Research map / Architectures and efficiency / Attention and transformers
FRACTAL: State Space Model with Fractional Recurrent Architecture for Computational Temporal Analysis of Long Sequences
Mengqi Li, Wensheng Lin, Jinshuai Yang, Lixin Li
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 |
|---|---|---|---|
| replace uniform HiPPO measure with fractional power-law measure and multi-alpha filter bank | S5 · LRA ListOps | test accuracy (%): 61.10 -> 61.85 | Improved |
| replace S5 initialization with fractional measure | S5 · LRA Text | test accuracy (%): 88.72 -> 89.10 | Improved |
| replace S5 initialization with fractional measure | S5 · LRA average | test accuracy (%): 87.04 -> 87.11 | No clear effect |
| replace S5 initialization with fractional measure | S5 · LRA Path-X | test accuracy (%): 98.62 -> 98.39 | No clear effect |
| replace S5 initialization with fractional measure | S5 · LRA Image | test accuracy (%): 87.59 -> 87.30 | Hurt |
| replace S5 initialization with fractional measure | S5 · LRA Pathfinder | test accuracy (%): 95.04 -> 94.80 | Hurt |
| use analytic closed-form input projection initialization instead of random Gaussian (ablation) | FRACTAL · LRA training | training loss early epochs | Improved |
| use analytic closed-form input projection initialization instead of random Gaussian (ablation) | FRACTAL · LRA final accuracy | test accuracy (%) | No clear effect |
Closest papers by meaning
- Fractional State Space Transition for Long Sequence Modeling (NeurIPS 2026 · similarity 0.92)
- LongSpike: Fractional Order Spiking State Space Models for Efficient Long Sequence Learning (NeurIPS 2026 · similarity 0.84)
- Sequence Modeling with Spectral Mean Flows (NeurIPS 2025 · similarity 0.83)
- FACTS: A Factored State-Space Framework for World Modelling (ICLR 2025 · similarity 0.82)
- Quantifying Memory Utilization with Effective State-Size (ICML 2025 · similarity 0.82)
- Parallel-in-Time Training of Recurrent Neural Networks for Dynamical Systems Reconstruction (NeurIPS 2026 · similarity 0.82)
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.
