AdaC-GRPO: Towards Mastering Long-Horizon Visual Sudoku Puzzles via Adaptive Curriculum Learning
While reinforcement learning with verifiable reward (RLVR) enhances reasoning in large language models, its efficacy in long-horizon multimodal tasks remains underexplored.
ProofPaper ↗
Key points
- 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.
- We propose Adaptive Curriculum Group Relative Policy Optimization (AdaC-GRPO) to overcome this.
- 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.
Sources (1)
- [1]AdaC-GRPO: Towards Mastering Long-Horizon Visual Sudoku Puzzles via Adaptive Curriculum LearningarXiv (AI, ML, NLP, CV, robotics, multi-agent) · Oct 9, 08:08 AM
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.
Extractive summary: sentences quoted from the sources.
Before this
- Oct 9, 2026Introducing Clef-omni with full multimodality, plus a faster Clef and a cheaper Clef-flash
- Oct 9, 2026The Thinking Leakage in Reinforcement Learning for LLMs
- Oct 8, 2026Residual Advantage: Student-Relative Teacher Guidance for RL with Verifiable Rewards
- Oct 8, 2026AdaptEvo: Adaptive Agent Learning with Evolving Supervision
- Oct 7, 2026SPLIT-RL: Staged Perception-Language Reasoning Training with Claim-Level Advantages
- Oct 7, 2026On the Clock: Towards Punctual and Productive Time-Budgeted AI Agents