D\'ej\`aQ: Open-Ended Evolution of Diverse, Learnable and Verifiable Problems

TL;DR

D'ejàQ introduces a dynamic framework that evolves diverse mathematical problems alongside model training, enhancing reasoning abilities by adapting problems based on the model's progress and employing LLM-driven mutations.

Contribution

It presents a novel evolutionary approach to generate and adapt training problems dynamically, moving beyond static datasets to improve reasoning in language models.

Findings

Generated meaningful, novel problems through LLM mutations

Improved reinforcement learning training with evolving problems

Analyzed problem validity and computational costs

Abstract

Recent advances in reasoning models have yielded impressive results in mathematics and coding. However, most approaches rely on static datasets, which have been suggested to encourage memorisation and limit generalisation. We introduce D\'ej\`aQ, a framework that departs from this paradigm by jointly evolving a diverse set of synthetic mathematical problems alongside model training. This evolutionary process adapts to the model's ability throughout training, optimising problems for learnability. We propose two LLM-driven mutation strategies in which the model itself mutates the training data, either by altering contextual details or by directly modifying problem structure. We find that the model can generate novel and meaningful problems, and that these LLM-driven mutations improve RL training. We analyse key aspects of D\'ej\`aQ, including the validity of generated problems and computational overhead. Our results underscore the potential of dynamically evolving training data to enhance mathematical reasoning and indicate broader applicability, which we will support by open-sourcing our code.