- Rating
- 3(+1)
- Confidence
- 2(−1)
- Informativeness
- 2
- Presentation
- 3
- Soundness
- 2
Read review
Paper summary
This paper proposes SPARC, a plug-and-play framework for generalized category discovery that improves assignment stability through spectral regularization, combining image-level consistency, token-level consistency, and stable semantic token selection.
Critical strengths
- The paper explores spectral assignment consistency in GCD, providing a meaningful perspective on reducing sensitivity to appearance variations.
- Extensive experiments across multiple datasets and backbones demonstrate consistent average improvements on fine-grained benchmarks.
Critical weaknesses
- Although the paper emphasizes spectral consistency as its central contribution, Table 3 shows that SST alone provides the largest individual performance improvement. This raises concerns about whether the main gains originate from spectral regularization or simply from replacing the CLS token with a more discriminative local representation.
- Table 3 suggests complementary effects among ICR, SST, and TCR, but notably omits the ICR+TCR configuration without SST readout replacement. This missing combination prevents a complete assessment of the interaction between the two proposed spectral regularizers and makes it difficult to isolate their contributions from SST.
- Several related modular GCD approaches have emerged during 2025–2026. Since SPARC is explicitly presented as a plug-and-play framework, comparisons should evaluate these methods under the same host architecture and experimental settings, rather than relying primarily on standalone performance comparisons.
- Although SPARC improves its host baselines on average, its absolute performance on coarse-grained benchmarks remains below several existing approaches, with some configurations showing negligible or negative improvements. The paper lacks a convincing explanation of why the proposed spectral regularization is substantially more effective on fine-grained datasets and whether its benefits are fundamentally task-dependent.
Questions for authors
please see weaknesses.
2 comments
Response to Reviewer uzgu (2/2)
[Q3] Plug-and-play
Thank you for your comments. Our earlier emphasis on standalone absolute numbers was mainly to show that SPARC can push the accuracy boundary of fine-grained GCD; for a plug-and-play claim, a same-host protocol is the fairer yardstick, and we adopt it here. We report modular add-ons on fixed SimGCD and SelEx hosts (DINO / DINOv2), with host-relative , against recent modular methods (e.g., APL, ConGCD, LEAH, CPF-GCD). SimGCD and SelEx represent parametric versus non-parametric pipelines.
From this table, we can observe that our SPARC consistently improves over both SimGCD and SelEx hosts under DINO and DINOv2. For example, on DINO + SimGCD, SPARC raises the average All ACC from 56.1 to 64.7, outperforming APL (60.4), ConGCD (57.1), LEAH (64.4), and CPF-GCD (59.1). Similar gains are observed on SelEx and DINOv2, showing that SPARC is a general and effective plug-and-play module.
Table. Performace on different hosts and backbones
| Backbone | Method | Venue | CUB | Stanford-Cars | FGVC-Aircraft | Average |
|---|---|---|---|---|---|---|
| DINO | SimGCD | ICCV 2023 | 60.3 / 65.6 / 57.7 | 53.8 / 71.9 / 45.0 | 54.2 / 59.1 / 51.8 | 56.1 / 65.5 / 51.5 |
| DINO | + APL | CVPR 2025 | 64.5 / 68.1 / 62.1 | 60.1 / 77.6 / 51.2 | 56.6 / 60.2 / 54.8 | 60.4 / 68.6 / 56.0 |
| DINO | + ConGCD | ICCV 2025 | 61.6 / 65.1 / 59.5 | 54.5 / 72.2 / 47.8 | 55.0 / 58.4 / 53.1 | 57.1 / 65.2 / 53.5 |
| DINO | + LEAH | AAAI 2026 | 69.4 / 76.3 / 65.9 | 65.5 / 80.4 / 58.2 | 58.3 / 64.7 / 55.1 | 64.4 / 73.8 / 59.7 |
| DINO | + CPF-GCD | ICML 2026 | 63.9 / 65.5 / 63.1 | 57.6 / 73.7 / 49.8 | 55.8 / 59.1 / 54.2 | 59.1 / 66.1 / 55.7 |
| DINO | + SPARC | – | 68.0 / 71.7 / 66.1 | 67.6 / 81.2 / 61.0 | 58.5 / 66.2 / 54.7 | 64.7 / 73.0 / 60.6 |
| Inc. | +7.7 / +6.1 / +8.4 | +13.8 / +9.3 / +16.0 | +4.3 / +7.1 / +2.9 | +8.6 / +7.5 / +9.1 | ||
| DINO | SelEx | ECCV 2024 | 73.6 / 75.3 / 72.8 | 58.5 / 75.6 / 50.3 | 57.1 / 64.7 / 53.3 | 63.1 / 71.9 / 58.8 |
| DINO | + ConGCD | ICCV 2025 | 81.7 / 80.4 / 82.4 | 57.5 / 77.5 / 47.9 | 62.5 / 70.2 / 58.7 | 67.3 / 76.0 / 63.0 |
| DINO | + CPF-GCD | ICML 2026 | 79.8 / 80.5 / 79.4 | 58.9 / 79.0 / 49.2 | 61.8 / 69.0 / 58.1 | 66.8 / 76.2 / 62.2 |
| DINO | + SPARC | – | 79.6 / 76.5 / 81.1 | 68.6 / 76.1 / 65.0 | 62.4 / 67.1 / 60.1 | 70.2 / 73.2 / 68.7 |
