DIO：精炼互信息与因果链以增强机器抽象推理能力
DIO: Refining Mutual Information and Causal Chain to Enhance Machine Abstract Reasoning Ability
AI总结：
本文针对瑞文渐进矩阵抽象推理任务，提出通过收紧互信息下界和引入元数据监督来精炼因果链建模，显著提升判别准确率并首次实现开放式答案生成。
摘要：
尽管深度学习取得了广泛成功，其抽象推理瓶颈依然存在。我们针对瑞文渐进矩阵（RPM）这一用于模式、推理和问题解决智能的基准展开研究。我们建模了完整的因果链：图像 $\ ightarrow$ 属性 $\ ightarrow$ 渐进模式 $\ ightarrow$ 一致性 $\ ightarrow$ 答案，并构建了基线模型 DIO。然而，DIO 的互信息下界目标并未嵌入人类逻辑：该下界较为松弛且基于统计，忽略了因果性的主客体关联。因此，我们提出三项改进。1) Brando 引入可训练的负选项以收紧变分下界。2) WORLD 将生成替换为高斯混合特征模型，提供无限的加权负样本，进一步收紧下界。3) DIEGO 增加元数据监督，以纠正“属性 $\ ightarrow$ 模式”的语义鸿沟，使表征与人类规则对齐。这些升级显著提升了判别式 RPM 的准确率，并首次使 DIO 能够在开放式 RPM 中生成有效答案。该工作为抽象推理研究提供了因果驱动的设计准则、目标精炼策略以及跨模态洞见。
英文摘要：
Despite deep learning's broad success, its abstract-reasoning bottleneck persists. We tackle Raven's Progressive Matrices (RPM), the benchmark for pattern, reasoning and problem-solving intelligence. We model the full causal chain image $\rightarrow$ attributes $\rightarrow$ progressive patterns $\rightarrow$ consistency $\rightarrow$ answer and build the baseline DIO. Yet DIO's mutual-information lower-bound objective does not embed human logic: the bound is loose and statistic-based, ignoring causal subject-object links. We therefore present three refinements. 1) Brando introduces trainable negative options to tighten the variational bound. 2) WORLD replaces generation with a Gaussian-mixture feature model that supplies infinite, weighted negatives, further tightening the bound. 3) DIEGO adds metadata supervision to rectify the "attributes $\rightarrow$ patterns" semantic gap, aligning representations with human rules. These upgrades substantially boost discriminative RPM accuracy and, for the first time, let DIO generate valid answers in open-ended RPM. The work provides causal-driven design guidelines, objective-refinement strategies and cross-modal insights for abstract-reasoning research.
机构、总结、中文摘要均由AI分析而来，可能存在错误，请以论文正文为准。
