Event
ICLR 2026 Paper Proposes DDCG and IVW-H Policy Gradient Methods
1 reports · 1 sources
An ICLR 2026 paper re-examines "whether differentiable simulators can yield better policy gradients," concluding that bias from discontinuous dynamics is not the main obstacle. The authors, Ku Onoda et al., propose a lightweight test, DDCG, which switches estimators in non-smooth regions; the paper claims it requires only one hyperparameter, performs robustly in controlled experiments, and remains reliable with small samples. The authors also propose per-step inverse variance implementation, IVW-H, for differentiable robot control tasks. The paper claims this method stabilizes variance and achieves good results without explicitly detecting discontinuities, and argues that in real-world deployment, variance control is often more critical than estimator switching.
Report timeline
- Can Differentiable Simulators Give Better Policy Gradients? ICLR 2026 Paper Proposes DDCG and IVW-H
An ICLR 2026 paper re-examines the question of whether differentiable simulators can give better policy gradients, arguing that bias from discontinuous dynamics is not the main obstacle. The authors propose DDCG, a lightweight test that switches estimators in nonsmooth regions; it requires only one hyperparameter and remains reliable with small samples. They also propose IVW-H, a per-step inverse-variance implementation that stabilizes variance in differentiable robot control tasks without explicit discontinuity detection. The paper concludes that estimator switching can improve robustness in controlled experiments, while in real deployment variance control is often more critical.
arXiv cs.RO ·