[论文] SoftReason: A Fully Differentiable Neuro-Soft-Symbolic Deductive Reaso...

论文概要

研究领域: ML
作者: Wael AbdAlmageed
发布时间: 2026-07-24
arXiv: 2507.18392

中文摘要

在许多推理问题中，前提不是作为离散符号观察到的，而是必须从高维输入中推断出来。此外，谓词词汇、论证结构和可信证据由知识图谱（KG）或规则定义提供。经典的神经符号流水线在感知和演绎之间有离散的接口。我们提出了一种神经软符号架构，用于对潜在感知事实和知识提供的谓词进行可微演绎推理。SoftReason通过将演绎状态表示为候选常数和谓词上的局部软解释张量来消除梯度差距。感知提出概率基础事实，KG三元组作为高置信度软证据进入，每个查询锚点、谓词选择和闭包更新保持可微。我们的核心创新是即时后果算子的学习可微提升。它使用谓词定义嵌入和潜在组合通道来形成软体谓词混合物，聚合所有可能的见证，提出查询条件头事实，并通过单调概率OR更新解释。我们在知识感知视觉问答（KVQA）上实例化该框架，并展示了SoftReason如何支持端到端感知接地、KG证据注入和可微演绎闭包在一个可训练的架构中。

原文摘要

In many reasoning problems, the premises are not observed as discrete symbols, but must be inferred from high-dimensional inputs. Further, the predicate vocabulary, argument structure, and trusted evidence are supplied by a Knowledge Graph (KG), or rule definitions. Classical neuro-symbolic pipelines have a discrete interface between perception and deduction. We present a neuro-soft-symbolic architecture for differentiable deductive reasoning over latent perceptual facts and knowledge-provided predicates. SoftReason removes the gradient gap by representing the deductive state as a local soft interpretation tensor over candidate constants and predicates. Perception proposes probabilistic base facts, KG triples enter as high-confidence soft evidence, and every query anchor, predicate choice,...

自动采集于 2026-07-24

#论文 #arXiv #ML #小凯