Research map / Time series / Long-term forecasting
Liquid Structural State-Space Models
Ramin Hasani, Mathias Lechner, Tsun-Hsuan Wang, Makram Chahine, Alexander Amini, Daniela Rus
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 |
|---|---|---|---|
| add liquid kernel (PB, input-correlation terms) to S4-LegS | Liquid-S4 · Long Range Arena | average accuracy: 86.09 -> 87.32 | Improved |
| add liquid kernel (PB) to S4-LegS | Liquid-S4 · ListOps | accuracy: 59.60 -> 62.75 | Improved |
| add liquid kernel (PB) to S4-LegS | Liquid-S4 · Speech Commands full 35 labels, 16kHz | accuracy: 96.25 (best baseline S4D-Lin) -> 96.78 | Improved |
| add liquid kernel (PB) to S4-LegS | Liquid-S4 · Speech Commands full 35 labels, 8kHz | accuracy: 91.80 (S4D-Inv) -> 90.00 | Hurt |
| increase liquid order p (ablation) | Liquid-S4 · ListOps and IMDB | accuracy | Improved |
| add liquid kernel (PB) to S4-LegS | Liquid-S4 · BIDMC RR | RMSE: 0.247 -> 0.158 (vs S4-LegS) | Improved |
| add liquid kernel (PB) to S4-LegS | Liquid-S4 · sCIFAR | accuracy: 91.80 -> 92.02 | Improved |
| use simplified PB kernel instead of KB kernel (ablation) | Liquid-S4 · general | computational cost and performance | No clear effect |
| reduce state size to 7 (small kernel) for ListOps, IMDB, Speech Commands (ablation) | Liquid-S4 · ListOps, IMDB, Speech Commands | accuracy | No clear effect |
| use bidirectional instead of causal modeling (ablation) | Liquid-S4 · general | accuracy | Hurt |
Closest papers by meaning
- Elastic Spectral State Space Models for Train-Once Budgeted Inference (NeurIPS 2026 · similarity 0.86)
- State Space Models are Provably Comparable to Transformers in Dynamic Token Selection (ICLR 2025 · similarity 0.86)
- Parameter-Efficient Fine-Tuning of State Space Models (ICML 2025 · similarity 0.86)
- Structured Sparse Transition Matrices to Enable State Tracking in State-Space Models (NeurIPS 2025 · similarity 0.85)
- Longhorn: State Space Models are Amortized Online Learners (ICLR 2025 · similarity 0.85)
- HOPE for a Robust Parameterization of Long-memory State Space Models (ICLR 2025 · similarity 0.85)
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.
