paper-with-me

홈 › Papers

Does Reinforcement Learning Really Incentivize Reasoning Capacity in LLMs Beyond the Base Model?

2025-04-18 · Zhiqi Chen, Rui Lu, Andrew Zhao, Zhaokai Wang, Yang Yue, Shiji Song, Gao Huang

Reinforcement Learning with Verifiable Rewards (RLVR) has recently demonstrated notable success in enhancing the reasoning performance of large language models (LLMs), particularly on mathematics and programming tasks. Similar to how traditional RL helps agents explore and learn new strategies, RLVR is believed to enable LLMs to continuously self-improve, thus acquiring novel reasoning abilities beyond those of the corresponding base models. In this study we critically examine the current state of RLVR by systematically probing the reasoning capability boundaries of RLVR-trained LLMs across various model families, RL algorithms, and math, coding, and visual reasoning benchmarks, using pass@k at large k values as the evaluation metric. Surprisingly, we find that the current training setup does not elicit fundamentally new reasoning patterns. While RLVR-trained models outperform their base models at small k (e.g., k = 1), the base models achieve a higher pass@k score when k is large. Coverage and perplexity analyses show that the observed reasoning abilities originate from and are bounded by the base model. Treating the base model as an upper bound, our quantitative analysis shows that six popular RLVR algorithms perform similarly and remain far from optimal in leveraging the potential of the base model. By contrast, we find that distillation can introduce new reasoning patterns from the teacher and genuinely expand the model's reasoning capabilities. Overall, our findings suggest that current RLVR methods have not yet realized the potential of RL to elicit truly novel reasoning abilities in LLMs. This highlights the need for improved RL paradigms, such as continual scaling and multi-turn agent-environment interaction, to unlock this potential.

📄 PDF Abstract BibTeX arXiv:2504.13837

Code (0)

등록된 구현이 없습니다.

Tasks

MathVisual Reasoning

Methods 이 논문이 사용한 방법론

BASE 설명 없음

Similar Papers 제목 키워드 기반

Curriculum Reinforcement Learning Can Incentivize Reasoning Capacity in LLMs Beyond the Base Model

2026-06-21 · Pengxiang Cai, Tianchen Fang, Xiaohan Li, Qingyuan Zeng 외 arxiv

Reinforcement learning with verifiable rewards (RLVR) is widely viewed as a promising path toward continuously improving large language models. Recent works, however, suggest that mainstream RLVR often reallocates sampli…

Reinforcement Learning

Incentivizing Reasoning from Weak Supervision

2025-05-26 · Yige Yuan, Teng Xiao, Shuchang Tao, Xue Wang 외

Large language models (LLMs) have demonstrated impressive performance on reasoning-intensive tasks, but enhancing their reasoning abilities typically relies on either reinforcement learning (RL) with verifiable signals o…

reinforcement-learningReinforcement LearningReinforcement Learning (RL)

Diversity-Incentivized Exploration for Versatile Reasoning

2025-09-30 · Zican Hu, Shilin Zhang, Yafu Li, Jianhao Yan 외 arxiv

Reinforcement Learning with Verifiable Rewards (RLVR) has emerged as a crucial paradigm for incentivizing reasoning capabilities in Large Language Models (LLMs). Due to vast state-action spaces and reward sparsity in rea…

Reinforcement Learning

Metis-RISE: RL Incentivizes and SFT Enhances Multimodal Reasoning Model Learning

2025-06-16 · Haibo Qiu, Xiaohan Lan, Fanfan Liu, Xiaohu Sun 외

Recent advancements in large language models (LLMs) have witnessed a surge in the development of advanced reasoning paradigms, which are now being integrated into multimodal large language models (MLLMs). However, existi…

Multimodal ReasoningReinforcement Learning (RL)

Reinforcement Learning with Verifiable Rewards Implicitly Incentivizes Correct Reasoning in Base LLMs

2025-06-17 · Xumeng Wen, Zihan Liu, Shun Zheng, Zhijian Xu 외

Reinforcement Learning with Verifiable Rewards (RLVR) has emerged as a promising paradigm for advancing the reasoning capabilities of Large Language Models (LLMs). However, a critical paradox clouds its efficacy: RLVR-tu…