Gespeichert in:
Bibliographische Detailangaben
Hauptverfasser: Liu, Yihao, Li, Shuocheng, Cao, Lang, Xie, Yuhang, Zhou, Mengyu, Dong, Haoyu, Ma, Xiaojun, Han, Shi, Zhang, Dongmei
Format: Preprint
Veröffentlicht: 2025
Schlagworte:
Online-Zugang:https://arxiv.org/abs/2506.01096
Tags: Tag hinzufügen
Keine Tags, Fügen Sie den ersten Tag hinzu!
_version_ 1866915434154950656
author Liu, Yihao
Li, Shuocheng
Cao, Lang
Xie, Yuhang
Zhou, Mengyu
Dong, Haoyu
Ma, Xiaojun
Han, Shi
Zhang, Dongmei
author_facet Liu, Yihao
Li, Shuocheng
Cao, Lang
Xie, Yuhang
Zhou, Mengyu
Dong, Haoyu
Ma, Xiaojun
Han, Shi
Zhang, Dongmei
contents Large language models are increasingly used for complex reasoning tasks where high-quality offline data such as expert-annotated solutions and distilled reasoning traces are often available. However, in environments with sparse rewards, reinforcement learning struggles to sample successful trajectories, leading to inefficient learning. At the same time, these offline trajectories that represent correct reasoning paths are not utilized by standard on-policy reinforcement learning methods. We introduce SuperRL, a unified training framework that adaptively alternates between RL and SFT. Whenever every rollout for a given instance receives zero reward, indicating the absence of a learning signal, SuperRL falls back to SFT on the curated offline data. Extensive experiments across diverse reasoning benchmarks show that SuperRL surpasses vanilla RL by delivering higher sample efficiency, stronger generalization, and improved robustness under sparse rewards.
format Preprint
id arxiv_https___arxiv_org_abs_2506_01096
institution arXiv
publishDate 2025
record_format arxiv
spellingShingle SuperRL: Reinforcement Learning with Supervision to Boost Language Model Reasoning
Liu, Yihao
Li, Shuocheng
Cao, Lang
Xie, Yuhang
Zhou, Mengyu
Dong, Haoyu
Ma, Xiaojun
Han, Shi
Zhang, Dongmei
Artificial Intelligence
Large language models are increasingly used for complex reasoning tasks where high-quality offline data such as expert-annotated solutions and distilled reasoning traces are often available. However, in environments with sparse rewards, reinforcement learning struggles to sample successful trajectories, leading to inefficient learning. At the same time, these offline trajectories that represent correct reasoning paths are not utilized by standard on-policy reinforcement learning methods. We introduce SuperRL, a unified training framework that adaptively alternates between RL and SFT. Whenever every rollout for a given instance receives zero reward, indicating the absence of a learning signal, SuperRL falls back to SFT on the curated offline data. Extensive experiments across diverse reasoning benchmarks show that SuperRL surpasses vanilla RL by delivering higher sample efficiency, stronger generalization, and improved robustness under sparse rewards.
title SuperRL: Reinforcement Learning with Supervision to Boost Language Model Reasoning
topic Artificial Intelligence
url https://arxiv.org/abs/2506.01096