SEIF: Self-Evolving Reinforcement Learning for Instruction Following

TL;DR

SEIF introduces a self-evolving reinforcement learning framework that enhances large language models' instruction-following abilities by dynamically generating and filtering challenging instructions, leading to consistent improvements across models.

Contribution

The paper presents a novel self-evolving framework with four roles that co-evolve to improve instruction-following in LLMs, addressing limitations of static or costly supervision methods.

Findings

SEIF consistently improves instruction-following across multiple models.

Dynamic instruction difficulty evolution enhances model learning.

Effective training strategy involves early solid foundation followed by moderation.

Abstract

Instruction following is a fundamental capability of large language models (LLMs), yet continuously improving this capability remains challenging. Existing methods typically rely either on costly external supervision from humans or strong teacher models, or on self-play training with static-difficulty instructions that cannot evolve as the model's capabilities improve. To address these limitations, we propose SEIF (Self-Evolving Reinforcement Learning for Instruction Following), a self-evolving framework for enhancing the instruction-following ability of LLMs. SEIF forms a closed self-evolution loop that improves the model's instruction-following ability, where instruction difficulty evolution and model capability evolution reinforce each other. SEIF consists of four roles: an Instructor that generates increasingly challenging instructions, a Filter that removes conflicting or invalid instructions to ensure data quality, a Follower that learns to follow evolved instructions, and a Judger that provides reward signals for reinforcement learning. The Instructor and Follower are alternately trained and co-evolve throughout the process. Experiments across multiple model scales and architectures show that SEIF consistently improves instruction-following performance, suggesting strong generality. Further analyses reveal the sources of improvement and identify an effective training strategy for self-evolution on open-ended tasks: sufficient early-stage training to build a solid foundation, followed by moderate late-stage training to mitigate overfitting and achieve better final performance. The code and data are publicly available at https://github.com/Rainier-rq1/SEIF.