive candidate yields a peak accuracy of 91.743%, compared with 91.514% for full SUPRO. The full method, however, reaches 90% at 320 iterations, compared with 390 iterations without the reactive candidate. This experiment therefore associates the reactive reference with earlier target attainment, while the variant without it obtains a slightly higher later peak.
Regarding the matching entries, Table 3 summarizes each trajectory through peak accuracy, the first iteration attaining that peak, and the first iteration reaching 90%. Accuracy is evaluated on 872 examples, and evaluations occur every 10 iterations (Table 5). Matching values of these discrete summaries do not imply identical updates or trajectories.
The prediction controls provide a concrete example. Under the original single-seed control protocol, retaining C/U, candidate selection, and mixing while changing only the prediction residual gives:
| Prediction configuration | Accuracy at step 500 (%) | Peak accuracy (%) |
|---|---|---|
| Full learned prediction | 90.8257 | 91.5138 |
| Without the learned P residual | 90.8257 | 91.2844 |
| Linear extrapolation | 89.9083 | 91.2844 |
Although linear extrapolation and removal of the learned prediction residual attain the same peak accuracy, their step-500 accuracies differ by 0.9174 percentage points. The complete method attains the highest peak and matches the higher final accuracy. These controls use the original ablation protocol and are presented separately from the additional periodic-correction experiments.
The no-mixing and two-step-GD variants also have different update definitions. The former retains learned response construction and candidate selection, whereas the latter directly applies two gradient-based response updates. Figure 3 provides their trajectories, including a small difference around iteration 270 despite their matching summaries. We will make the relationship between these trajectory-level observations and Table 3 more explicit.
The intended interpretation is therefore based on both target attainment and attained accuracy. In the original component study, full SUPRO reaches 90% earlier than each ablated configuration, while the supplementary wall-clock results evaluate the complete practical configuration with its additional computational costs included.
Q1 Wall-clock efficiency and computational cost
1. Wall-clock efficiency and computational cost
We thank the reviewer for the careful assessment and for recognizing the prediction-dependent analysis and the distinction between fixed-penalty and original bilevel stationarity. We appreciate the concern that candidate construction and online training introduce costs beyond an outer iteration. Our practical objective is to improve performance within a given training budget.
We evaluated four methods at adapter ranks 4, 8, and 16, using three paired seeds per configuration: 36 completed BiDoRA runs. Within each rank and seed, the methods share initial task parameters, data splits, and upper-/lower-level minibatch sequences. All runs use FP32, batch size 32, and 500 outer steps.
Timing includes prediction, correction, candidate evaluation, online network training, data handling, and subsequent validation. It starts after initial validation and excludes startup, model loading, initial warmup/validation, and final saving.
Accuracy and time are mean ± sample standard deviation; memory entries are means across three seeds.
SUPRO achieves the highest mean final accuracy and shortest training time at each rank under these configurations. Its 500-step time reduction ranges from 6.38% to 9.70%, including online learning and candidate-evaluation costs. At rank 16, mean final accuracy exceeds MEHA, BOME, and two-step GD by 2.29, 2.45, and 1.07 percentage points, respectively. The additional peak memory is approximately 1.7–2.9 GiB.
For 500 steps, SUPRO uses 2,385 task-gradient requests and 74 response-network backward passes. MEHA/BOME use 3,000 task-gradient requests, and two-step GD uses 2,500. This corresponds to 20.5% and 4.6% fewer task-gradient requests. Network backward passes are reported separately because these operations have different costs; objective-evaluation costs are included in timing.
For reproducibility, this additional SUPRO configuration predicts at every step and performs full correction and selection during the first 30 steps and every 10 steps thereafter: 77 selection events and 423 prediction-only steps. BOME uses three inner steps.
At ranks 4/8, the main upper- and lower-variable learning rates warm up to 2.5×10−6 over 100 steps. SUPRO increases them to 5×10−6 over steps 101–200; the baselines retain 2.5×10−6. The corresponding main learning rates are halved at rank 16. These results assess the complete practical configurations under the declared schedules; the original component controls are discussed separately below.