内部动作图的校准测试：无全局仿射闭包的状态信号
A Calibrated Test of Internal Action Maps: State Signals Without Global Affine Closure
浏览论文内容
中文总结 AI 辅助
该研究测试无全局仿射闭包时内部动作图的校准情况，以Qwen/Qwen3-4B为对象发现状态可用性等可分离，仅h28/h36有因果效应，重拟合未通过组合测试。
中文摘要 AI 辅助
隐藏状态信号可以是可解码的或具有因果可用性的，而无需支持可复用的动作图。我们测试在无源情况下拟合的动作图是否能达到其自然的动作后激活状态并进行组合。我们将这些测试组织为证据格，并在已知的仿射S₅载体上验证几何分支：所有保留源的折叠均通过单步、组合、逆、解码和交换性门控。结构化曲率和保留域共轭性会使误差单调上升，但仅23/30个最强单元翻转了闭包门控，从而对校准进行了有界而非普适性的限定。在预训练后的Qwen/Qwen3-4B中，冻结的最终标记h28仿射图的保留实体误差均值为.519，而测试域内交叉拟合的误差为.398。7次随机实体拆分和图几何不支持纯粹的实体特异性解释。更早的h4/h16层更适合拟合单步转换，但h4的冲突状态解码能力较弱，且词汇控制问题仍未解决。从一个冻结检查点再生的3个匹配干预数据集仅在h28/h36处显示出因果效应。结果感知的重拟合将h28单步误差降至.474（加权后为.469），但无任何重拟合通过组合测试。学习到的有限世界同样保留了相对代数信号或共享图表，而无需保留源的仿射闭包。在测试的载体中，状态可用性、因果使用、局部几何和可复用闭包是可分离的。该结果仅适用于一个预训练模型、采样的最终标记层、两个有限世界以及测试的仿射或诊断函数类。
英文摘要
A hidden state signal can be decodable or causally usable without supporting a reusable action map. We test whether action maps fitted without a source reach its natural post-action activation and compose. We organize the tests as an evidence lattice and validate the geometric branch on a known affine S_5 carrier: all held-source folds pass one-step, composition, inverse, decoding, and commutativity gates. Structured curvature and held-domain conjugacy raise error monotonically, but only 23/30 strongest cells flip a closure gate, bounding rather than universalizing calibration. In post-trained Qwen/Qwen3-4B, frozen final-token h28 affine maps have mean held-entity error .519, versus .398 for within-test-domain cross-fit. Seven randomized entity splits and map geometry do not support a purely entity-specific account. Earlier h4/h16 layers fit one-step transitions better, but h4 conflict-state decoding is weak and lexical controls remain unresolved. Three matched intervention datasets regenerated from one frozen checkpoint show causal effects only at h28/h36. Outcome-aware refitting improves h28 one-step error to .474 (.469 with weighting), yet no refit passes composition. Learned finite worlds likewise preserve relative algebraic signals or shared charts without held-source affine closure. Within the tested carriers, state availability, causal use, local geometry, and reusable closure are separable. The result is limited to one pretrained model, sampled final-token layers, two finite worlds, and the tested affine or diagnostic function classes.
发表机构
- Zhejiang University（浙江大学）
机构由 AI 辅助整理，请以论文原文为准。
