Weiqin Chen, Xinmeng Huang, Dongsheng Ding, Santiago Paternain
原文摘要
Safe reinforcement learning from human feedback (Safe-RLHF) seeks to improve helpfulness while satisfying harmlessness constraints. Existing primal-dual approaches can converge to feasible policies but do not prevent constraint violations during training. We study a stronger requirement, safe exploration for RLHF, under which every policy executed during training, including auxiliary exploration policies, remains feasible. We propose Zeroth-Order Log-Barrier Policy Gradient (ZO-LBPG), a reward-model-free method that optimizes a log-barrier objective using zeroth-order estimates derived directly from human preferences. To ensure safety under noisy feedback, we combine conservative constraint margin estimates with coupled rollouts and an adaptive step size that prevents updates from crossing constraint boundaries. We prove that, with high probability, all iterates and exploration policies remain strictly feasible, and that ZO-LBPG globally converges to an approximately optimal constrained policy. We establish a sample complexity of $\widetilde{\mathcal O}(\varepsilon^{-6}+d^3\varepsilon^{-5})$ trajectories and $\widetilde{\mathcal O}(\varepsilon^{-8}+d^5\varepsilon^{-7})$ human queries. Experiments on a synthetic CMDP and a language-model alignment task corroborate the training-time safety guarantees.
Tao Wang, Shuo Li, Yan Sun, Dongsheng Ding, Edgar Dobriban
原文摘要
Reinforcement learning with verifiable rewards (RLVR) has emerged as a central paradigm for improving the reasoning capabilities of large language models. Group-based policy optimization methods, such as GRPO, typically allocate a fixed number of rollouts to every prompt. This uniform allocation can be inefficient: it over-allocates compute to prompts whose sampled groups are already saturated while under-exploring prompts for which additional samples may reveal useful correct trajectories. To address this limitation, we introduce hit utility, the posterior probability that at least one rollout in a proposed additional allocation for a prompt will be correct. Building on this notion, we propose Hit-Utility Optimal Rollout Allocation (HORA), a learning-free rollout allocation policy that maximizes total posterior hit utility within each allocation batch. HORA adaptively reallocates rollout budgets while leaving the downstream reward evaluation and group-based advantage estimator unchanged. Across two training sets, four mathematical reasoning benchmarks, and three model families at multiple scales, HORA preserves comparable Pass@1 and improves Pass@K over compute-matched GRPO, and it outperforms related allocation baselines at large K. It is also drop-in compatible with other group-based estimators such as RLOO. Ablation studies indicate that the uniform prior used by HORA is competitive with five prompt-conditioned learned-prior alternatives.