NObSP: Functional Decomposition of Neural Networks via Oblique Subspace Projections
本文提出NObSP方法,通过斜投影分解神经网络预测,解决深度神经网络决策解释问题,支持局部和全局功能分析。
本文提出NObSP方法,通过斜投影分解神经网络预测,解决深度神经网络决策解释问题,支持局部和全局功能分析。
研究通过审计三种AI平台对20个心理健康问题的引用来源,使用分类器分析15,942条引用,揭示了引用分布和平台偏好差异。
研究探讨了为何某些团体能保持独立判断而其他团体则陷入一致性陷阱,通过建立模型分析社会认可对质疑文化的影响及动态恢复机制。
This work addresses the high computational overhead and latency commonly incurred by local small language models in structured reasoning due to reliance on repeated sampling or multiple model invocations. The authors propose VFR-LLM, a novel neuro-symbolic framework that, for the first time, integrates finite-domain logic into local small models. By introducing a symbolic layer that type-checks and formalizes input problems while ensuring consistency and traceability, the approach delegates deterministic reasoning to a dedicated solver, requiring only a single model call. Evaluated on pure prioritization tasks, VFR-LLM achieves an accuracy of 0.983—substantially outperforming self-consistency methods (0.700)—and attains 0.933 accuracy on extended logical reasoning tasks compared to 0.283 for baseline approaches, demonstrating significantly reduced latency without compromising precision.
This study investigates how long-term seismic risk shapes national identity, distinguishing for the first time between its distributive (resource-dependence) and expressive (emotional attachment) dimensions. Integrating World Values Survey data with subnational seismic risk measures, the authors employ spatial matching, panel regressions, and quasi-experimental designs embedded within a social interaction theoretical framework. They find that residents in high-risk areas exhibit significantly stronger national pride, willingness to fight for their country, and in-group prioritization—but only in institutional contexts where the state and religion are highly aligned. Short-term seismic events show no general effect, influencing only older, less mobile populations. The findings reveal that persistent, unavoidable environmental threats shape national identity through mechanisms moderated by institutional and religious contexts.
本文提出NObSP方法,通过斜投影分解神经网络预测,解决深度神经网络决策解释问题,支持局部和全局功能分析。
研究通过审计三种AI平台对20个心理健康问题的引用来源,使用分类器分析15,942条引用,揭示了引用分布和平台偏好差异。
研究探讨了为何某些团体能保持独立判断而其他团体则陷入一致性陷阱,通过建立模型分析社会认可对质疑文化的影响及动态恢复机制。
This work addresses the high computational overhead and latency commonly incurred by local small language models in structured reasoning due to reliance on repeated sampling or multiple model invocations. The authors propose VFR-LLM, a novel neuro-symbolic framework that, for the first time, integrates finite-domain logic into local small models. By introducing a symbolic layer that type-checks and formalizes input problems while ensuring consistency and traceability, the approach delegates deterministic reasoning to a dedicated solver, requiring only a single model call. Evaluated on pure prioritization tasks, VFR-LLM achieves an accuracy of 0.983—substantially outperforming self-consistency methods (0.700)—and attains 0.933 accuracy on extended logical reasoning tasks compared to 0.283 for baseline approaches, demonstrating significantly reduced latency without compromising precision.
This study investigates how long-term seismic risk shapes national identity, distinguishing for the first time between its distributive (resource-dependence) and expressive (emotional attachment) dimensions. Integrating World Values Survey data with subnational seismic risk measures, the authors employ spatial matching, panel regressions, and quasi-experimental designs embedded within a social interaction theoretical framework. They find that residents in high-risk areas exhibit significantly stronger national pride, willingness to fight for their country, and in-group prioritization—but only in institutional contexts where the state and religion are highly aligned. Short-term seismic events show no general effect, influencing only older, less mobile populations. The findings reveal that persistent, unavoidable environmental threats shape national identity through mechanisms moderated by institutional and religious contexts.