arXiv：cs.AI· Sida Li, Yuanxin Liu, Linli Yao, Shicheng Li, Xu Sun, Haoli Bai, Xiaohui Li·· 13 小时前
AdaC-GRPO：用自适应课程学习攻克长程视觉数独
AdaC-GRPO: Towards Mastering Long-Horizon Visual Sudoku Puzzles via Adaptive Curriculum Learning
AI 导读
针对长程多模态任务中 RLVR 的奖励稀疏问题，研究者提出 AdaC-GRPO，通过在线 rollout 准确率追踪模型"最近发展区"来动态构建训练批次，并发布按网格尺寸和推理步数分级的 FG-Sudoku 基准。
正文
Abstract:While reinforcement learning with verifiable reward (RLVR) enhances reasoning in large language models, its efficacy in long-horizon multimodal tasks remains underexplored. Visual Sudoku puzzles pose a representative challenge, requiring multi-step deductive reasoning over both local constraints and global dependency chains. To this end, we introduce FG-Sudoku, a fine-grained benchmark categorized by grid size and required deductive steps. Evaluations reveal that frontier MLLMs fail on complex long-range dependencies (e.g., 9*9 puzzles). During post-training, this near-zero success rate causes severe reward sparsity, halting standard RLVR. We propose Adaptive Curriculum Group Relative Policy Optimization (AdaC-GRPO) to overcome this. Unlike rigid or prior-based schedules, AdaC-GRPO dynamically constructs training batches by tracking the model's "Zone of Proximal Development" via online rollout accuracy. Experiments demonstrate that AdaC-GRPO enables Qwen2.5-VL-7B to approach the performance of Gemini-2.5-Flash while bridging 62.2% of the overall accuracy gap to Gemini-2.5-Pro, with its generalizability further validated across InternVL-3.5 and Qwen3-VL.
| Subjects: | Artificial Intelligence (cs.AI) |
| Cite as: | arXiv:2610.13184 [cs.AI] |
| (or arXiv:2610.13184v1 [cs.AI] for this version) | |
| https://doi.org/10.48550/arXiv.2610.13184 arXiv-issued DOI via DataCite (pending registration) |
Submission history
From: Sida Li [view email]
[v1]
Fri, 9 Oct 2026 08:08:42 UTC (1,297 KB)
来源：arXiv：cs.AI · arxiv.org