- Rating
- 2
- Confidence
- 4
- Informativeness
- 2
- Presentation
- 2
- Soundness
- 1
Read review
Paper summary
This paper studies cross-dataset node classification with graph-text foundation models. It proposes Perturbation Alignment (PA), which updates class text prototypes using target subgraph representations while keeping the pretrained encoders frozen. The paper also replaces average pooling with gated pooling. Experiments on seven target datasets show improved average accuracy, especially when class descriptions are unavailable.
Critical strengths
PA is simple and easy to integrate into existing graph-text models because it keeps the pretrained encoders fixed.
The ablation study is informative. PA alone improves the reported average accuracy from 62.01% to 62.94% with full descriptions, and from 53.76% to 57.17% with class names only. The latter setting provides the clearest evidence for the usefulness of PA.
Critical weaknesses
The margin-improvement guarantee is not established. Equation (9) implicitly uses
max_k(a_k - epsilon * b_k)
=
max_k(a_k) - epsilon * max_k(b_k),
which does not hold in general.
There is also ambiguity in the margin definition: the maximum appears to include the correct class, whereas a standard classification margin should use k != y. In addition, the proof does not account for the L2 normalization applied to the perturbed prototypes in Algorithm 1. These issues invalidate the current guarantee and require a corrected analysis or a weaker theoretical claim.
The experimental protocol depends on target-domain validation labels. Appendix A.4 states that epsilon and lambda are selected separately for each target dataset by grid search on the validation set. PA is therefore parameter-update-free, but the reported protocol is not strictly target-label-free. This should be reflected in comparisons with few-shot methods.
There are also reproducibility inconsistencies. Table 6 reports epsilon = 0 for the gated model on Instagram, while Table 3 reports an improvement from 63.04% to 64.20% after adding PA. In addition, Appendix A.4 states that the epsilon search range starts from 0.01, although Table 6 contains several epsilon = 0 entries.
Questions for authors
-
Can the authors provide a corrected margin analysis that accounts for L2 normalization? Is the maximum in Eqs. (7)-(9) intended to exclude the correct class?
-
How many labeled target-domain validation samples are used to select epsilon and lambda? Are the few-shot baselines given a comparable target-label budget?
-
What performance is obtained when the PA hyperparameters are fixed across target datasets or selected only from source-domain data?
-
Why does Table 6 report epsilon = 0 for the gated Instagram setting while Table 3 reports a clear gain from PA?
-
Why are epsilon = 0 values reported in Table 6 when Appendix A.4 states that the search range starts at 0.01?
5 comments
Weakness 2: Experimental protocol depends on target-domain validation labels; ε and λ selected via per-dataset grid search; PA should be reflected in few-shot comparisons.
Response:
We thank the reviewer for raising this concern. We position PA as a training-free cross-dataset generalization method, but relying on target-domain validation labels for hyperparameter selection does limit its zero-shot capability. To evaluate strict zero-shot performance, we add balanced PA, which uses no target labels. Its coefficient is computed from unlabeled data by gradient-norm balancing:
where provides numerical stability. We use the nearest shared candidate and fix for complete descriptions, for category-only templates, and when the corresponding source domain is absent. These settings are shared across datasets within each scenario.
Table 1. Accuracy (%) under the fixed, label-free balanced-PA configuration. Desc indicates whether full class descriptions are provided.
| Method | Desc | Cora (Academic) | History (E-commerce) | WikiCS (Wikipedia) | Instagram (Social) | Avg |
|---|---|---|---|---|---|---|
| None | ✓ | 67.04 ± 2.31 | 53.13 ± 0.26 | 71.44 ± 0.16 | 63.04 ± 0.32 | 63.66 |
| Compression only | ✓ | 66.97 ± 1.88 | 52.46 ± 0.28 | 71.74 ± 0.21 | 62.64 ± 0.38 | 63.45 |
| Decorrelation only | ✓ | 67.04 ± 2.05 | 53.52 ± 0.27 | 71.46 ± 0.22 | 63.30 ± 0.27 | 63.83 |
| Balanced PA | ✓ | 67.59 ± 2.05 | 53.32 ± 0.20 | 71.27 ± 0.23 | 63.10 ± 0.31 | 63.82 |
| None | × | 66.05 ± 2.23 | 52.80 ± 0.29 | 67.44 ± 0.31 | 43.37 ± 0.62 | 57.42 |
| Compression only | × | 66.73 ± 2.48 | 52.12 ± 0.31 | 67.62 ± 0.35 | 44.06 ± 0.70 | 57.63 |
| Decorrelation only | × | 65.68 ± 2.27 | 50.76 ± 0.31 | 67.21 ± 0.43 | 43.25 ± 0.52 | 56.73 |
| Balanced PA | × | 65.38 ± 1.88 | 55.04 ± 0.22 | 66.77 ± 0.18 | 42.95 ± 0.33 | 57.53 |
Balanced PA improves the average over no PA from 63.66% to 63.82% with complete descriptions and from 57.42% to 57.53% with class names. The component differences also show why a shared coefficient is a robust default rather than a guaranteed optimum: decorrelation is more useful with complete descriptions, whereas compression is more useful with category names.