| Inc. | +6.0 / +1.2 / +8.3 | +10.1 / +0.5 / +14.7 | +5.3 / +2.4 / +6.8 | +7.1 / +1.3 / +9.9 | ||
| DINOv2 | SimGCD | ICCV 2023 | 71.5 / 78.1 / 68.3 | 71.5 / 81.9 / 66.6 | 63.9 / 69.9 / 60.9 | 69.0 / 76.6 / 65.3 |
| DINOv2 | + APL | CVPR 2025 | 75.1 / 79.1 / 73.2 | 73.4 / 87.6 / 66.7 | 68.8 / 74.1 / 66.6 | 72.4 / 80.3 / 68.8 |
| DINOv2 | + SPARC | – | 80.3 / 82.8 / 79.0 | 79.8 / 90.5 / 74.7 | 73.9 / 74.9 / 73.3 | 78.0 / 82.7 / 75.7 |
| Inc. | +8.8 / +4.7 / +10.7 | +8.3 / +8.6 / +8.1 | +10.0 / +5.0 / +12.4 | +9.0 / +6.1 / +10.4 | ||
| DINOv2 | SelEx | ECCV 2024 | 87.4 / 85.1 / 88.5 | 82.2 / 93.7 / 76.7 | 79.8 / 82.3 / 78.6 | 83.1 / 87.0 / 81.3 |
| DINOv2 | + ConGCD | ICCV 2025 | 86.3 / 87.4 / 85.8 | 79.8 / 93.1 / 73.3 | 81.7 / 83.3 / 81.0 | 82.6 / 87.9 / 80.0 |
| DINOv2 | + SPARC | – | 90.5 / 88.6 / 91.5 | 84.0 / 93.5 / 79.7 | 81.6 / 83.1 / 80.8 | 85.4 / 88.4 / 84.0 |
| Inc. | +3.1 / +3.5 / +3.0 | +1.8 / -0.2 / +3.0 | +1.8 / +0.8 / +2.2 | +2.3 / +1.4 / +2.7 |
[Q4] Performance on coarse-grained benchmarks
Thank you for your insightful comments. Indeed, the magnitude of the gain is task-dependent, but our SPARC improves the host on both fine- and coarse-grained benchmarks on average. From the rightmost 3 columns of both Tab.1 and Tab.2, we can observe consistent improvements of our methods across different datasets, hosts and backbones. The reason that SPARC does better on fine-grained datasets is likely that fine-grained categories often differ in local parts and textures; SST + ICR/TCR directly target this local-cue bottleneck, which leads to larger gains. However, on coarse-grained data, classes are already well separated by global semantics, so there is little headroom left. Additional local spectral constraints therefore bring only modest margins and can occasionally over-regularize. Nevertheless, we still observe notable improvements - for example, + 4.0 in All and + 4.8 in New on SelEx + SPARC with DINO as well as + 2.8 in All and +3.7 in New on SimGCD + SPARC with DINOv2 for ImageNet-100 - and the overall average improvement remains positive.
We hope this addresses your concerns and welcome any follow-up questions.
Response to Reviewer uzgu (1/2)
[Q1] Ablation
Thank you for your constructive reivew. Your concern relates to the effectiveness of two distinct pieces of SPARC: (i) the Stable Semantic Token (SST) and (ii) spectral assignment regularizers (ICR/TCR). Firstly, we would like to emphasize that SST itself is a core contribution, not a side effect to be explained away. It provides a frequency-domain route to more discriminative local cues. To our knowledge, this form of spectral token selection for GCD has not appeared in prior work. Secondly, from Tab. 3 we can conclude that the spectral regularizers still help beyond the readout, even though SST alone already provide a huge improvement. Holding the readout fixed, adding ICR+TCR lifts SST from 61.1 to 64.7 Avg All (CUB 63.3→68.0, Stanford-Cars 63.0→67.6, FGVC-Aircraft 57.1→58.5). Thirdly, we further conduct an ablation on the SelEx host. In this scenario, SST is not used as a readout, which allows us to clearly figure out the contribution of spectral regularizers. In this table, by comparing the second and third row with the baseline (first row), we can see that each one regularizer alone yields an improvemen. Besides, Avg All rises from 63.1 to 70.2 (CUB 73.6→79.6, Stanford-Cars 58.5→68.6, FGVC-Aircraft 57.1→62.4) with two regularizers working together.
Table. Component ablation (SelEx host, DINO)
ICR/TCR ablation on SelEx (DINO); SST is not applied as a CLS replacement. Metrics are All / Old / New ACC (%).
[Q2] Ablation
Thank you for pointing this out. We would like to showcase the full component grid in the SimGCD host ablation table below, including the ICR+TCR configurtaion without SST. With the complete 3 components ablation table, we observe that the qualitative picture is consistent with the paper, that is, three components are complementary and generally most effective when used jointly.
Additionlly, since SST is not as a readout for SelEx, we provide a 2 components ablation on SelEx as above, which perfectly isolates their contribution from SST. This table shows that ICR and TCR both help improve accuracy for SelEx and that they works better together (Average All moves from 63.1 to 66.8 with ICR, to 68.7 with TCR, and to 70.2 with both).
Table. Component ablation (SimGCD host, DINO)
Component ablation of ICR, SST, and TCR on CUB, Stanford-Cars, and FGVC-Aircraft. Metrics are All / Old / New ACC (%).