Target Policy Optimization

TL;DR

Target Policy Optimization (TPO) is a new RL method that separates the construction of target distributions from policy fitting, improving performance especially in sparse reward settings.

Contribution

TPO introduces a novel approach that decouples target distribution creation from policy updates, enhancing stability and effectiveness in reinforcement learning tasks.

Findings

TPO matches existing methods on easy tasks.

TPO outperforms other algorithms under sparse rewards.

TPO is effective across various RL benchmarks.

Abstract

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 $p^\theta - q$, 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.