Table 2. Accuracy (%) when the corresponding source domain is included or excluded. Balanced PA uses the shared label-free settings; searched PA uses validation labels.
| Source domain | PA | Cora (Academic) | Computers (E-commerce) | Instagram (Social) |
|---|---|---|---|---|
| Included | None | 67.04 ± 2.31 | 63.94 ± 0.56 | 63.04 ± 0.32 |
| Excluded | None | 55.35 ± 0.84 | 47.05 ± 0.68 | 47.65 ± 0.99 |
| Excluded | Balanced | 67.77 ± 0.77 | 52.42 ± 0.14 | 49.24 ± 0.79 |
| Excluded | Searched | 71.16 ± 0.97 | 52.59 ± 0.38 | 53.95 ± 0.38 |
When the academic, e-commerce, or social source domain is removed, balanced PA improves Cora, Computers, and Instagram by 12.42, 5.37, and 1.59 percentage points, respectively, without target validation labels. Searched PA is higher on all three because it adapts and to each target setting. The corresponding grouped bar charts are shown in Figure 4(a--b), and the raw values are in Appendix A.7.
Figure 3(b) already compares PA against few-shot methods. Few-shot baselines require gradient-based training to optimize class-specific prompt vectors from labeled examples, whereas PA is entirely training-free—a practical advantage of our approach. Regarding label budgets, num_shot denotes the number of labeled examples per class. PA uses a validation set of 50 samples. At num_shot=10, few-shot methods consume 10×N labeled examples (where N is the number of classes), which substantially exceeds the 50 unlabeled samples used by PA on all datasets except Instagram.
Question 1: Can the authors provide a corrected margin analysis that accounts for L2 normalization? Is the maximum in Eqs. (7)-(9) intended to exclude the correct class?
Response:
Yes. As detailed in our response to Weakness 1, the revised analysis explicitly includes the post-update L2 normalization and defines the signed margin as . The three revised theorems establish local descent of the normalized objective, a bound on margin and error-rate changes, and a sufficient condition for local risk reduction. Complete proofs are in Appendix A.8.
Question 2: How many labeled target-domain validation samples are used to select ε and λ? Are the few-shot baselines given a comparable target-label budget?
Response:
Figure 3(b) provides the direct few-shot comparison: num_shot denotes the number of labeled examples per class, whereas PA uses a validation set of 50 samples. Thus, at num_shot=10, few-shot methods use labeled examples for classes, which exceeds PA's 50 unlabeled samples on all datasets except Instagram. Balanced PA uses no target labels and is the strict-zero-shot counterpart.
Question 3: What performance is obtained when PA hyperparameters are fixed across target datasets or selected only from source-domain data?
Response:
This is evaluated by balanced PA in Weakness 2. It fixes across datasets and uses , , or by scenario, without target labels. It improves the average from 63.66% to 63.82% with complete descriptions and from 57.42% to 57.53% with class names (Table 1), and improves Cora, Computers, and Instagram by 12.42, 5.37, and 1.59 points when their corresponding source domains are removed (Table 2). Validation-selected PA is higher because it adapts the two hyperparameters to each target setting.
Question 4: Why does Table 6 report ε = 0 for the gated Instagram setting while Table 3 reports a clear gain from PA?
Response:
We apologize for this inconsistency. It was caused by an error in processing the reported hyperparameter table. The actual setting for gated Instagram used , not , and the corresponding final accuracy has been corrected in the revised manuscript from the erroneous 64.20% to 63.29%.
Question 5: Why are ε = 0 values reported in Table 6 when Appendix A.4 states that the search range starts at 0.01?
Response:
We apologize for the reporting error. The search-range entry and several selected and values in the previous version were affected by the same data-processing issue. In the revised manuscript, we have corrected the search ranges to and .
