Research map / Architectures and efficiency / Spiking networks
Otters: An Energy-Efficient Spiking Transformer via Optical Time-to-First-Spike Encoding
Zhanglu Yan, Jiayi Mao, Qianhui Liu, Fanfan Li, Tao Luo, Gang Pan, Bowen Zhu, Weng-Fai Wong
Read the paperCheck your idea against 137,217 papers
AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.
Measured effects reported in the paper
| What was changed | Where | Measured | Effect |
|---|---|---|---|
| Use optoelectronic decay plus 1-bit KV instead of traditional digital TTFS encoding (ablation) | Otters attention block · SST-2 energy | total energy (mJ): 5.12 -> 4.06 mJ (26.1% more energy for traditional TTFS) | Improved |
| Reduce KV precision from 4-bit to 1-bit (ablation) | Otters · SST-2 | energy and accuracy: energy 4.49 -> 4.06 mJ; accuracy 91.51 -> 91.28 | Mixed |
| Add hardware-aware training (10% noise) to Otters (ablation) | Otters · SST-2, noise 0.12 on O(t) | accuracy: 73.8 -> 85.3 | Improved |
| Add hardware-aware training (20% noise) to Otters (ablation) | Otters · SST-2, noise 0.20 on O(t) | accuracy: 53.3 -> 80.8 | Improved |
| Add hardware-aware training at low noise (ablation) | Otters · SST-2, noise 0.04 on O(t) | accuracy: 89.9 -> 89.3 (HAT1), 87.4 (HAT2) | Hurt |
| Use Otters 1-bit conversion versus 1-bit Sorbet SNN | spiking BERT · GLUE average | average score: 79.80 -> 83.22 | Improved |
| Use Otters instead of SpikingLM on energy | attention block · SST-2 | total energy (mJ): 7.2 -> 4.06 | Improved |
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 2 of 2 on most shared results.
- SpikingBERT — SpikingBERT: Distilling BERT to Train Spiking Language Models Using Implicit Differentiation (AAAI 2024) · better on 4 of 4 shared results
- Sorbet — Sorbet: A Neuromorphic Hardware-Compatible Transformer-Based Spiking Language Model (ICML 2025) · better on 1 of 1 shared result
Matched by method name in results tables read by AI. Authors' own tables tend to favour the newer method, and a comparison can be missed when a method is renamed.
Closest papers by meaning
- Are Conventional SNNs Really Efficient? A Perspective from Network Quantization (CVPR 2024 · similarity 0.88)
- Spike-driven Transformer V2: Meta Spiking Neural Network Architecture Inspiring the Design of Next-generation Neuromorphic Chips (ICLR 2024 · similarity 0.87)
- Adaptive Calibration: A Unified Conversion Framework of Spiking Neural Networks (AAAI 2025 · similarity 0.87)
- Spiking Token Mixer: An event-driven friendly Former structure for spiking neural networks (NeurIPS 2024 · similarity 0.86)
- The Dormant Spiking Neuron: A State-Driven Mechanism for Efficient Spiking Neural Networks (NeurIPS 2026 · similarity 0.86)
- High-Performance Temporal Reversible Spiking Neural Networks with $\mathcal{O}(L)$ Training Memory and $\mathcal{O}(1)$ Inference Cost (ICML 2024 · similarity 0.86)
Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.
