Playpen: An Environment for Exploring Learning Through Conversational Interaction
This work addresses the challenge of enhancing LLM learning through conversational interaction, offering a new environment for research, though it is incremental in exploring existing methods on a novel feedback source.
The paper tackled the problem of using Dialogue Games as feedback signals for post-training LLMs, finding that interactive learning with GRPO showed balanced improvements without skill loss, while SFT improved performance on unseen instances but negatively impacted other skills.
Interaction between learner and feedback-giver has come into focus recently for post-training of Large Language Models (LLMs), through the use of reward models that judge the appropriateness of a model's response. In this paper, we investigate whether Dialogue Games -- goal-directed and rule-governed activities driven predominantly by verbal actions -- can also serve as a source of feedback signals for learning. We introduce Playpen, an environment for off- and online learning through Dialogue Game self-play, and investigate a representative set of post-training methods: supervised fine-tuning; direct alignment (DPO); and reinforcement learning with GRPO. We experiment with post-training a small LLM (Llama-3.1-8B-Instruct), evaluating performance on unseen instances of training games as well as unseen games, and on standard benchmarks. We find that imitation learning through SFT improves performance on unseen instances, but negatively impacts other skills, while interactive learning with GRPO shows balanced improvements without loss of skills. We release the framework and the baseline training setups to foster research in the promising new direction of learning in (synthetic) interaction.