Guardado en:
Detalles Bibliográficos
Autor principal: Chitra, Tarun
Formato: Preprint
Publicado: 2025
Materias:
Acceso en línea:https://arxiv.org/abs/2504.09777
Etiquetas: Agregar Etiqueta
Sin Etiquetas, Sea el primero en etiquetar este registro!
_version_ 1866915240763981824
author Chitra, Tarun
author_facet Chitra, Tarun
contents Chain-of-thought reasoning enables large language models to solve multi-step tasks by framing problem solving as sequential decision problems. Outcome-based rewards, which provide feedback only on final answers, show impressive success, but face challenges with credit assignment and slow convergence. In contrast, procedure-based rewards offer efficient step-level feedback, but typically require costly human supervision. We introduce \emph{Backwards Adaptive Reward Shaping} (BARS), a no-regret framework that converts sparse outcomes-based rewards into effective procedure-based signals. BARS uses sparse rewards generated from terminal-state priors and cover trees to scale rewards while preventing exploitation. With Bellman contraction and $(Δ, ε)$-gap rewards, our backward Euler solver achieves $ε$-accuracy in $O\left((R_{\max}/Δ)\log(1/ε)\right)$ iterations with $O(\log T)$ dynamic regret over $T$ rounds. Our analysis, based on generic chaining, continuous scaling limits, and non-linear Feynman-Kac bounds, connects recent outcome-based methods' empirical successes with the benefits of intermediate supervision. Combined, this provides the first rigorous no-regret algorithm for outcome reward shaping, providing a theoretical foundation for the empirical success of DeepSeek's R1.
format Preprint
id arxiv_https___arxiv_org_abs_2504_09777
institution arXiv
publishDate 2025
record_format arxiv
spellingShingle Reasoning without Regret
Chitra, Tarun
Machine Learning
Artificial Intelligence
Chain-of-thought reasoning enables large language models to solve multi-step tasks by framing problem solving as sequential decision problems. Outcome-based rewards, which provide feedback only on final answers, show impressive success, but face challenges with credit assignment and slow convergence. In contrast, procedure-based rewards offer efficient step-level feedback, but typically require costly human supervision. We introduce \emph{Backwards Adaptive Reward Shaping} (BARS), a no-regret framework that converts sparse outcomes-based rewards into effective procedure-based signals. BARS uses sparse rewards generated from terminal-state priors and cover trees to scale rewards while preventing exploitation. With Bellman contraction and $(Δ, ε)$-gap rewards, our backward Euler solver achieves $ε$-accuracy in $O\left((R_{\max}/Δ)\log(1/ε)\right)$ iterations with $O(\log T)$ dynamic regret over $T$ rounds. Our analysis, based on generic chaining, continuous scaling limits, and non-linear Feynman-Kac bounds, connects recent outcome-based methods' empirical successes with the benefits of intermediate supervision. Combined, this provides the first rigorous no-regret algorithm for outcome reward shaping, providing a theoretical foundation for the empirical success of DeepSeek's R1.
title Reasoning without Regret
topic Machine Learning
Artificial Intelligence
url https://arxiv.org/abs/2504.09777