LLM 在扩展搜索空间上的优化式推理研究:OPT* 任务族
Optimization-like Reasoning in LLMs over Expanding Search Spaces
AI 导读
研究者提出 OPT*,一个用于训练与评估大语言模型逐步优化式推理的可扩展任务族:每个任务提供可行性检查器与评估器,通过复杂度参数扩展搜索空间而无需新的人工标注。
正文 · 原文
Abstract:Verifiable reward training has improved mathematical and coding reasoning, but these domains capture only part of step-by-step decision making. Many real-world tasks require finding a high-value feasible plan among many valid alternatives. We introduce OPT*, a scalable family of optimization-style tasks for training and evaluating LLM step-by-step optimization-like reasoning along a complexity axis: each task provides a feasibility checker and evaluator, while a complexity parameter expands the search space without requiring new human labels. This motivates studying these tasks in two regimes: (i) solver-guided online policy optimization, which uses a solver as a value oracle for partial states and applies rank-based reward shaping to reinforce better next steps, and (ii) search-based offline RL when such solvers are unavailable. Theoretically, we relate success in large search spaces to the information a reasoner extracts per unit of search budget. Empirically, we ablate the ingredients that make search efficient on OPT* and show that training on OPT* improves step-by-step optimization-like reasoning.
| Subjects: | Artificial Intelligence (cs.AI) |
| Cite as: | arXiv:2606.05464 [cs.AI] |
| (or arXiv:2606.05464v2 [cs.AI] for this version) | |
| https://doi.org/10.48550/arXiv.2606.05464 arXiv-issued DOI via DataCite |
Submission history
From: Nicolas Astorga [view email]
[v1]
Wed, 3 Jun 2026 21:43:38 UTC (1,089 KB)
[v2]
Tue, 17 Nov 2026 00:52:17 UTC (1,862 KB)
来源:arXiv cs.AI · arxiv.org