提出可微分的意识单元，让机器在故障中快速自适应。
The Reflexive Integrated Information Unit: A Differentiable Primitive for Artificial Consciousness
- 用元状态和广播缓存增强递归单元，模拟自我感知。
- 故障后13步内恢复90%奖励，速度是普通网络两倍。
- 将意识研究转为可计算、可训练的数学问题，适合脑科学与AI融合研究者。
人工意识研究缺乏类似感知机的可复制、可训练基础模块。本文提出反射式整合信息单元（RIIU），一种递归神经单元，在隐藏状态 $h$ 外增加两个向量：(i) 元状态 $μ$ 记录单元自身的因果痕迹；(ii) 广播缓冲区 $B$ 将该痕迹暴露给网络其余部分。通过滑动窗口协方差与可微分的 Auto-$Φ$ 代理，每个 RIIU 能在线最大化局部信息整合。我们证明：(1) RIIUs 可端到端反向传播；(2) 模块可加性组合；(3) 在梯度上升下实现 $Φ$-单调可塑性。在八路网格世界任务中，四层 RIIU 代理在执行器故障后仅13步即恢复超过90%奖励，速度是参数匹配的GRU的两倍，且持续保持非零 Auto-$Φ$ 信号。通过将“类意识”计算压缩至单元尺度，RIIUs 将哲学争议转化为可实证的数学问题。
原文摘要 · Abstract (English)
Research on artificial consciousness lacks the equivalent of the perceptron: a small, trainable module that can be copied, benchmarked, and iteratively improved. We introduce the Reflexive Integrated Information Unit (RIIU), a recurrent cell that augments its hidden state $h$ with two additional vectors: (i) a meta-state $μ$ that records the cell's own causal footprint, and (ii) a broadcast buffer $B$ that exposes that footprint to the rest of the network. A sliding-window covariance and a differentiable Auto-$Φ$ surrogate let each RIIU maximize local information integration online. We prove that RIIUs (1) are end-to-end differentiable, (2) compose additively, and (3) perform $Φ$-monotone plasticity under gradient ascent. In an eight-way Grid-world, a four-layer RIIU agent restores $>90\%$ reward within 13 steps after actuator failure, twice as fast as a parameter-matched GRU, while maintaining a non-zero Auto-$Φ$ signal. By shrinking "consciousness-like" computation down to unit scale, RIIUs turn a philosophical debate into an empirical mathematical problem.
Thank you to arXiv for use of its open access interoperability. PaperDance 不是 arXiv 官方产品；中文卡片由大模型生成，请以原文为准。