arXiv 论文提出可微精确学习框架 DFST,实现极长序列泛化
Differentiable Exact Learning
这篇来自 arXiv 的论文做了件挺有意思的事:用有限训练数据教会神经网络精确执行图灵机模拟和二进制算术,还能从模型里提取出可读的离散控制器并给出证明。做状态追踪、形式化方法或可解释性研究的开发者值得把开源代码跑一遍。
论文《Differentiable Exact Learning》提出将无界外部环境与状态追踪神经控制器耦合的训练框架,训练数据来自策略轨迹观测(PTO),即只暴露观测与动作、隐藏内部状态的专家行为轨迹。论文同时提出 Differentiable Finite-State Transducers(DFST),采用多重线性更新并支持 log-parallel 训练,可模拟任意有限状态控制器。
在小规模训练数据下,紧凑的 DFST 在算术与通用模拟任务上实现极长序列泛化。作者将该配方迁移到 RNN 与 GRU 上同样复现成功,总结为:状态追踪控制器 + 无界外部环境 + PTO 等于极长泛化。度量提取协议还能从训练后的 DFST 中恢复出可解释的离散控制器,并证明其与目标算术、通用模拟算法等价;Softmax DFST 只需对转移概率取 argmax,即可精确执行通用图灵机模拟与二进制算术。
论文承认精确学习的证明依赖若干假设,但给出了经验支持。代码、训练配方、模型权重与证明检查器已在 GitHub(fanagor/DEL)开源。
Differentiable Exact Learning
We introduce a framework for differentiable exact learning that couples an unbounded external environment to a state-tracking neural controller, which we train on policy-trajectory observations (PTOs): expert behavior traces that expose observations and actions while hiding internal states. In addition to the learning framework, we introduce Differentiable Finite-State Transducers (DFSTs) which use multilinear updates, support log-parallel training, and emulate arbitrary finite-state controllers, yielding computational universality through environmental interaction. From tiny training datasets, compact DFSTs achieve extreme length generalization on arithmetic and universal-simulation tasks. We replicate the success of DFSTs on RNNs and GRUs, thereby supporting a broader learning recipe: State-Tracking Controller + Unbounded External Environment + PTOs = Extreme Length Generalization. Beyond length generalization, our metric-extraction protocol recovers interpretable discrete controllers from trained DFSTs, which we prove equivalent to the target arithmetic and universal-simulation algorithms. Furthermore, for our Softmax DFST models, simply taking $\mathrm{argmax}$ over the transition probabilities yields discrete controllers which provably exactly perform universal Turing-machine simulation and binary arithmetic. Proving exact learning for the neural controllers requires assumptions, which we empirically support. We make our code, training recipes, model weights, and proof checkers publicly available at https://github.com/fanagor/DEL.
本条情报归属 AITOP「论文」核心栏目,重点聚焦北美前沿实验室与硅谷科技巨头。在跨区域产业观察坐标中,该动态直接关联「AI 开发者工具链、Coding Agent 与工程化实践」研究主线,反映出当前技术演进与商业落地的最新风向。涉及核心实体与技术议题包括:DFST、RNN、GRU、图灵机,读者可通过下方标签进入对应专题页查阅历史脉络与关联对比。
信源核验说明:本文原始线索采编自「arXiv cs.AI」(@Hristo Papazov, Francesco D'Angelo, Nicolas Flammarion),经 AITOP 多维评分管线结构化提炼。