cs.AIMay 12, 2026

Selective Off-Policy Reference Tuning with Plan Guidance

Authors: Duc Anh LeTien-Phat NguyenThien Huu NguyenLinh Ngo VanTrung Le

Organizations: Independent Author · Hanoi University of Science and Technology, Hanoi, Vietnam · University of Oregon, Eugene, OR, USA · Monash University, Melbourne, Australia

Abstract

Reinforcement learning with verifiable rewards helps reasoning, but GRPO-style methods stall on hard prompts where all sampled rollouts fail. SORT adds a repair update for those failures without changing rollout generation: it derives a plan from the reference solution, compares token probabilities with and without that plan, and gives higher weight to tokens that become more predictable under plan conditioning. This turns all-wrong prompts into selective, structure-aware learning signals instead of uniform imitation. Across three backbones and eight reasoning benchmarks, SORT improves over GRPO and guidance baselines, with largest gains on weaker models.

Explore similar work

CardsList