analiza aiqbench
This research introduces APIVIS, a framework that enhances mathematical reasoning in large language models by integrating value-guided search into reinforcement learning with verifiable rewards. By combining direct and searched responses to generate informative relative rewards, the method ensures that search-induced improvements effectively update the policy rather than merely increasing trajectory diversity. The approach demonstrates substantial performance gains on standard benchmarks, offering a robust solution for optimizing complex reasoning tasks within AI systems.
Improving Math Reasoning through Value-guided Informative Search
arXiv:2610.01080v1 Announce Type: new Abstract: Reinforcement learning with verifiable rewards (RLVR) has substantially improved the mathematical reasoning capabilities of large language models. Recent work introduces search into RLVR rollouts to increase trajectory diversity, but diversity alone does not ensure that the search-induced rollout policy improves upon the current policy. To address this gap, we propose APIVIS, a training-time framework that adapts finite-budget Gumbel search to chunk-level mathematical reasoning. APIVIS combines direct and searched responses within each rollout group, allowing improvements found by search to produce informative relative rewards. It further applies selective supervision to search-improved tokens, preserving a learning signal when uniform group rewards render GRPO ineffective. We show that exact value-guided selection improves the expected verifier reward at each searched state and that this guarantee extends to the complete rollout policy, with a corresponding approximate guarantee under bounded value-estimation error. Experiments on widely recognized mathematical reasoning benchmarks and different model scales demonstrate substantial improvements over competitive search-based methods, validating the effectiveness of APIVIS.
Abstract:Reinforcement learning with verifiable rewards (RLVR) has substantially improved the mathematical reasoning capabilities of large language models. Recent work introduces search into RLVR rollouts to increase trajectory diversity, but diversity alone does not ensure that the search-induced rollout policy improves upon the current policy. To address this gap, we propose APIVIS, a training-time framework that adapts finite-budget Gumbel search to chunk-level mathematical reasoning. APIVIS combines direct and searched responses within each rollout group, allowing improvements found by search to produce informative relative rewards. It further applies selective supervision to search-improved tokens, preserving a learning signal when uniform group rewards render GRPO ineffective. We show that exact value-guided selection improves the expected verifier reward at each searched state and that this guarantee extends to the complete rollout policy, with a corresponding approximate guarantee under bounded value-estimation error. Experiments on widely recognized mathematical reasoning benchmarks and different model scales demonstrate substantial improvements over competitive search-based methods, validating the effectiveness of APIVIS.
Artificial Intelligence (cs.AI)
arXiv:2610.01080 [cs.AI]
(or arXiv:2610.01080v1 [cs.AI] for this version)
https://doi.org/10.48550/arXiv.2610.01080 Focus to learn more arXiv-issued DOI via DataCite (pending registration)
arXivLabs is a framework that allows collaborators to develop and share new arXiv features directly on our website.
Both individuals and organizations that work with arXivLabs have embraced and accepted our values of openness, community, excellence, and user data privacy. arXiv is committed to these values and only works with partners that adhere to them.
Have an idea for a project that will add value for arXiv's community? Learn more about arXivLabs.
Text integral sub licența deschisă a publicației · 253 cuvinte
Subiecte conexe
Nicole’s guide to writing and editing
Quantum Frontiers · 26 aug. 2026
Synthesis Superintelligence: from Semiconductors to Superconductors — Periodic Labs’ Liam Fedus and Ekin Dogus Cubuk
Latent Space · 8 oct. 2026
Links #5: 2026/07
LessWrong · 31 iul. 2026
ReST-RL: Reinforcing LLM Reasoning through Unified Self-Training and Value-Guided Search
arXiv · 9 sept. 2026