AI 中文总结
本研究通过六十个四单元系统的预注册实验，证实了证据遮蔽机制能显著提升组合泛化准确率，但角色信息的作用及机制归因仍待进一步探索。
AI 中文摘要
限制一个模块能读取的内容，可能会改善系统学习计算的结果。我们在一个预注册的确认研究中对此进行了测试，该研究涉及六十个四单元系统，它们共享一个冻结的语言模型骨干，并通过学习到的连续数据包进行通信。五个条件在六个初始化簇中变化了证据遮蔽、所有权标记以及用中性填充物替换外部证据，每个簇有两个数据顺序，在一个全新的任务世界中。在两种机制下都有标记可用时，遮蔽在保留的两操作和三操作组合上的准确率提高，中位数配对差异分别为0.846和0.859；所有十二对均超过了要求的边际，完整的预注册行为标准通过。未标记的复制也通过了。没有全局可见的系统通过标记跟随检查，因此可用角色信息的效果仍未解决。填充条件产生了七个完整的泛化器，但其分解标准不具结论性。在所有十八个审计的遮蔽系统中，数据包干预在符合条件的案例上遵循了预测的中间值变化；这些有限的、以成功为条件的审计并未确立中介效应。结果确认了所测试遮蔽机制的巨大优势，同时其更精细的归因和普适性仍待探索。协议、结果和检查点均已公开。
英文摘要
Restricting what a module can read may improve what a system learns to compute. We test this in a preregistered confirmation with sixty four-cell systems sharing a frozen language-model backbone and communicating through learned continuous packets. Five conditions vary evidence masking, ownership markers, and replacement of foreign evidence with neutral filler (task-irrelevant text of the same token length), across six initialization clusters, each with two data orders, on one fresh task world. With markers available in both regimes, masking improved accuracy on held-out two- and three-operation compositions by median paired differences of 0.846 and 0.859; all twelve pairs cleared the required margins, and the full preregistered behavioral criterion passed. The unmarked replication also passed. No marked global-visibility (G+) system passed the marker-following check, so the effect of usable role information remains unresolved. The filler condition yielded seven full generalizers, but its decomposition criteria were inconclusive. Packet interventions in all eighteen audited masked systems followed the predicted intermediate-value changes on eligible cases; these finite, success-conditioned audits do not establish mediation. The results confirm a large advantage of the tested masking regime, while leaving its finer attribution and generality open. Protocols, results, and checkpoints are public.
Comments15 pages, 2 figures, 5 tables. Preregistered fresh-world confirmation of arXiv:2608.20054; related companion study: arXiv:2609.11365. Revised exposition, statistical and audit clarifications, AI-use disclosure, and table layout; numerical results unchanged
