Target Policy Optimization
This addresses optimization instability in RL for AI practitioners, though it appears incremental as it builds on existing policy-gradient frameworks.
The paper tackles the problem of policy-gradient methods overshooting or undershooting updates in reinforcement learning by introducing Target Policy Optimization (TPO), which separates the selection of completions from parameter updates, resulting in matching or outperforming methods like PG, PPO, GRPO, and DG, especially under sparse reward conditions.
In RL, given a prompt, we sample a group of completions from a model and score them. Two questions follow: which completions should gain probability mass, and how should the parameters move to realize that change? Standard policy-gradient methods answer both at once, so the update can overshoot or undershoot depending on the learning rate, clipping, and other optimizer choices. We introduce \emph{Target Policy Optimization} (TPO), which separates the two questions. Given scored completions, TPO constructs a target distribution $q_i \propto p_i^{\,\mathrm{old}} \exp(u_i)$ and fits the policy to it by cross-entropy. The loss gradient on sampled-completion logits is $p^θ- q$, which vanishes once the policy matches the target. On tabular bandits, transformer sequence tasks, and billion-parameter LLM RLVR, TPO matches PG, PPO, GRPO, and DG on easy tasks and substantially outperforms them under sparse reward. Code is available at https://github.com/JeanKaddour/tpo.