dorsal/arxiv
View SchemaDPWriter: Reinforcement Learning with Diverse Planning Branching for Creative Writing
| Authors | Qian Cao, Yahui Liu, Wei Bi, Yi Zhao, Ruihua Song, Xiting Wang, Ruiming Tang, Guorui Zhou, Han Li |
|---|---|
| Categories | |
| ArXiv ID | 2601.09609vv1 |
| URL | https://arxiv.org/abs/2601.09609 |
| License | http://creativecommons.org/licenses/by-nc-sa/4.0/ |
Abstract
Reinforcement learning (RL)-based enhancement of large language models (LLMs) often leads to reduced output diversity, undermining their utility in open-ended tasks like creative writing. Current methods lack explicit mechanisms for guiding diverse exploration and instead prioritize optimization efficiency and performance over diversity. This paper proposes an RL framework structured around a semi-structured long Chain-of-Thought (CoT), in which the generation process is decomposed into explicitly planned intermediate steps. We introduce a Diverse Planning Branching method that strategically introduces divergence at the planning phase based on diversity variation, alongside a group-aware diversity reward to encourage distinct trajectories. Experimental results on creative writing benchmarks demonstrate that our approach significantly improves output diversity without compromising generation quality, consistently outperforming existing baselines.
{
"annotation_id": "9d69ff28-0848-4fff-ad07-6081ad8e6425",
"date_created": "2026-02-17T05:53:19.966000Z",
"date_modified": "2026-02-17T05:53:19.966000Z",
"file_hash": "2afef5662b4c090a637fd7e8e0354bad65abdf4c95733b925f22c4a63ac9f9da",
"private": false,
"record": {
"abstract": "Reinforcement learning (RL)-based enhancement of large language models (LLMs) often leads to reduced output diversity, undermining their utility in open-ended tasks like creative writing. Current methods lack explicit mechanisms for guiding diverse exploration and instead prioritize optimization efficiency and performance over diversity. This paper proposes an RL framework structured around a semi-structured long Chain-of-Thought (CoT), in which the generation process is decomposed into explicitly planned intermediate steps. We introduce a Diverse Planning Branching method that strategically introduces divergence at the planning phase based on diversity variation, alongside a group-aware diversity reward to encourage distinct trajectories. Experimental results on creative writing benchmarks demonstrate that our approach significantly improves output diversity without compromising generation quality, consistently outperforming existing baselines.",
"arxiv_id": "2601.09609",
"authors": [
"Qian Cao",
"Yahui Liu",
"Wei Bi",
"Yi Zhao",
"Ruihua Song",
"Xiting Wang",
"Ruiming Tang",
"Guorui Zhou",
"Han Li"
],
"categories": [
"cs.CL",
"cs.AI"
],
"license": "http://creativecommons.org/licenses/by-nc-sa/4.0/",
"title": "DPWriter: Reinforcement Learning with Diverse Planning Branching for Creative Writing",
"url": "https://arxiv.org/abs/2601.09609",
"version": "v1"
},
"schema_id": "dorsal/arxiv",
"source": {
"execution_id": "8a965d4f-3f45-430f-8b93-a5b044636762",
"id": "arXiv Dataset",
"type": "Model",
"variant": "snapshot-2026-01-17",
"version": "0.1.0"
},
"user_id": 1000002
}