Divide-Fuse-Conquer: Eliciting "Aha Moments" in Multi-Scenario Games
This work addresses generalization issues in multi-scenario games for reinforcement learning with LLMs, representing an incremental improvement over existing methods.
The paper tackles the challenge of poor generalization in multi-scenario reinforcement learning for large language models by proposing the Divide-Fuse-Conquer framework, which groups games, trains specialized models, and fuses parameters, resulting in Qwen2.5-32B-Align achieving performance comparable to Claude3.5 with 7 wins and 4 draws across 18 TextArena games.
Large language models (LLMs) have been observed to suddenly exhibit advanced reasoning abilities during reinforcement learning (RL), resembling an ``aha moment'' triggered by simple outcome-based rewards. While RL has proven effective in eliciting such breakthroughs in tasks involving mathematics, coding, and vision, it faces significant challenges in multi-scenario games. The diversity of game rules, interaction modes, and environmental complexities often leads to policies that perform well in one scenario but fail to generalize to others. Simply combining multiple scenarios during training introduces additional challenges, such as training instability and poor performance. To overcome these challenges, we propose Divide-Fuse-Conquer, a framework designed to enhance generalization in multi-scenario RL. This approach starts by heuristically grouping games based on characteristics such as rules and difficulties. Specialized models are then trained for each group to excel at games in the group is what we refer to as the divide step. Next, we fuse model parameters from different groups as a new model, and continue training it for multiple groups, until the scenarios in all groups are conquered. Experiments across 18 TextArena games show that Qwen2.5-32B-Align trained with the Divide-Fuse-Conquer strategy reaches a performance level comparable to Claude3.5, achieving 7 wins and 4 draws. We hope our approach can inspire future research on using reinforcement learning to improve the generalization of LLMs.