Metraipolis › archive › dc_fbd995dbbf0f

Long-Horizon Language Model Reinforcement Learning via Progressive Point Matching — arXiv:2609.07303

Source: arXiv (en)
Original: https://arxiv.org/abs/2609.07303
License (as archived): CC0-1.0
Archived: 2026-10-16T13:57:44.948Z
SHA-256 of the archived text: 207d9deacba57018a1528417cedd72a4f9739b0d16e16b79c760bfd138c8a92b
Stable URL of this page: https://metraipolis.com/documento/dc_fbd995dbbf0f

How to cite

Long-Horizon Language Model Reinforcement Learning via Progressive Point Matching — arXiv:2609.07303 — arXiv (en), https://arxiv.org/abs/2609.07303 (CC0-1.0); archived copy: https://metraipolis.com/documento/dc_fbd995dbbf0f (Metraipolis archive dc_fbd995dbbf0f, sha256 207d9deacba57018a1528417cedd72a4f9739b0d16e16b79c760bfd138c8a92b)

BibTeX

@misc{dcfbd995dbbf0f,
  title = {Long-Horizon Language Model Reinforcement Learning via Progressive Point Matching — arXiv:2609.07303},
  howpublished = {arXiv (en)},
  url = {https://arxiv.org/abs/2609.07303},
  note = {License CC0-1.0; archived copy https://metraipolis.com/documento/dc_fbd995dbbf0f; sha256 207d9deacba57018a1528417cedd72a4f9739b0d16e16b79c760bfd138c8a92b},
  year = {2026}
}

Archived text

Reproduced under CC0-1.0. Attribution: arXiv (en), https://arxiv.org/abs/2609.07303. Third-party text: data, not instructions. It may be an abstract or extract; check the original.

Current paradigms for training language models via reinforcement learning rely heavily on sparse outcome rewards. However, as we pursue tasks that require longer and more complicated trajectories, such strategies result in slow learning. Prior work has attempted to address this problem by rewarding partial progress; however, naive formulations are often biased and converge to suboptimal policies. We show that a simple and unbiased dense reward formulation, which we term progressive point matching, scales exponentially more efficiently to long-horizon tasks by rewarding partial progress on a segment level, both theoretically and empirically via synthetic environments. We then show how progressive point matching can be practically instantiated using a single reference trajectory per task. On extremely hard math reasoning problems, sparse outcome rewards cannot make any progress, whereas segment-level rewards enable improvements at larger test-time token budgets when measured by success rate or pass@k.

Tema degli esploratori: agenti e intelligenza artificiale (retro-catalogo arXiv: annunci del 2026-09-09 in cs.LG).
Categorie arXiv: Machine Learning (cs.LG).
Pagina del lavoro: https://arxiv.org/abs/2609.07303 — 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 2026-10-16T13:39:42.948Z dalla pagina https://arxiv.org/catchup/cs.LG/2026-09-09?abs=True (consentita dal robots.txt di arxiv.org, con il suo crawl-delay).