发表机构
University of Almería; Research Centre CIAIMBITAL（阿尔梅里亚大学; CIAIMBITAL研究中心）
机构由 AI 辅助整理，请以论文原文为准。AI 中文总结
本研究构建含自适应停机等模块的极简完整认知架构，探究功能是涌现还是需计算，发现价值需计算实现，还验证了停机机制、自洽性等相关结论。
AI 中文摘要
认知架构不止是进行推理的模块，它还必须决定思考时长以及哪些内容值得投入精力。我们构建了一个极简但完整的系统——具备自适应停机机制的循环推理器、稳态控制场以及价值模块，并针对每个部分提出问题：该功能是从梯度下降中涌现的，还是必须通过计算实现？能力是涌现的。停机似乎也是涌现的，且比所有可预先判定的内容更具价值，但这种表象是工具性的：匹配平均计算量下的收益从0.467（均匀分布）升至0.546（难度相关），再到0.698（事前价值），而进一步升至0.921（事后自我观测）的结果无法通过验证。PonderNet式停机机制返回隐藏状态的停机加权混合，而强制深度基线仅返回单个隐藏状态，语言头仅在该混合上训练；均衡读出会消除原生执行的表观优势（残差+0.000 [0.000, 0.000]）。价值并非涌现的：训练后的耦合完全未捕捉到显式分配器能完全捕捉的收益（+0.151，路由相关性+0.79），因此收益所需的二阶决策必须通过计算实现，至少在价值与内容正交的情况下是如此，本研究正是通过构造实现了这一点。在冻结的大语言模型（LLM）执行器上，相同工具显示自洽投票为可测边界（+0.0236 [+0.0150, +0.0326]），而样本间一致性作为停机信号几乎毫无价值，其质量集中于错误答案。我们断言的每个零结果都对应一种机制和一个正对照，该方案也是本研究的贡献之一。执行我们可证伪的预测，承诺下的价值在悬崖成本族中收益为+0.1312 [+0.1124, +0.1502]，约为平滑族估计值的7倍——这并非因为悬崖事前转移了信息，而是因为它将可实现范围扩大了5.1倍（5.1x [3.4, 8.2]）。
英文摘要
A cognitive architecture is more than the module that reasons: it must also decide how long to think and what deserves the effort. We built a minimal but complete system - a recurrent reasoner with adaptive halting, a homeostatic control field, and a value module - and asked of each part: does this function emerge from gradient descent, or must it be computed? Competence emerges. Stopping appears to emerge too, and to be worth more than everything decidable in advance, but that appearance is instrumentation: payoff at matched mean compute climbs from 0.467 (uniform) through 0.546 (difficulty) to 0.698 (ex-ante value), and the further climb to 0.921 (posterior self-observation) does not survive audit. PonderNet-style halting returns a halting-weighted mixture of hidden states while forced-depth baselines return one, and the language head is trained on the mixture alone; equalizing the readout annihilates the apparent advantage of native execution (residual +0.000 [0.000, 0.000]). Value does not emerge: trained couplings capture zero of a payoff an explicit allocator captures completely (+0.151, routing correlation +0.79), so the second-order decisions that pay must be computed, at least where value is orthogonal to content, as here by construction. On a frozen LLM actuator the same instruments show self-consistency voting to be a measured bound (+0.0236 [+0.0150, +0.0326]) and inter-sample agreement nearly worthless as a stopping signal, its mass concentrating on wrong answers. Every null we assert carries a mechanism and a positive control, and the protocol is part of the contribution. Executing our own falsifiable prediction, value under commitment pays +0.1312 [+0.1124, +0.1502] in a cliff-cost family, some seven times the smooth-family estimate - not because the cliff shifts information ex ante, but because it multiplies the attainable range fivefold (5.1x [3.4, 8.2]).
Comments16 pages, 3 figures. Code, preregistrations and results: https://github.com/fmarrabal/miuracognitive
