SEIF: Self-Evolving Reinforcement Learning for Instruction Following

Hugging Face Daily Papers Papers

Summary

This paper introduces SEIF, a self-evolving reinforcement learning framework that enhances LLM instruction-following capabilities through iterative difficulty adaptation and co-training of instructor and follower components.

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.
Original Article
View Cached Full Text

Cached at: 05/12/26, 07:29 AM

Paper page - SEIF: Self-Evolving Reinforcement Learning for Instruction Following

Source: https://huggingface.co/papers/2605.07465 Published on May 8

·

Submitted byhttps://huggingface.co/dd12345789

rainon May 12

Abstract

A self-evolving reinforcement learning framework enhances large language model instruction-following capabilities through iterative difficulty adaptation and co-training of instructor and follower components.

Instruction following is a fundamental capability oflarge 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-EvolvingReinforcement Learningfor Instruction Following), a self-evolving framework for enhancing theinstruction-followingability of LLMs. SEIF forms a closedself-evolutionloop that improves the model’sinstruction-followingability, whereinstruction difficulty evolutionandmodel capability evolutionreinforce 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 providesreward signalsforreinforcement 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 improvesinstruction-followingperformance, suggesting strong generality. Further analyses reveal the sources of improvement and identify an effective training strategy forself-evolutionon 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.

View arXiv pageView PDFGitHub2Add to collection

Models citing this paper0

No model linking this paper

Cite arxiv.org/abs/2605.07465 in a model README.md to link it from this page.

Datasets citing this paper0

No dataset linking this paper

Cite arxiv.org/abs/2605.07465 in a dataset README.md to link it from this page.

Spaces citing this paper0

No Space linking this paper

Cite arxiv.org/abs/2605.07465 in a Space README.md to link it from this page.

Collections including this paper0

No Collection including this paper

Add this paper to acollectionto link it from this page.

Similar Articles

SEAL: Synergistic Co-Evolution of Agents and Learning Environments

arXiv cs.CL

SEAL proposes a closed-loop framework for jointly evolving LLM agents and their training environments, using diagnosis-guided labels to align both sides. It achieves substantial gains in multi-turn tool-use tasks with only 400 training samples, demonstrating improved robustness and out-of-distribution transfer.

Improving instruction hierarchy in frontier LLMs

OpenAI Blog

OpenAI presents a training approach using instruction-hierarchy tasks to improve LLM safety and reliability by teaching models to properly prioritize instructions based on trust levels (system > developer > user > tool). The method addresses prompt-injection attacks and safety steerability through reinforcement learning with a new dataset called IH-Challenge.