LATIDIA · Investigación

AdaC-GRPO: Hacia el dominio de los rompecabezas de sudoku visual de largo horizonte a través del aprendizaje curricular adaptativo

arXiv: 2610.13184v1Tipo de anuncio: nuevo Resumen: Si bien el aprendizaje por refuerzo con recompensa verificable (RLVR) mejora el razonamiento en modelos de lenguaje grandes, su eficacia en tareas multimodales de largo horizonte sigue siendo poco explorada.

La noticia

arXiv:2610.13184v1 Announce Type: new 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