- Rating
- 2
- Confidence
- 4
- Informativeness
- 2
- Presentation
- 3
- Soundness
- 2
Read review
Paper summary
This paper addresses the concern that sparse query–key matching in Spiking Transformers may restrict long-range information exchange. It proposes Gated Spike Axial Propagation (GSAP), which uses axial propagation for global contextual interaction and receiver-conditioned gating to regulate context utilization. GSAP further incorporates a local feature pathway and an input residual connection to integrate local and global information while preserving spiking representations and temporal dynamics. Experiments demonstrate improvements over the baseline models on image classification, event-based recognition, and semantic segmentation. Ablation studies and token-perturbation analyses further investigate the contributions of individual components and the information propagation properties of the proposed mechanism.
Critical strengths
-
The paper evaluates GSAP on multiple static-image classification, event-based recognition, and semantic segmentation benchmarks, demonstrating its applicability across different tasks and backbone architectures.
-
The paper identifies a potential limitation of communication based on sparse query–key coincidences and presents a clearly motivated design that separates spatial context propagation from receiver-conditioned context selection.
Critical weaknesses
-
Insufficient comparison with stronger ImageNet baselines, such as QKFormer and SDT-v3 [1].
-
GSAP is structurally highly similar to ParCNetV2 [2]. Both employ large one-dimensional convolutional kernels along two spatial axes to aggregate information, a local feature branch, a gating-like branch, and related feature fusion operations.
-
The ablations are restricted to relatively small datasets, and several reported performance differences are below 0.2 percentage points. Without results across multiple random seeds or uncertainty estimates, it is difficult to determine whether these differences reflect reliable improvements or training variability. Moreover, the local-only variant performs worse than the baseline, leaving the standalone benefit of the local pathway unclear.
[1] Scaling Spike-driven Transformer with Efficient Spike Firing Approximation Training. TPAMI 2025.
[2] ParCNetV2: Oversized Kernel with Enhanced Attention. ICCV 2023.
Questions for authors
-
The paper states that SSA is severely affected by sparsity, and Figure 1 shows that SSA is sparser than GSAP. Why, then, does SSA consume more energy than GSAP?
-
What are the core methodological differences between GSAP and ParCNetV2, beyond the introduction of spiking neurons? Which experiments demonstrate the benefits of these differences?
-
How does GSAP affect temporal spike patterns?
5 comments
Response to Reviewer CM3Z (2/2)
Tab.8's network-wide structural comparison uses a uniform effective activity rate of . These supplementary Spikingformer token mixer estimates use actual spike activity from trained checkpoints:
| Dataset | Mixer | GFLOPs | eMAC (M) | SOP (B) | Energy (mJ) |
|---|---|---|---|---|---|
| ImageNet | SSA | 31.445 | 3.874 | 3.835 | 3.4697 |
| ImageNet | GSAP | 22.803 | 11.519 | 3.175 | 2.9109 |
| CIFAR-100 | SSA | 1.309 | 0.316 | 0.09448 | 0.086485 |
| CIFAR-100 | GSAP | 0.937 | 0.940 | 0.08476 | 0.080613 |
Both use : ImageNet uses 8–768 and 1,000 validation images; CIFAR-100 uses 4–384 and 10,000 test images. GFLOPs counts dense computation; eMAC is MAC-equivalent floating-point arithmetic. Energy coefficients are 4.6 pJ/eMAC and 0.9 pJ/SOP.
Relative to SSA on ImageNet and CIFAR-100, GSAP reduces arithmetic energy by 16.10% and 6.79%, SOPs by 17.21% and 10.28%, and dense computation by 27.48% and 28.43%, respectively. Savings from fewer spike accumulations outweigh the added floating-point arithmetic costs.
