Helmholtz-SDE closes the approximation gap in simulation-free latent SDE inference
canaesseth · x · 2026-10-06
- The paper (arXiv:2606.16138, Smith, Trippe & Linderman) tackles recovering dynamical systems from noisy observations via latent SDEs fitted with variational inference.
- The authors show existing simulation-free VI algorithms pay for efficiency: their parameterizations restrict the approximate posterior to a subset of SDEs available to simulation-based methods, degrading inference and parameter learning.
- Helmholtz-SDE optimizes over path laws compatible with prescribed marginals, recovers dynamics more faithfully—largest gains under high posterior uncertainty—and matches simulation-based VI at a fraction of the runtime. Code is available.
More from Research
- PerturBot Breaks Shortcut Priors in Vision-Language-Action Models With Perturbative Training — Mingyu Liu · 2026-10-06
- Georgia Tech's TextReg Fixes Prompt Distributional Overfitting, Gains Up to +11.8% OOD — GeorgiaTech · 2026-10-06
- SourceLearn Builds Source-Specific Agent Competence, Wins 13 of 15 Benchmarks — GeorgiaTech · 2026-10-06
- Representation-Space MMD Post-Training Boosts Diffusion LMs, More Parallel Decoding at 16B — yresearch · 2026-10-06
- Attention Relay makes embedding models instruction-aware without training via LLM attention weights — _reachsumit · 2026-10-06
- Programmatic Search Agents boost task success by up to 7.56 points over query-based agents — _reachsumit · 2026-10-06