The findings support on-policy distillation as an effective warmup for agentic RLVR and identify initial reward discovery as a mechanism that can contribute to the observed acceleration.
Abstract
Reinforcement learning with a verifiable reward (RLVR) offers a scalable approach to training language-model agents, yet sparse outcome rewards can leave early training with little signal for policy improvement. We identify an On-Policy Acceleration Phenomenon: in our main comparisons, RLVR initialized with on-policy distillation reaches high performance earlier in training and achieves both higher average performance during subsequent RLVR and higher final performance than the alternative baselines. Motivated by this observation, we study On-Policy Warmup (OPW), a teacher-guided stage in which the student trains with teacher supervision on its own interaction trajectories before transitioning to RLVR. Unlike imitation on fixed teacher-generated trajectories, OPW targets states induced by the student's own decisions, including imperfect actions and recovery situations. We provide a theoretical explanation by connecting on-policy reverse-KL distillation to trajectory-level distribution matching. Under a competent teacher and sufficiently small population distillation loss, this connection yields a lower bound on initial verifier success and a corresponding bound on reward-discovery complexity. For group-relative RLVR, we further characterize when increased success probability produces more reward-informative groups. Together, our findings support on-policy distillation as an effective warmup for agentic RLVR and identify initial reward discovery as a mechanism that can contribute to the observed acceleration.
The results support OPD as preparation for RL and favor forward KL when OPD is followed by RL in the authors' comparison, which highlights the importance of both objective choice and the states receiving supervision for subsequent RL.
This work proposes self-instructing policy optimization (SIPO) with a contrastive self-teacher to provide dense credit, and demonstrates that SIPO outperforms both RLVR and OPSD baselines without an external teacher or additional generation.
Zhenrui Yue, Hui-Min Zeng, Yue-Qi Wang et al.· arXiv.org· 1 citation
This work reformulates the implicit reward of sampled-token OPD based on trajectory correctness, then applies a ReLU gating mechanism to ensure that correct trajectories receive non-negative rewards and incorrect ones receive non-positive rewards, making it readily combinable with any policy gradient algorithm, such as...
Modern language models undergo reinforcement learning with verifiable rewards (RLVR) on top of already-trained checkpoints. A key promise of RLVR is the discovery of new reasoning strategies. In principle, a model can sample novel ideas absent from its prior training data. In practice, however, augmenting RLVR with str...
This work proposes Reasoning-Progress-Aware Reward Filtering for On-Policy Distillation (R2-OPD), which constructs two within-trajectory rankings of reasoning spans, one from teacher-derived rewards and the other from independently estimated progress reward.
Chen Yang, Hai-Yuan Wan, Rengrong Xiong et al.· 2 citations
TIDE dynamically rebalanced teacher guidance and reward optimization should be dynamically rebalanced over training and jointly allocated across turns, and experiments support the effectiveness of TIDE's adaptive OPD--RL coordination.
AI coding agents are increasingly help developers fix bugs and build features largely on their own, but every interaction with their models consumes tokens. In each session, an agent sends instructions, tool descriptions, and session history to its model. Most efforts to reduce this cost focus on individual tasks. Yet developers return to the same […] The post TokenEconomizer: Learning from Experience to Make AI Coding Agents More Efficient appeared first on Microsoft Research.