LeWi：让神经元学习低维嵌入和局部连接，提升样本效率
Learning Computational Structure for Sample-Efficient Prediction
一篇很新颖的架构论文：给每个神经元学一个位置，只连附近的邻居，17 个世界模型任务里赢了 16 个，做模型结构的人可以看看。
论文针对高维学习中样本复杂度受环境维度制约的问题，提出 Learned Wiring（LeWi）方法：每个神经元拥有可学习的低维嵌入，只与邻近神经元连接，嵌入与权重联合训练。作者构建了 CSFRBench 基准，训练时隐藏目标的组合计算结构，发现稠密网络和稀疏基线未能利用这一结构。在 32 个由独立采样计算图生成的目标上，LeWi 的平均学习曲线面积比稠密网络低 27%，是不提供目标图情况下最省样本的方法。在 17 个世界模型预测任务（覆盖控制、导航、操作）中，LeWi 在 16 个任务上超过稠密基线，并在 119 次基线对比中赢下 99 次。
Learning Computational Structure for Sample-Efficient Prediction
Statistical learning in high dimensions suffers from the curse of dimensionality. Fortunately, many functions of practical interest are compositionally sparse, meaning they admit a composition of local functions, each depending on only a few variables, allowing sample complexity to depend on local arity rather than ambient dimension. Yet realizing this benefit requires exploiting hidden compositional structure, which theory does not establish for practical gradient-based training. We study this gap with CSFRBench, a controlled benchmark that hides the target's compositional computation during training but retains it for diagnosis, and find that dense networks and sparsity-based baselines leave much of that benefit unrealized. Inspired by the spatial organization of biological neural circuits, we introduce Learned Wiring (LeWi): each neuron has a learned low-dimensional embedding, only nearby neurons connect, and embeddings and weights are trained jointly. LeWi improves sample efficiency on targets with hierarchical structure, intermediate reuse, or independent modules. Across 32 additional targets generated by independently sampled computation graphs, LeWi is the most sample-efficient method not given the target graph, with 27% lower mean learning-curve area than dense networks. Dense networks perform better on targets with a single high-arity function. LeWi's representations and connectivity align with the target computation. On 17 world-model prediction tasks spanning control, navigation, and manipulation, LeWi beats the dense baseline on 16 and wins 99/119 baseline comparisons. These results show that jointly learning connectivity and weights can exploit computational structure to improve sample efficiency.