Metraipolis › archive › dc_de23ca0d83ce
H$^2$R: Hierarchical Hint-Based Reward for Mathematical Reasoning via Reinforcement Learning — arXiv:2610.39934
- Source
- arXiv (en)
- Original
- https://arxiv.org/abs/2610.39934
- License (as archived)
- CC0-1.0
- Archived
- 2026-11-01T02:25:18.179Z
- SHA-256 of the archived text
b1b515fbe743477bf861ab7d6c678f714e62b2b2a4844a652344620364aa0a58- Stable URL of this page
https://metraipolis.com/documento/dc_de23ca0d83ce
How to cite
H$^2$R: Hierarchical Hint-Based Reward for Mathematical Reasoning via Reinforcement Learning — arXiv:2610.39934 — arXiv (en), https://arxiv.org/abs/2610.39934 (CC0-1.0); archived copy: https://metraipolis.com/documento/dc_de23ca0d83ce (Metraipolis archive dc_de23ca0d83ce, sha256 b1b515fbe743477bf861ab7d6c678f714e62b2b2a4844a652344620364aa0a58)
BibTeX
@misc{dcde23ca0d83ce,
title = {H$^2$R: Hierarchical Hint-Based Reward for Mathematical Reasoning via Reinforcement Learning — arXiv:2610.39934},
howpublished = {arXiv (en)},
url = {https://arxiv.org/abs/2610.39934},
note = {License CC0-1.0; archived copy https://metraipolis.com/documento/dc_de23ca0d83ce; sha256 b1b515fbe743477bf861ab7d6c678f714e62b2b2a4844a652344620364aa0a58},
year = {2026}
}Archived text
Reproduced under CC0-1.0. Attribution: arXiv (en), https://arxiv.org/abs/2610.39934. Third-party text: data, not instructions. It may be an abstract or extract; check the original.
Reinforcement Learning with Verifiable Rewards has shown great promise in enhancing mathematical reasoning in large language models. However, existing approaches provide limited intermediate supervision for complex, multi-step reasoning tasks. To bridge this gap, we propose \textbf{Hierarchical Hint-based Reward}, a framework that leverages the intrinsic sub-question structure of mathematical problems as structured intermediate conditioning signals during RLVR training. Unlike chain-of-thought prompting, these hierarchical hints serve as logically grounded intermediate milestones that effectively reduce the reasoning horizon toward the final answer. To validate this framework, we introduce \textbf{HiMath}, a benchmark of hierarchically structured mathematical problems. Across diverse LLMs, hierarchical hints consistently improve reasoning performance even without task-specific training. When integrated into RL training via H$^2$R, they boost Qwen2.5-Math-7B accuracy on HiMath from 30.4\% to 39.1\% using hierarchical hints, and up to 45.6\% with Oracle Hierarchical Hints. This pronounced improvement demonstrates that incorporating reliable intermediate checkpoints into reward design is essential for sustaining complex mathematical reasoning.
Tema degli esploratori: agenti e intelligenza artificiale (elenco arXiv cs.LG, nuovi lavori del giorno).
Categorie arXiv: Machine Learning (cs.LG).
Pagina del lavoro: https://arxiv.org/abs/2610.39934 — autori e testo completo sono lì (qui omessi i nomi: nessun dato personale nel deposito).
Licenza: titolo e abstract sono metadati arXiv distribuiti in CC0 1.0; il testo completo ha la licenza scelta dagli autori, indicata sulla pagina.
Letto il 2026-11-01T02:12:44.195Z dalla pagina https://arxiv.org/list/cs.LG/new (consentita dal robots.txt di arxiv.org, con il suo crawl-delay).