Writing-RL: Advancing Long-form Writing via Adaptive Curriculum Reinforcement Learning

TL;DR

This paper introduces Writing-RL, an adaptive reinforcement learning framework that enhances long-form writing in large language models by using curriculum strategies and discriminative rewards.

Contribution

It proposes a novel adaptive curriculum reinforcement learning framework with three key components to improve long-form writing beyond supervised fine-tuning.

Findings

Writing-RL improves long-form writing performance over strong SFT baselines.

Models trained with long-output RL generalize well to long-input reasoning tasks.

The framework effectively advances long-form writing capabilities in 7B-scale models.

Abstract

Recent advances in Large Language Models(LLMs) have enabled strong performance in long-form writing, but current training paradigms remain limited: Supervised Fine-Tuning (SFT) remains constrained by data saturation and performance ceilings, while Reinforcement Learning with Verifiable Reward (RLVR), though successful in verifiable domains like math and code, cannot be directly migrated to open-ended long-form writing due to a lack of ground-truths. To further advance long-form writing, we present Writing-RL: an Adaptive Curriculum Reinforcement Learning framework to advance long-form writing capabilities beyond SFT. The framework consists of three key components: Margin-aware Data Selection strategy that prioritizes samples with high learning potential, Pairwise Comparison Reward mechanism that provides discriminative learning signals in the absence of verifiable rewards, and Dynamic Reference Scheduling approach, which plays a critical role by adaptively adjusting task difficulty based on evolving model performance. Experiments on 7B-scale writer models show that Writing-RL effectively improves long-form writing performance over strong SFT baselines. Furthermore, we observe that models trained with long-output RL generalize surprisingly well to long-input reasoning tasks, potentially offering a promising perspective for rethinking long-context training.