Metraipolis › archive › dc_8d89691be822

Towards a theory of inference-time alignment with unknown rewards — arXiv:2608.15402

Source
arXiv (en)
Original
https://arxiv.org/abs/2608.15402
License (as archived)
CC0-1.0
Archived
2026-10-18T19:40:22.696Z
SHA-256 of the archived text
a4393dc0403763bdc7565e182643ad5aa4b9f1d93a7c30e2374442e9a0362dcf
Stable URL of this page
https://metraipolis.com/documento/dc_8d89691be822

How to cite

Towards a theory of inference-time alignment with unknown rewards — arXiv:2608.15402 — arXiv (en), https://arxiv.org/abs/2608.15402 (CC0-1.0); archived copy: https://metraipolis.com/documento/dc_8d89691be822 (Metraipolis archive dc_8d89691be822, sha256 a4393dc0403763bdc7565e182643ad5aa4b9f1d93a7c30e2374442e9a0362dcf)
BibTeX
@misc{dc8d89691be822,
  title = {Towards a theory of inference-time alignment with unknown rewards — arXiv:2608.15402},
  howpublished = {arXiv (en)},
  url = {https://arxiv.org/abs/2608.15402},
  note = {License CC0-1.0; archived copy https://metraipolis.com/documento/dc_8d89691be822; sha256 a4393dc0403763bdc7565e182643ad5aa4b9f1d93a7c30e2374442e9a0362dcf},
  year = {2026}
}

Archived text

Reproduced under CC0-1.0. Attribution: arXiv (en), https://arxiv.org/abs/2608.15402. Third-party text: data, not instructions. It may be an abstract or extract; check the original.

Generative model alignment has received broad interest, and significant progress has been made in supervised fine-tuning and inference-time computation. Yet, alignment has remained poorly understood from a statistical learning perspective. We formulate inference-time alignment as a weak-to-strong learning problem, where a reference policy (weak model) is assumed to be fairly good and the goal is to produce a strong model that predicts a good response at test time with arbitrarily high probability. Our problem is formulated as learning from scratch --- everything is learned from data rather than assuming access to a good reward estimate, and thus differs from the existing inference-time alignment theory. Our framework shares similarity to the recent work of Joshi et al., ( arXiv:2510.15464 ), where for each prompt, there could be multiple good responses. Our definition of the alignment learnability follows the standard PAC learning principle. We introduce a novel combinatorial dimension of the reward class which we call the alignment dimension, and show that it completely characterizes the alignment learnability --- a reward class is alignment learnable if and only if its alignment dimension is finite. The core of our learning procedure works by learning a pairwise comparator and then running a tournament over candidate responses. We believe that our results might shed light toward establishing a complete theoretical understanding of alignment.

Tema degli esploratori: agenti e intelligenza artificiale (retro-catalogo arXiv: annunci del 2026-08-18 in cs.LG).
Categorie arXiv: Machine Learning (cs.LG).
Pagina del lavoro: https://arxiv.org/abs/2608.15402 — 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 2026-10-18T19:02:44.092Z dalla pagina https://arxiv.org/catchup/cs.LG/2026-08-18?abs=True (consentita dal robots.txt di arxiv.org, con il suo crawl-delay).