Find a story
Search Spins
Search titles, summaries, and missing voices across published articles — press releases, announcements, and media coverage.
0 results for “safety constraints”
Boundary-Seeking Policy Gradient for Safe Reinforcement Learning
A new reinforcement learning algorithm called Boundary-Seeking Policy Gradient (BSPG) is introduced to improve safety-constrained optimization by explicitly guiding policies to the active constraint boundary—rather than settling inside the feasible region—yielding tighter constraint satisfaction and higher reward in simulation.
Aug 12, 2026
Safe Bayesian Optimization with Counterfactual Policies
Researchers introduced a new method called 'Safe Bayesian Optimization with Counterfactual Policies' that integrates conformal prediction to estimate uncertain counterfactual baselines, enabling optimization under safety constraints where the safe reference point is unobserved.
Jul 9, 2026
Safe Inference-Time Alignment via Lagrangian Reward Augmentation
A new research paper proposes Lagrangian Reward Augmentation (LARA), a framework to integrate explicit safety constraints into inference-time alignment of frozen language models by dualizing constrained optimization and calibrating a single dual variable on a small dataset.
Jul 8, 2026