(b) Savings Mainly Come from Input Projection
The mixer energy breakdown (mJ/image) identifies the savings:
| Energy source | ImageNet SSA | ImageNet GSAP | CIFAR-100 SSA | CIFAR-100 GSAP |
|---|---|---|---|---|
| Input projection | 2.8117 | 1.7198 | 0.072870 | 0.048306 |
| Spatial aggregation | 0.0094 | 0.0282 | 0.000774 | 0.001584 |
| Output projection | 0.6307 | 1.1099 | 0.011386 | 0.026398 |
| Scaling and fusion | 0.0178 | 0.0530 | 0.001455 | 0.004325 |
| Total | 3.4697 | 2.9109 | 0.086485 | 0.080613 |
SSA's Q/K/V projections consume 81.04% and 84.26% of mixer arithmetic energy on ImageNet and CIFAR-100. GSAP's feature and gating projections save about 1.0919 and 0.0246 mJ/image, covering higher spatial aggregation, output projection, scaling and fusion costs, for net savings of 0.5588 and 0.005872 mJ/image, respectively. Lower input projection cost is the main source of savings on both datasets.
Under the same CIFAR-100 counting method, GSAP and the spiking adaptation of ParCNetV2 use 0.080613 and 0.092018 mJ/image of mixer arithmetic energy, respectively, a 12.39% reduction for GSAP.
4. Temporal Spike Patterns
(a) Firing and Temporal Persistence
We measured per-step firing before Block 8's output projection for Spikingformer-8-768 and GSAP-8-768 on the same 1,000 ImageNet validation images (). To assess sustained early-information propagation, we cleared GSAP's inter-axis LIF membrane state each step and tracked remote firing changes after removing first-step central-source spikes.
| Step | SSA firing rate | GSAP firing rate | Remote response: native GSAP | Remote response: state cleared |
|---|---|---|---|---|
| 1 | 3.0% | 18.9% | 49.7% | 49.7% |
| 2 | 8.5% | 39.2% | 50.1% | 33.0% |
| 3 | 10.1% | 36.3% | 46.1% | 26.3% |
| 4 | 11.3% | 42.0% | 41.7% | 22.4% |
GSAP has higher firing rates at every step. State clearing leaves first-step remote responses unchanged but weakens later responses, from 41.7% to 22.4% at step 4. Top-1 remains close (78.4% vs 78.2%, a gap of 0.2 pp). Thus, the inter-axis membrane state helps sustain early information's effects on later remote firing.
(b) Prediction Timing and Online Learning
Each model's original classification head uses the mean of the spatially pooled features over the first t simulation steps:
| Steps | SSA (%) | GSAP (%) | Gain and 95% interval (pp) |
|---|---|---|---|
| 1 | 53.3 | 63.5 | +10.2 [7.1, 13.3] |
| 2 | 70.9 | 75.6 | +4.7 [2.0, 7.2] |
| 3 | 73.0 | 76.8 | +3.8 [1.3, 6.1] |
| 4 | 74.9 | 78.4 | +3.5 [1.2, 5.7] |
GSAP leads by 10.2 pp at step 1 and stays ahead thereafter. SSA correctly classifies 50.6% of images at every step (1–4), versus 60.9% for GSAP, showing that GSAP makes correct predictions earlier and maintains them. The 95% intervals use paired image resampling at fixed checkpoints.
Training results are from the original Fig. 4 online-learning experiment with the GSAP propagation variant without receiver gating, using , 200 epochs and seed 2022.
| Dataset | OTTT-A: SSA → propagation variant (gain) | OTTT-O: SSA → propagation variant (gain) |
|---|---|---|
| CIFAR-10 | 82.51 → 83.97 (+1.46) | 89.30 → 91.19 (+1.89) |
| CIFAR-100 | 51.48 → 62.07 (+10.59) | 63.05 → 68.49 (+5.44) |
Accuracy is in %; gains in parentheses are in pp. OTTT-A accumulates gradients across steps before updating; OTTT-O updates each step. The propagation variant is more accurate under both. Fig. 4's mean cross-step gradient cosine similarity rises from 0.164 to 0.311, indicating more consistent cross-step gradient directions.
