Research map / Probabilistic and scientific modeling / Variational and generative sampling
SIXO: Smoothing Inference with Twisted Objectives
Dieterich Lawson, Allan Raventos, Andrew Warrington, Scott Linderman
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 FIVO with SIXO-q (single future observation) | FIVO · Stochastic volatility model | train 4-particle bound: 6921.29 -> 6928.90 | Improved |
| replace FIVO with SIXO-DRE (all future observations) | FIVO · Stochastic volatility model | train 4-particle bound: 6921.29 -> 6931.51 | Improved |
| replace FIVO with SIXO-DRE | FIVO · Stochastic volatility model | train BPF 2048-particle bound: 7020.14 -> 7019.42 | No clear effect |
| replace FIVO with SIXO-DRE | FIVO · Stochastic volatility model | test BPF 2048-particle bound: 3352.71 -> 3354.27 | Improved |
| replace FIVO with SIXO-q | FIVO · Stochastic volatility model | test BPF 2048-particle bound: 3352.71 -> 3353.45 | Improved |
| replace FIVO with SIXO | FIVO · Hodgkin-Huxley model | test method 256-particle bound: -50.25 -> -47.94 | Improved |
| replace FIVO with SIXO | FIVO · Hodgkin-Huxley model | relative parameter error: 0.46 -> 0.02 | Improved |
| add gradient clipping to FIVO | FIVO · Hodgkin-Huxley model | relative parameter error: 0.46 -> 0.63 | Hurt |
| add gradient clipping to FIVO | FIVO · Hodgkin-Huxley model | test method 256-particle bound: -50.25 -> -51.43 | Hurt |
| use DRE-trained twists in SMC instead of bootstrap particle filter | bootstrap particle filter · Hodgkin-Huxley synthetic data | log marginal lower bound: -74.89 -> -73.88 | Improved |
| use SIXO-u instead of FIVO | FIVO · Gaussian drift diffusion | bound gap | Improved |
| use SIXO-DRE (biased lower-variance gradients) instead of SIXO-u | SIXO-u · Gaussian drift diffusion | gradient variance | Improved |
Later papers that compared with this method
1 later accepted paper reports SIXO as a baseline in their results tables; 0 of them report a better number on most of the results they share with it.
- NAS-X: Neural Adaptive Smoothing via Twisting (NeurIPS 2023 · better on 1 of 8 shared results)
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
- Online Variational Sequential Monte Carlo (ICML 2024 · similarity 0.87)
- Variational Inference for Sequential Data with Future Likelihood Estimates (ICML 2020 · similarity 0.86)
- Divide and Couple: Using Monte Carlo Variational Objectives for Posterior Approximation (NeurIPS 2019 · similarity 0.86)
- Reverse Diffusion Sequential Monte Carlo Samplers (NeurIPS 2025 · similarity 0.85)
- Probabilistic Inference in Language Models via Twisted Sequential Monte Carlo (ICML 2024 · similarity 0.85)
- VISA: Variational Inference with Sequential Sample-Average Approximations (NeurIPS 2024 · similarity 0.84)
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.
