The Majority is not always right: RL training for solution aggregation
This work addresses the challenge of efficiently scaling test-time compute for LLMs in reasoning tasks, offering a novel method that is incremental but provides practical gains.
The paper tackles the problem of aggregating multiple candidate solutions from large language models to improve reasoning accuracy, proposing a reinforcement learning-trained aggregator that outperforms rule-based and reward-model baselines across benchmarks and generalizes to solutions from differing models while using fewer tokens.
Scaling up test-time compute, by generating multiple independent solutions and selecting or aggregating among them, has become a central paradigm for improving large language models (LLMs) on challenging reasoning tasks. While most prior work relies on simple majority voting or reward model ranking to aggregate solutions, these approaches may only yield limited benefits. In this work, we propose to learn aggregation as an explicit reasoning skill: given a set of candidate solutions, we train an aggregator model to review, reconcile, and synthesize a final, correct answer using reinforcement learning from verifiable rewards. A key ingredient is careful balancing of easy and hard training examples, allowing the model to learn both to recover minority-but-correct answers as well as easy majority-correct answers. Empirically, we find our method, AggLM, outperforms both strong rule-based and reward-model baselines, across multiple benchmarks. Furthermore, it generalizes effectively to solutions from differing models, including stronger ones than contained in the training data, all while requiring substantially fewer tokens than majority voting with larger numbers of solutions.