Thanks again for suggestions that improved our analysis and presentation. We hope this addresses your questions. We will add the relevant results and analyses in future versions for clearer, more complete arguments.
Thank you again for your thoughtful review. We wanted to follow up on your question about ParCNetV2. In the same Spikingformer backbone, GSAP and the spiking adaptation of ParCNetV2 achieved similar mean accuracy across three seeds, with lower estimated mixer arithmetic energy for GSAP on CIFAR-100.
We would be happy to clarify any remaining questions about the comparison or the other results in our rebuttal. If the additional evidence addresses your main concerns, we would appreciate it if you would reconsider your overall assessment.
Response to Reviewer CM3Z (1/2)
Thank you for noting GSAP's similarity to ParCNetV2 and asking about energy and spike timing. Below we address these questions and stronger ImageNet baselines, and give full three-seed ablations and additional temporal analyses.
1. Stronger ImageNet Architectures
We compared QKFormer and its GSAP variant at 10–384 under identical settings.
GSAP saves 0.78M parameters and about 76.1% attention energy, with an observed Top-1 gain of +0.17 percentage points (pp), showing successful adaptation to the stronger QKFormer with energy and performance gains. We plan to report QKFormer + GSAP results at more model scales in future versions.
SDT-V3 mainly optimizes SDT-series neuronal firing patterns. Like Spikingformer, it is a single-stage spiking Transformer. We hope to explore compatibility between GSAP and SDT-V3's neuronal mechanisms in future work.
2. ParCNetV2 and Component Benefits
(a) Structure and Direct Comparison
Both mixers use large axial kernels, local features and gating, with these structural differences:
Based on the official implementation, our ParCNetV2 spiking adaptation in Spikingformer retains shared projected features and joint gating, maintaining binary synaptic inputs. Three-seed results:
GSAP's mean accuracy gains over the spiking adaptation are 0.15, 0.74, and 0.01 pp, respectively. On CIFAR-100, its estimated energy is 12.39% lower: 0.080613 versus 0.092018 mJ/image under the same counting method.
(b) Three-Seed Ablations and Local Branch
Three-seed ablations examine the components and propagation path:
GSAP's mean gains over SSA are 0.25, 0.20, and 0.53 pp, respectively. Its mean accuracy is also higher than that of the parallel axial variant tested on each dataset.
The local branch supplies neighborhood detail; the global branch aggregates long-range information. On CIFAR-100, local-only yields 78.83%, below SSA's 79.97%; both branches yield 80.22%. Full GSAP improves global-only mean accuracy by 0.73, 0.64, and 0.16 pp, respectively, supporting effective joint local–global modeling.
The 2D variant replaces both axial operators, the inter-axis LIF and the row shortcut, so we compare the performance and cost of the full paths. Its mean accuracy is higher on CIFAR-100 and N-Caltech101; GSAP is better on CIFAR10-DVS. At k=15, axial kernels use 30C weights versus 225C for 2D, giving GSAP 86.7% fewer propagation kernel parameters. This illustrates the trade-off between accuracy and propagation kernel cost.
3. Sparse Matching and Energy
(a) Sparse Matching and Energy in SSA
Q–K matching sparsity and neuronal firing sparsity measure different properties: zero-inner-product token-pair rates versus zero-spike rates in neuronal outputs. For binary Q/K in one head and step, only co-firing in corresponding channels, (Qi,c,Kj,c)=(1,1), contributes nonzero values to QiKj⊤.
For Qi=(1,0) and Kj=(0,1), both groups fire but their inner product is zero: an unmatched token pair does not mean its tokens did not fire. Projections have already occurred; these spikes may match other tokens. To estimate total arithmetic energy, we sum costs across input projection, spatial aggregation, output projection, fusion and other stages, accounting for actual spike activity, operator scale and execution method.