We apologize that the initial submission was not as polished as it should have been in terms of experimental depth, theoretical development, and presentation quality. In this revision, we have done our utmost to address these shortcomings through comprehensive refinements across all three dimensions. We are truly sorry for the rough state of the original manuscript, and we deeply appreciate the reviewer's patience and understanding in evaluating the earlier version.
Weakness 1: Margin-improvement guarantee not established; Eq. (9) incorrectly separates maximization; ambiguity in margin definition including correct class; L2 normalization ignored; requires corrected analysis.
Response:
We thank the reviewer for identifying these errors. We sincerely apologize for this oversight. We have replaced it with three theorems that establish local objective descent, bound changes in classification margins and error rates, and give a sufficient condition for local risk reduction under normalized updates. The revised statements are reproduced below.
First, we establish that the normalized gradient update decreases the combined objective locally.
Theorem 1 (Objective descent under a normalized gradient update).
Given N≥2 unit-norm pretrained class templates Z={zk}k=1N, a unit-norm subgraph encoding h, and the combined objective
L(Z)=Nλ∑j=1Nh⊤zj+N(N−1)1∑i=1N∑j=izi⊤zj,
suppose that the current templates are not a stationary point of this objective under the unit-norm constraints. Then a single ordinary gradient-descent step followed by normalization strictly decreases the combined objective for all sufficiently small positive step sizes.
Theorem 1 justifies the normalized PA update as a local descent step for the combined objective.
Second, we characterize how the step size controls PA's effect on classification correctness. Theorem 2 identifies a margin neighborhood containing all samples whose correctness may change, while preserving correct classification for samples with sufficiently large positive margins.
Theorem 2 (Classification stability under a normalized gradient update).
Given N≥2 unit-norm class templates Z and a unit-norm encoding h, apply one gradient-descent step to L in Theorem 1 with λ≥0, followed by normalization. For true label y∈{1,…,N}, let m(0):=h⊤zy−maxk=yh⊤zk and m(ϵ) denote the original and updated margins. For 0<ϵ<N/(λ+2),
∣m(ϵ)−m(0)∣≤N4ϵ(λ+2).
Thus any sample with m(0)>4ϵ(λ+2)/N remains correctly classified. For any target distribution, let R(0) and R(ϵ) be the error rates before and after updating each input separately, with Z, λ, ϵ, and the tie-breaking rule fixed. Then
\le\Pr\!\left(|m(0)|\le\frac{4\epsilon(\lambda+2)}{N}\right).$$ For fixed $N$ and $\lambda$, increasing $\epsilon$ within the admissible range widens this neighborhood. Its probability mass bounds the possible accuracy change, quantifying the scope for correcting errors and classification stability outside the neighborhood. Finally, we give a sufficient condition for a local reduction in classification risk under normalized perturbations. **Theorem 3 (Local classification-risk reduction under a normalized perturbation).** Given $N\ge2$ unit-norm class templates $\mathbf Z$ and a random unit-norm encoding $\mathbf H$ with true label $Y$, perturb each template by $\epsilon\mathbf v_k(\mathbf H)$ and normalize it, where the directions are bounded, measurable, and independent of $\epsilon$. Let $m(0)$ and $m(\epsilon)$ be the margins defined in Theorem 2, and let $D:=m'(0+)$ denote the right derivative in $\epsilon$. Suppose that $m(0)$ has a density $f_m$ continuous and positive at zero; a version of the conditional distribution of $D$ given $m(0)=t$ converges weakly to the same limit from both sides as $t\to0$; and $m(\epsilon)=m(0)+\epsilon D+o(\epsilon)$ holds uniformly, up to null events, on $\{|m(0)|\le C\epsilon\}$ for every fixed $C>0$ independent of $\epsilon$. Writing $a:=\lim_{t\to0}\mathbb E[D\mid m(0)=t]$, the classification error $R(\epsilon)$ under any fixed tie-breaking rule satisfies $$R(\epsilon)=R(0)-\epsilon f_m(0)a+o(\epsilon).$$ Thus $a>0$ guarantees a strict error reduction for all sufficiently small positive step sizes. Theorem 3 links local accuracy improvement to the mean first-order margin change near the decision boundary. $a>0$ is a sufficient condition for risk reduction. For space considerations, complete proofs of all three theorems appear in **Appendix A.8**. Appendix A.9 checks the boundary condition ($a>0$).