结构化稀疏记忆用于循环推理
Structured Sparse Memory for Recurrent Reasoning
浏览论文内容
中文总结 AI 辅助
针对ARC推理任务，提出CHARM模型，结合循环推理、结构化任务记忆（CoSE）和合成数据，大幅减少记忆参数并提升性能，在ARC-AGI-1和ARC-AGI-2上分别达到84%和46.7%的pass@2。
中文摘要 AI 辅助
从零训练的循环模型最近在ARC风格的推理任务上变得具有竞争力，但围绕小型循环骨干网络的常规框架忽视了系统的两个重要部分：任务条件记忆和合成增强数据。我们通过CHARM研究这一机制，CHARM是一个紧凑的混合ARC模型，结合了循环推理、结构化任务记忆、合成数据和推理时聚合。在现有方法中，任务条件记忆提供了大量隐藏的容量来源，其大小超过循环骨干网络的30倍。我们引入了一种用于任务条件化的组合稀疏嵌入（CoSE），在受控ARC消融实验中，将学习的任务记忆参数减少了超过90%，同时提高了pass@2。对于循环骨干网络，循环深度仅在平衡学习视野时才有帮助。结合这些要素，我们的系统在ARC-AGI-1上达到84%的pass@2，在ARC-AGI-2公开评估上达到46.7%的pass@2。结构化记忆的优势也泛化到未见过的谜题和其他领域。我们的代码、数据集和模型检查点可在以下网址获取：此https URL。
英文摘要
Recurrent models trained from scratch have recently become competitive on ARC-style reasoning tasks, but the usual framing around small recurrent backbones overlooks two important parts of the system: task-conditioned memory and synthetic augmentation data. We study this regime through CHARM, a compact hybrid ARC model that combines recurrent reasoning with structured task memory, synthetic data, and inference-time aggregation. In existing approaches, task-conditioned memory supplies a large hidden source of capacity, reaching more than 30x the size of the recurrent backbone. We introduce a compositional sparse embedding (CoSE) for task conditioning that reduces learned task-memory parameters by over 90% while improving pass@2 in controlled ARC ablations. For the recurrent backbone, recurrent depth helps only when balanced with learning horizon. Combining these ingredients, our system reaches 84% pass@2 on ARC-AGI-1 and 46.7% pass@2 on ARC-AGI-2 public evaluation. The benefits of structured memory also generalize to unseen puzzles and other domains. Our code, dataset, and model checkpoints are available at https://github.com/water-vapor/charm.
